The options market also showed no signs of panic, with 7-day and 30-day at-the-money implied volatility at 28.8 and 32.6 respectively, standing at the lower end of the recent range; the 7-day 25-delta risk reversal quickly narrowed from -7.39 to -2.10, indicating a significant alleviation of short-end downside skew.

On the macro level, the U.S. July ISM Manufacturing PMI rose to 55.6 (a new high in over four years), but employment data softened, with ADP private employment adding only 44,000 jobs; the market awaits the non-farm payroll data to be released later that day (The Wall Street Journal expects an increase of approximately 83,000 jobs). In addition, the situation in the Strait of Hormuz has not been fully resolved, with Brent crude oil returning above $83; yen intervention and Japanese government bond yield trends continue to drive global liquidity expectations.

Previously, OpenAI, Anthropic, and Meta successively disclosed that their experimental models accidentally gained internet access permissions in test environments and intruded into external systems. Hinton pointed out that these events highlight the growing complexity of frontier AI models and warned of a surge in malicious cyberattacks in the future. He estimated that the probability of advanced AI posing an existential threat to humans is between 10% and 20%, and urged the industry to take potential risks seriously now, rather than waiting for problems to arise before responding.

The report states that industry giants expected to attend include global mining giant Rio Tinto Group, Australian BHP Group, U.S. Freeport-McMoRan Inc., U.S. Mountain Pass Materials Company, U.S. Rare Earth Company, U.S. Energy Fuels Inc., and Canadian Metal Company, among others. According to sources, the Trump administration plans to announce multiple deals and memorandums of understanding. (Jin10)

In Galaxy's Q2 financial report, two business engines are operating in different ways.

One follows the ups and downs of the crypto market, with falling coin prices continuing to drag down book performance; the other extends upward from the data halls in Texas, with servers racked, power delivered, and rent recognized gradually with capacity.

Loss Narrowed by $131 Million, Proprietary Investment Portfolio Remains Main Drag

On August 5, Galaxy Digital announced its Q2 2026 results. The company's net loss for the quarter was $85 million, a reduction in loss of $131 million compared to Q1's $216 million; adjusted gross profit was $43 million, compared to a loss of $88 million in Q1. Adjusted EBITDA was a loss of $77 million, compared to a loss of $188 million in Q1.

Simply put, adjusted EBITDA is an operating metric derived from net profit by excluding interest, taxes, depreciation and amortization, equity incentives, and certain one-off items, primarily used to compare business performance across different quarters.

Breaking down by business segment, the digital assets and data centers operating businesses contributed a combined $86 million in adjusted gross profit, with combined adjusted EBITDA of $1 million. Among them, the digital assets business achieved $66 million in adjusted gross profit, up 34% quarter-over-quarter, with adjusted EBITDA at a loss of $11 million; data centers achieved $20 million in adjusted gross profit and $11 million in adjusted EBITDA.

What kept Galaxy's overall adjusted EBITDA in the loss zone was mainly the company's own investment portfolio. The financial report classifies this part under the "Treasury and Corporate Affairs" (Treasury & Corporate) segment. This can be understood as a basket of assets held by Galaxy using its own funds, including digital asset spot, derivatives and ETFs, as well as listed company stocks, venture capital, private equity and fund investments.

In Q2, this segment recorded an adjusted gross loss of $42 million and an adjusted EBITDA loss of $78 million. Galaxy stated that the main reason was unrealized losses on related digital assets and investment positions. The digital assets and data centers operating businesses contributed a combined $1 million in adjusted EBITDA; after adding this segment, Galaxy's overall adjusted EBITDA turned to a loss of $77 million.

As of June 30, the net exposure of this investment portfolio was $1.16 billion. Among them, venture capital and fund investments were $606 million, the largest item; Bitcoin exposure was $400 million; other token exposure was $76 million; Solana exposure was $58 million; other liquid investments were $19 million. The financial report specifically noted that Bitcoin and Solana exposures include not only spot but also derivatives, shorts and other hedging positions, wrapped tokens and related investment vehicles, so the above amounts cannot be directly understood as spot coin holdings.

In terms of the balance sheet, as of June 30, Galaxy's total assets increased from $9.992 billion to $10.844 billion, up 9% quarter-over-quarter; total equity decreased from $2.779 billion to $2.720 billion; cash and stablecoins decreased from $2.605 billion to $2.459 billion. During the same period, the aforementioned net exposure to digital assets and investments decreased from $1.362 billion to $1.160 billion. These are changes in ending balances and belong to different financial metrics compared to the net loss for the quarter.

Trading Volume Declines, But Adjusted Gross Profit of Digital Assets Business Rebounds

An interesting contrast also emerged within the digital assets business.

Adjusted gross profit for global markets business increased from $31 million to $49 million, up 58% quarter-over-quarter; the number of counterparties increased from 1,691 to 1,741, with average loan size remaining largely flat at approximately $1.4 billion. The company disclosed that its trading volume decreased 7% quarter-over-quarter, and stated that industry trading volume declined by double digits during the same period. What can be confirmed here is that Galaxy maintained its relative share in a weaker market, but this cannot yet be used to infer that long-term profitability has stabilized.

Data for asset management and infrastructure solutions was more affected by coin prices. Galaxy disclosed that at the end of Q2, assets under management and staked assets totaled approximately $7.1 billion, down 12% quarter-over-quarter, mainly due to the decline in digital asset prices. Among them, ETF-related assets were $1.805 billion, alternative assets were $2.553 billion, and staked assets were $2.790 billion.

133MW Starts Generating Rent, Helios Moves from Construction Site to Income Statement

The most substantive change in Q2 occurred in West Texas. Galaxy has delivered Helios Phase I's total 200MW power to CoreWeave, corresponding to 133MW of critical IT load, and completed Phase I delivery as planned. Rent was recognized gradually with delivered capacity in Q2, making the data centers segment a revenue-generating operating business for the first time

Aerial view of Galaxy Helios Data Center Campus

After completion of delivery, Galaxy expects Helios Phase I to generate approximately $80 million in lease revenue per quarter starting from Q3 2026, with project-level adjusted EBITDA margin exceeding 90%. The base term of the CoreWeave lease is 15 years; the total signed critical IT load for all three phases is 526MW.

The company expects that over the entire lease term, the average annual revenue from this portion of leases will exceed $1.2 billion, and the average lease-level adjusted EBITDA margin is expected to exceed 90%.

This business is also more capital intensive. Q2 data center capital expenditures were $448 million, higher than Q1's $354 million. As of the end of the quarter, total liabilities classified under the data centers business in the financial report increased from $1.33 billion to $1.548 billion.

On July 28, a project company under Galaxy completed a $3.5 billion private placement of senior secured notes maturing in 2031, and the proceeds will be used for Helios Phase II construction. Phase II plans to add 260MW of critical IT capacity, with data halls expected to start delivery from Q2 2027.

As for the most prominent "over 5.7GW" in the financial report, qualifiers must be added. It refers to the potential power pipeline Galaxy has arranged in Texas, not capacity that is already powered or leased out. Helios currently has approved power capacity exceeding 1.6GW, with two additional 1GW load applications in the ERCOT interconnection process; the newly acquired Merlin, Caspian, and Selene three campuses have potential capacities of approximately 500MW, 700MW, and 900MW respectively, among which Merlin's initial agreement supports approximately 74MW.

From Institutional OTC Prediction Markets to On-Chain Funding Rate Products, What Does Galaxy Want to Capture?

If Helios provides longer-term lease contracts, Galaxy's digital assets business is attempting to turn institutional services into reusable products.

In Q2 and after the quarter, Galaxy launched institutional OTC prediction market trading, on-chain funding rate products (GOFR), and Galaxy Curator successively.

Among them, the on-chain funding rate product GOFR integrates floating rates from on-chain lending markets such as Aave, Morpho, Spark, and Kamino into a dynamically rebalanced funding rate; customers face Galaxy directly, with Galaxy handling wallets, smart contract execution and collateral monitoring. The company committed to investing up to $100 million of its own capital as first-loss protection, but the scope of protection is subject to specific terms.

Galaxy Curator builds institutional vault strategies based on Morpho and distributes them through Fireblocks Earn, allowing institutions to access on-chain yield products within existing approval, signing and strategy control processes. They expanded Galaxy's service boundaries, but did not disclose revenue separately in this quarter's financial report, so a more accurate description is "product capability expansion" rather than "growth curve has been validated".

Compliance channels are also advancing simultaneously. In May, the New York State Department of Financial Services granted BitLicense and money transmitter licenses to GalaxyOne Prime NY, allowing it to provide regulated digital asset trading and custody services to institutions in New York State.

In August, Galaxy also announced a multi-year cooperation with The Bank of New York Mellon (BNY), planning to provide staking support for BNY's digital asset custody platform and participate in platform infrastructure construction as a design partner.

Mike Novogratz summarized Galaxy's strategy in the concurrent CEO letter as the convergence of two forces: financial activities migrating on-chain, and the development of artificial intelligence continuously pushing up demand for power, land and data centers. This is management's explanation of the company's direction, not the financial results themselves. The information truly landing on the statements in Q2 is more plain: crypto asset prices can still significantly affect profits and losses, the resilience of digital assets operating business has improved, and data centers began contributing positive adjusted EBITDA for the first time.

The tests facing Galaxy have thus become concrete. On-chain products need to turn institutional cooperation into recurring revenue, and the Texas campuses need to turn potential GW into billable MW according to milestones. In the financial report, on one side is the coin price curve that will still fluctuate, on the other side is the electricity meter that has already started turning. The weight of the next phase depends on whether the latter can light up one by one on time and on budget.

The US July Non-Farm Payrolls report will be released tonight Beijing time. The market consensus expects approximately 80,000 new jobs, but multiple forward-looking indicators are sending mixed signals. Some institutions have provided forecasts far below the consensus. Whether the "Weak July" curse can be broken has become the biggest suspense in the current market.

The market expectation range is unusually wide, varying from a high of 157,000 to a low of 40,000. Goldman Sachs predicts an increase of 75,000, slightly below the consensus; Vanguard gives an extremely low forecast of only 18,000, believing that this spring's employment data was artificially inflated by factors such as weather, World Cup staffing, and early hiring by local governments, facing significant correction pressure in July. Meanwhile, ADP private sector employment data increased by only 44,000, significantly missing expectations, further exacerbating concerns about downside risks.

For the Federal Reserve, the current policy focus has clearly shifted to inflation rather than employment. Multiple officials have recently described the labor market as "stable." A strong employment report will reinforce expectations of "higher for longer" interest rates, thereby putting pressure on rate-sensitive assets; conversely, if the data is weak, it may push market pricing towards a mild rate cut direction.

"Weak July" Curse: Consistently Missing Expectations for Nearly Three Years

One of the backgrounds of this report that attracts the most market attention is the pattern of July employment data consistently disappointing in recent years.

According to Goldman Sachs research reports, over the past three years, the US July non-farm employment increase averaged 66,000 lower than the three-month average at that time, and averaged 35,000 lower than market consensus. These weaker-than-expected data were also accompanied by significant downward revisions of the previous two months' data, with an average revision magnitude of 112,000.

Goldman Sachs economists Ronnie Walker and Jessica Rindels listed this pattern as one of the core bases for downside risk in the report. Multiple alternative employment growth indicators they tracked averaged 65,000 in July, lower than June's 79,000.

In addition, Barclays analysts also warned that June employment data itself carries significant revision risk—the data was based on only about half the usual survey response rate, with the US Bureau of Labor Statistics (BLS) relying heavily on model estimates rather than actual reported data. Barclays expects the revision magnitude to be large, but the direction remains unclear.

World Cup Effect and Low Layoffs Provide Support

Not all signals point downward. Several data points provide phased support for the employment market.

The World Cup staffing effect is a significant plus in Goldman Sachs' forecast. Data from Homebase shows that during the survey reference weeks from June to July, employment growth in World Cup host cities was significantly faster than in other regions. Goldman Sachs estimates this effect could contribute about 10,000 to July non-farm payrolls, mainly concentrated in leisure and hospitality, professional business services, and trade and transportation industries. However, the same data also shows that this effect began to recede after the July reference period ended.

Layoff data also presents positive signals. Initial jobless claims in July fell to 210,000 during the BLS survey window, lower than June's 224,000; the week coinciding with the survey window was even lower at 188,000, the lowest level since September 1969. The number of layoffs announced by companies reported by Challenger, Gray & Christmas decreased by 12,000 month-over-month to 33,000 in July, the lowest since July 2024.

Government department hiring also shows signs of warming up. After contracting continuously for about a year and a half previously, government department employment added an average of 12,500 per month over the past four months, and government job vacancies have also rebounded recently.

Labor Force Participation Rate and Unemployment Rate: Potential Hidden Concerns

One of the cores of the employment report is the direction of the unemployment rate and the changes in the labor force participation rate behind it.

Goldman Sachs expects the July unemployment rate to rise slightly from 4.2% to 4.3%, higher than the consensus expectation of flat. Goldman Sachs believes this stems partly from a reversal of the significant decline in the June labor force participation rate—June's participation rate plummeted to 61.5%, the lowest since March 2021, and also the lowest level since June 1976 excluding the COVID-19 pandemic period; among them, the core working age (25 to 54 years old) participation rate recorded the largest single-month drop in history excluding April 2020.

Vanguard Group economists expect that as this batch of workers who exited the labor market seek jobs again, but the speed of finding jobs is slower than the willingness to return, the unemployment rate will face upward pressure, with their year-end unemployment rate forecast at 4.6%.

Citi economist Veronica Clark pointed out that the employment market currently presents a balanced state of "low hiring, low layoffs," a pattern particularly unfavorable for new job seekers. She expects the unemployment rate to rise above 4.5% within a few months, at which point market focus will shift back to rate cut expectations, with Citi's baseline scenario being a restart of rate cuts in the fourth quarter of this year.

Federal Reserve Stance: Inflation Priority, Employment Stability Secondary

The guiding significance of this non-farm data for monetary policy will mainly be reflected in whether it reinforces or loosens the baseline expectation of "higher for longer" interest rates.

Fed Governor Waller described the labor market as "robust and stable," Logan called it "robust and slightly improved," Schmid considered it "broadly balanced," Paulson and Hammack stated it has tended towards stability, Barkin was the most cautious in wording, calling the market "does not feel tight." Officials overall view inflation as a more urgent policy challenge than employment.

It is worth noting that Oxford Economics pointed out that even if July hourly wages rise by 0.4% month-over-month, the annual rate is only 3.6%, still consistent with the Federal Reserve's 2% inflation target, and wage pressure is currently not considered a significant inflation risk. According to Bloomberg, analysts believe a strong employment report could push up real yields, especially given Waller previously stated "the market has to some extent done some of the tightening work for the Federal Reserve."

Good News Turns Bad?

JPMorgan Chase's market intelligence department believes that this non-farm data will be traded on the logic of "good news is bad news"—a strong employment number will reinforce the pricing of "higher for longer" interest rates, pushing up rates and suppressing rate-sensitive sectors; if the data is moderately weak, it may drive yields down, market pricing moves slightly towards a dovish direction, and the equity market may react positively.

JPMorgan Chase provides the following detailed scenario analysis:

• If non-farm exceeds 150,000, the S&P 500 Index is expected to fall 50 to 175 basis points, with a probability of 10%;
• If non-farm is between 100,000 and 150,000, the index falls 50 basis points to rises 25 basis points, probability 25%;
• If non-farm is between 60,000 and 100,000, the index falls 25 basis points to rises 50 basis points, probability 30%;
• If non-farm is between 20,000 and 60,000, the index rises 25 to 75 basis points, probability 25%;
• If non-farm is below 20,000, the index falls 125 basis points to rises 50 basis points, probability 10%.

The options market's pricing of this non-farm data is relatively restrained, with implied volatility for contracts expiring on August 7 at only about 0.7%, reflecting that the market has partially digested uncertainty against the background of eased geopolitical tensions previously. As of the midday session on August 6, the 2-year US Treasury yield fell from a recent high of 4.35% to about 4.24%.

Mai Leduc Blount, Head of Product at Bridge, stated that these approvals will allow EU businesses to build stablecoin and payment products within a regulatory framework. According to the latest update from the European Securities and Markets Authority (ESMA), Bridge's addition brings the total number of authorized Electronic Money Token (EMT) issuers in the EU to 42; during the same period, three entities from Germany newly obtained CASP authorization, raising the total number of authorized CASPs in the EU to 324.

Binance is truly angry this time.

Searching through public records, the Binance system rarely sues partners for breach of commercial cooperation agreements.

Its most famous lawsuit occurred seven or eight years ago, a financing dispute with Sequoia Capital.

This time, it has aimed its spear at a U Card payment company called RedotPay, which has cooperated with it twice and is preparing to ring the bell in the US, and directly sued the three founders personally.

The claimed amount is $472.8 million, roughly equivalent to two and a half years of revenue for this payment company.

The timing of Binance's move is also subtle.

At this time, RedotPay is considering going public with a valuation of about $4 billion, while also negotiating a new round of financing.

At 11:30 noon today, the Supreme Court of Singapore held a procedural meeting that could be concluded in fifteen minutes.

Forty-eight hours before this hearing, Bloomberg obtained a Hong Kong court document: three entities affiliated with Binance sued RedotPay's three co-founders: Gao Zhangpeng, who is the company's current CEO; Chan Wa Choi; and Yao Chao.

In the complaint, Binance used a very heavy word: fraud scheme.

The allegations are: more than 470,000 users originally belonging to Binance Card were directed to RedotPay; funds from Binance Pay were not isolated as agreed, but instead were allowed, or even encouraged, to be used to top up RedotPay cards.

Binance calculated the claim of approximately $472.8 million based on a lifetime value of $925 per user.

Binance's Most Troublesome Year

The starting point of the story between RedotPay and Binance was three years ago.

2023 was Binance's most troublesome year.

In June, the US SEC sued Binance and Zhao Changpeng.

In July, Visa stopped issuing new Binance co-branded cards in Europe.

In August, Mastercard announced the termination of cooperation with Binance Card, and some users in South America were notified that their cards could no longer be used.

In October, Binance issued another announcement stating that Visa debit card services in the European region would be terminated.

But Binance provided an alternative solution in the announcement: switch to Binance Pay.

This was a forced strategic turn for Binance.

Binance Card was gone, but the demand for users to spend crypto assets did not disappear. Binance still had a payment channel, Binance Pay, but this channel still needed polishing, especially in operations.

At this time, a company called RedotPay was also established in Shenzhen and Hong Kong.

RedotPay's legal entity is called Red Dot Technology Limited, registered in Hong Kong in 2023. Its main product is a Visa card plus a stablecoin wallet: users top up USDT or BTC into the App, and can swipe cards, withdraw cash, or bind Apple Pay on the Visa network, with the system automatically completing the settlement from crypto assets to fiat currency in the background.

Its App was launched in August 2023, and physical cards were issued in October.

In the Chinese community, it is more commonly called "Little Red Card".

The person standing behind it is Yuan Dawei.

This early co-founder of Huobi left Huobi in 2015 and founded CoolWallet in Beijing in 2016, one of the first people in China to work on digital asset hardware storage; according to industry reports, RedotPay was initially invested and incubated by him.

He started researching Bitcoin in 2010, and the trust accumulated in the early Bitcoin holder and miner circles was the resource this new company could initially mobilize.

Therefore, the explanation in the industry at the time for "RedotPay and Binance cooperating" was very simple: Yuan Dawei had a good relationship with the Binance system, so he got the initial contact and cooperation.

After all, initially, the company RedotPay itself was smaller than outsiders imagined.

"Because RedotPay was initially a project with very few people itself, and the relationships inside were quite complicated," a source told BlockBeats. "Mr. Yuan should just be a shareholder, not really involved much in actual operations. There were very few full-time employees at RedotPay in the early days. The technical product side were not full-time people from RedotPay, but part-time."

On the card issuance side, RedotPay's initial cards were issued in cooperation with Hong Kong card issuance service provider Reap. "Generally speaking, it was packaged as a way of issuing cards for enterprises," the source explained.

Reap is a Hong Kong fintech company holding Visa principal membership, and its main business is white-label card issuance and cross-border payment for other enterprises.

According to his inference, the group of people who founded RedotPay should also be operating another entity at the same time, using the identity of this operating company to do operations and promotion of payment business for Binance Pay, similar to a contractor.

That company may have no relationship with RedotPay in terms of equity, "it's just that some actual controllers behind them overlap".

In other words, in the summer of 2023, RedotPay had a usable card, a team with almost no people, and a volume that kept failing to take off.

"There was no volume at the beginning," he said. "I know quite clearly that their first wave of significant volume came from Binance."

The First Fallout Between Binance and RedotPay

In November 2023, a Binance affiliate signed the first cooperation agreement with RedotPay.

This is the time point mentioned in the Hong Kong complaint.

The announcement later deleted by RedotPay but left by web archives described this relationship unequivocally: users open the Binance App, enter the mini-program, select RedotPay, enter the amount, click "Pay with Binance Pay", and get a top-up redemption code.

The announcement said this path compressed top-up time from minutes to seconds.

The fees were listed clearly: Binance Pay charges 1%, and the RedotPay mini-program charges 1%.

The announcement also specifically introduced a function called Lightning Deposit, allowing Binance users to directly deposit USD stablecoins from their accounts into RedotPay.

Cooperation announcement released by RedotPay in 2024

For RedotPay, this channel saved the most expensive part of customer acquisition: persuading a stranger to download the App, register, complete KYC, and deposit funds.

It no longer needed to find users one by one; the users were right at the door.

But this first agreement did not survive half a year.

Public reports citing the complaint stated that it broke down less than six months after signing, with the dispute point being that funds from Binance Pay were used to top up prepaid RedotPay cards.

Here appeared the most glaring contradiction in the entire case: RedotPay's public announcement at the time stated that direct card top-up was possible.

For the same path, on one side was the product selling point on the official website, and on the other was the overstep in Binance's eyes.

Whether the promotion ran ahead of the contract, whether certain specific paths were non-compliant, or whether the two parties had different understandings of "top-up account" and "top-up consumable card", we have no other information to judge before the original contract is made public.

But this fallout did not stop RedotPay.

In 2024, in documents for Series A financing, RedotPay still listed cooperation with Binance as a factor for "accelerating user adoption", and explicitly stated that users could deposit directly from Binance Pay to RedotPay cards.

These were all told to investors as RedotPay's growth story.

Two Years, $1 Billion Valuation

Perhaps the relationship was strong enough, or perhaps "connections were properly managed".

In March 2025, Binance and RedotPay sat back at the negotiation table.

They initiated the second cooperation agreement.

According to public reports recounting the complaint, the core of the second agreement was: funds from Binance could be used to exchange for fiat currency, transfer within the RedotPay App, and purchase RedotPay's own products, but could not directly become consumable balance on RedotPay cards, and must be isolated and labeled in accounting.

Simply put: Binance's money could still enter RedotPay's door, but could not exit RedotPay's door.

Why did Binance fixate on this step? Because account balance and card balance are two different worlds.

Account balance is the endpoint of a transfer; card balance is the starting point of a consumption relationship.

Once users put money into the card wallet, they will bind Apple Pay, buy coffee, pay subscriptions, and pay phone bills; their daily cash flow will grow on this platform from then on.

This is also where U Cards truly make money: exchange rate spreads, card swipe fees, balance retention, not that small card opening fee.

So-called fund isolation is not putting money into different drawers, but labeling it with a source tag in the system: this money comes from Binance Pay, can only follow the few paths allowed by the contract, cannot be commingled with ordinary balance, and even less automatically become swipable balance.

After the second cooperation began, RedotPay's financing pace began to accelerate.

Also in March 2025, RedotPay announced the completion of a $40 million Series A financing, led by Lightspeed, with participation from HSG and Galaxy Ventures.

In September, Coinbase Ventures entered, adding $47 million in strategic financing, and the company's valuation stood at $1 billion for the first time, officially becoming a unicorn.

In December, Goodwater Capital led a $107 million Series B, with Pantera Capital, Blockchain Capital, and Circle Ventures following, and HSG continuing to increase investment.

Within one year, RedotPay raised $194 million.

The operating data in the same financing materials were equally steep: as of November 2025, registered users exceeded 6 million, covering more than 100 markets; annualized payment volume exceeded $10 billion, and annualized revenue exceeded $150 million.

Annual payment volume tripled, with 3 million new users added.

A company that only had a few full-time employees two years ago and relied on others to issue cards, just like that, reached a valuation of $1 billion.

Mass Emails, "Prying Away" Binance Customers

Binance's allegation is: more than 470,000 users originally belonging to Binance Card were directed to RedotPay.

How did RedotPay "pry away" users? We did not see a reasonable explanation on various channels.

However, a source pointed out to us that the most critical move, and also the most primitive move, for RedotPay in the business war was: sending emails.

According to the complaint, Binance discovered the problem in March 2026: RedotPay was still allowing and encouraging unisolated Binance Pay funds to be used for prohibited purposes, including topping up RedotPay cards.

Then on April 3, Binance stopped relevant support, giving the reason externally as reviewing merchant partners.

Then, four months later, Binance submitted a complaint to the Hong Kong court.

But in the source's memory, Binance might have discovered problems with RedotPay in 2023, during the time when Binance Card had problems and cards in the hands of many users could not be swiped.

"The few people doing operations for Binance Pay should be the founding team of RedotPay. Those few people had Binance customer emails, so they basically just sent emails directly to customers, directing traffic to RedotPay," according to this source.

"This is definitely not normal." When asked if this counts as standard operation in the industry, he answered quickly, "If it was not explicitly agreed by Binance, then it is definitely not normal business behavior."

But immediately, this source also stated: "Speaking of this matter, I think it is unlikely that Binance knew nothing about it at the time."

He believes: An exchange willing to hand over this operation and the permission to access user data to an external team itself indicates the existence of a very deep personal relationship.

The more likely scenario is that Binance's core executives would not ask about this business line with insignificant volume at all, and the person specifically responsible for this project "turned a blind eye and tacitly approved, or even had verbal tacit approval".

Revenge, on the Road to RedotPay's Bell Ringing

In February 2026, Bloomberg reported that RedotPay is cooperating with JPMorgan, Goldman Sachs, and Jefferies to prepare for a New York listing, with a fundraising scale possibly exceeding $1 billion, valuation exceeding $4 billion, and could be realized as early as this year.

This will be one of the largest IPOs in the Asian stablecoin track this year.

But this road was not steady to begin with.

In March, Bloomberg reported again that RedotPay had lost at least five executives in the past 12 months, still had no CFO until the sprint for listing, and was simultaneously seeking up to $150 million in new financing.

A company with no CFO, executive turnover, and currently fundraising, received a bill equivalent to two and a half years of its own revenue at this moment.

Is this a coincidence?

"Actually, RedotPay should have had the ability to pay this one year ago," he guessed. Both parties most likely had been negotiating on this matter for a long time, but just couldn't reach an agreement, and now is the good timing for "revenge".

Because from RedotPay's perspective, for a payment company preparing to list in the US, the thing that should least appear in the prospectus is a cross-border lawsuit with no end in sight.

Underwriters will ask, investors will ask, regulators will ask.

In comparison, putting up a sum of money at once to exchange for a settlement agreement, turning the risk from "pending litigation" to "settled matter", is the fastest road to ringing the bell.

It seems that settlement should be a highly probable event.

"I feel the penalty amount Binance wants is a not too big, not too small amount," this source said. If the claim amount was opened to over $1 billion, the other party would directly stop negotiating and simply drag it out; but $470 million is a figure that "makes RedotPay bleed heavily, but is also a number they can grit their teeth and pay".

This is a quote designed for settlement.

The specific reserve ratios updated this time are BTC (103%), ETH (100%), TRX (106%), USDs (103%), HTX (102%), XRP (102%), DOGE (100%), SOL (102%). Users can view this on the HTX official website's "Assets - Proof of Reserves Report" page. In the future, HTX will continue to adhere to the 1:1 reserve principle, providing verifiable data, traceable on-chain assets, and regular information disclosure, allowing users to see the reserves, verify the assets, and use the platform with greater peace of mind.

Compiled by: AididiaoJP, Foresight News

Bitcoin and crypto traders haven't yet recovered from a massive attack worth about $100 million, which once ignited panic over a new round of price plummeting.

Since news of the hardware wallet Coldcard being attacked first broke, although the Bitcoin price has rebounded, it still hovers near recent lows. Traders are generally on edge, fearing another severe shock.

Against this backdrop, Bitcoin developers used AI tools to uncover nearly 5,000 security vulnerabilities across nearly 400 projects within just 24 hours. The situation was directly described as "extremely bad."

A team of Bitcoin developers composed of volunteers is conducting a large-scale, coordinated security audit. They have confirmed: the security status of the entire ecosystem is "extremely bad."

Within 24 hours, they scanned approximately 390 Bitcoin-related projects, cumulatively discovering 4,962 security vulnerabilities, including 85 critical-level vulnerabilities and 635 high-risk vulnerabilities. The vast majority of vulnerabilities have been verified by the project teams.

"We have grown to 16 people, distributed globally, working around the clock in shifts," Calle, an anonymous developer of the Cashu ecash protocol, wrote on X, "We are conducting a large-scale ecosystem security audit on the Bitcoin codebase."

The audit team is using Moonshot's Kimi K3 model—an open-weight artificial intelligence tool from China. Calle revealed that the team spends about $10,000 daily on computing power, a cost borne by OpenSats.

"We have been working day and night," Rob Hamilton, CEO of Bitcoin insurance company AnchorWatch and one of the audit members, also stated on X, adding that the team has discovered some "critical issues."

The efficiency of this audit is astonishingly high. Some developers claim that, on average, a critical vulnerability is uncovered almost every hour. AI is becoming an accelerator for both defenders and attackers—this has already begun to show in the recent Coldcard incident.

Over the past year, the Bitcoin price itself has fallen significantly, and the market was already highly sensitive to further declines. The sudden outbreak of hardware wallet security incidents has pushed the question of "whether self-custody is truly safe" back to the forefront.

Last week, Coldcard Bitcoin hardware wallets were exploited, and nearly 2,000 Bitcoin (worth just over $100 million) were drained from over 5,200 addresses in just a few days. The attackers exploited a key generation flaw that had existed for five years.

The Coldcard team has urgently called on users to transfer funds and repeatedly requested everyone on social media to "help spread the word."

"Please treat this as an emergency," the official Coldcard account wrote, "Migrate your funds immediately. Operate according to the recommendations for your device model, upgrade the device, generate a new seed, and carefully transfer funds... The threat is ongoing."

A wallet address associated with the hackers currently still holds about $36 million worth of Bitcoin, the vast majority of which is believed to be stolen proceeds. Since the incident was exposed, the address has successively received multiple transfers, some of which included messages attached via Bitcoin's OP_RETURN function.

Someone left a message stating: "I wash BTC, do KYC and cash out. I charge 10%." This was interpreted as money laundering solicitation, attempting to turn the hackers into clients. More messages were direct pleas to return the stolen Bitcoin.

On-chain analysts pointed out that the vulnerability has been made public with extremely high attention, and frontier large models are accessible to almost everyone, which means multiple hacker teams may already be simultaneously researching how to expand the results. "You are racing against time."

Cobra, another anonymous co-owner of Bitcoin.org, stated bluntly that he has a "very bad feeling"—AI has likely already been involved in this incident where Coldcard funds were drained.

This AI-driven vulnerability scanning, along with the previous massive theft from Coldcard, is pushing the security issues of the Bitcoin ecosystem to a new critical point. While developers use AI to accelerate vulnerability discovery, attackers may also use the same tools to accelerate exploitation. The window left for remediation and migration in between is being compressed.

Currently, the Bitcoin price is still fluctuating at low levels, and traders are waiting for the next possible shock. And this audit, which burns $10,000 worth of computing power daily, may just be the beginning of a broader security screening.

On August 4, a bombshell quietly appeared on the forum.

Justin Drake, along with five core developers, submitted a proposal named EIP-8363: when the staking rate reaches 50%, the protocol stops issuing new rewards to validators. All consensus layer earnings are burned.

48 hours later is the deadline for the All Core Developers Meeting to finalize candidate proposals for the Hegotá upgrade.

The entire ecosystem has only two days to digest this monumental shift in monetary policy. The forum instantly exploded, with core members of Aave, Lido, and ether.fi opening fire collectively.

01 What Exactly Does This Proposal Want to Burn?

Ethereum currently has over 40 million ETH staked, accounting for 33% of the total supply, with a net inflow of 1.75 million per month. The rise of liquid staking tokens, the maturity of institutional staking infrastructure, and the compliance of spot ETFs—three forces have simultaneously tightened the spring of staking growth.

The problem is that the current issuance curve has no brakes. Even if all ETH in the world were staked, the base yield for validators would still be 1.5%. As long as someone feels this yield covers the risk, the total staking amount will expand infinitely.

Justin Drake believes 30 million is enough to secure the network, Vitalik even thinks 15 million is sufficient. The current 40 million is at least double what is needed.

EIP-8363's solution is an increasing burn curve: the higher the staking amount, the larger the proportion of received rewards that are burned. At a 50% staking rate, burn everything, net yield goes to zero. Afterwards, validators can only survive on transaction priority fees and MEV.

The proposal does not expect to actually stop at 50%. Validators still have to bear hardware costs, slashing risks, and liquidity lock-ups; the market will demand a positive premium, and the equilibrium point will naturally fall back to a lower position.

Simply put, this is an automatically tightening faucet.

02 Intended to Prevent Centralization, the First to Die Are Independent Stakers

Consensus layer issuance accounts for 93% of validators' total income. Cutting this off treats all nodes equally in formula, but in reality, it precisely targets small capital.

The cost of running a node is rigid: hardware, bandwidth, electricity, operations time. Large service providers spread these across tens of thousands of validators, with marginal costs approaching zero. Home independent stakers must bear all expenses with 32 ETH per node.

Independent stakers currently account for only 5.4% of the total network staking amount and are shrinking year by year. Comprehensive income fell from 2.86% to 1.48%, nearly halved; they will certainly be the first to fall below the break-even line. The last practitioners of Ethereum's decentralization spirit will be physically erased from an economic level.

Opponents used models to deduce a more ironic future: On-chain LST users are highly sensitive to yield and will run as soon as yields drop. But exchanges like Coinbase have extremely low operating costs and are bound to a large number of ETF clients insensitive to yield. In four years, Coinbase might independently control over 33% of the network staking amount. The centralization the proposal aims to prevent is instead accelerated.

More hidden damage comes from MEV.

Consensus issuance is cut, and the weight of MEV in total income is passively amplified. Model calculations show that when staking amount reaches 48 million, MEV proportion surges from 7% to 19%; at 54 million, it approaches 30%.

MEV distribution is extremely uneven. Independent nodes may receive nothing for months, while large mining pools easily smooth variance based on validator base.

Currently, most MEV-Boost relays in the market follow the OFAC sanctions list and have a censorship tendency. When base yields are abundant, nodes still have the confidence to insist on connecting to neutral relays, sacrificing part of profits for decentralization ideals.

But when MEV becomes the lifeline for survival, choosing censored relays is no longer a moral choice, but a rule of business survival.

The proposal aims to maintain network neutrality, but the result greatly raises the financial threshold for maintaining neutrality. It defeats the purpose.

There is also a tax trap. The transition period design is "first double the bookkeeping rewards then burn half". In the US, UK, and Germany, staking rewards count as taxable income upon receipt. The validators' taxable base doubles, while net received yield actually decreases instead of increases. Aave founder Stani Kulechov directly called out: This is forcing home nodes out of compliant regions at the protocol level.

03 The Foundation of DeFi Lego Is Loosening

Ethereum's staking yield is regarded as the "risk-free benchmark rate" for the entire DeFi. All lending protocols, liquidity pool strategies, and interest-bearing asset pricing are strictly anchored to this benchmark. Remove the cornerstone, and the entire Lego tower will be repriced.

LST protocols hold over $42 billion in assets, relying on extracting 10% of staking rewards to operate. Yields halve, protocol income is forcibly halved, and the ability to reinvest in security audits and infrastructure maintenance drops significantly. Once investors determine that 1% yield cannot compensate for smart contract and de-peg risks, massive capital will sell off LST to swap back for native ETH, starting a discount spiral.

The more direct impact is on looped leverage. Many institutions perform leverage looping on Aave: deposit stETH, borrow WETH, then swap for stETH to deposit again. Under E-Mode, leverage is set to over 10x.

The premise of this strategy is that staking yield is higher than borrowing rates. Base yield falls from 2.6% to 1.2%, while borrowing rates maintain 1.5% in the short term. The spread flips from positive 1.1 percentage points to negative 0.3 percentage points.

The money printing machine becomes a daily loss machine.

Collective deleveraging means massive selling of stETH. Liquidity pools dry up, collateral shrinks and breaches liquidation thresholds, cascade liquidations. The scene of stETH severely de-pegging during the 2022 Terra collapse may replay.

04 Solana Is Waiting on the Side

If ETH's on-chain yield is pushed to zero, while stablecoin yields maintain 4% to 5%, the rational operation is: use ETH as collateral to borrow, then buy high-yield stablecoins. ETH becomes the Japanese Yen of the zero-interest era, used specifically for borrowing.

Meanwhile, Solana native staking yield exceeds 5%, and the Alpenglow upgrade plan compresses block finality from 12.8 seconds to about 150 milliseconds. Wall Street is already applying for Solana ETFs with staking dividends.

Ethereum has cumulatively attracted tens of billions of dollars in institutional funds through ETFs. These funds value predictable, continuous cash flow; Wall Street gave it a name called "Internet Bond". EIP-8363 zeros out this "coupon" with a mechanism not controlled by holders. No institution is willing to underwrite a financial instrument where "coupons can be eliminated by others' actions at any time".

For an issuance cut of about $1 billion per year, the cost could be the evaporation of tens of billions of dollars in institutional net inflows.

Aave DAO representative Marc Zeller publicly called for Lido, Aave, and ether.fi to ally, threatening to "directly reject EIP-8363" if necessary. EIP-8148 author Greg Koumoutsos questioned: Is there a plan to stuff the proposal into the upgrade within a feedback window of less than 48 hours? Idealistic researchers and builders bearing billions in real money have collided head-on at Ethereum's governance table.

Ethereum indeed needs a staking brake mechanism. But before having a complete plan to deal with LST de-pegging, institutional capital outflow, and validator centralization, the damage caused by slamming on the brakes may be far greater than the problem it aims to solve.

Two days to digest a monetary revolution. This problem has no standard answer.

Translated by: Chopper, Foresight News

Upbit launched a promotional event from July 26 to August 9, waiving the 0.05% transaction fee for stablecoin trades listed in the KRW market. Within the same week, the platform quickly listed USD stablecoins such as RLUSD and USDG. These measures aim to boost stablecoin trading volume.

This is not the first time Upbit has listed stablecoins, but it is the first time the platform has concentrated on launching multiple stablecoin-related businesses in a very short period. This article will analyze the logic behind Upbit's current push into the stablecoin sector from three dimensions: the stablecoin market landscape of Korean exchanges, trends in cross-border capital inflows and outflows of stablecoins, and the current regulatory environment.

Reshaping the Stablecoin Market Landscape of Korean Exchanges

In January 2025, the Korean stablecoin trading market was a duopoly: Upbit accounted for 53.5%, Bithumb accounted for 42.5%, with the two combined holding over 95% market share. Just 18 months later, the industry landscape was completely rewritten. As of June 2026, Coinone ranked first with a daily average stablecoin trading volume of 84.58 billion KRW (34.8%); Bithumb followed closely with 75.57 billion KRW (31.1%); Upbit was at 73.02 billion KRW (30.1%), evolving the market into a tripartite competition.

The trigger for the landscape change was Coinone completely waiving USDC trading fees starting from October 2025. At that time, other competitors' fees remained in the 0.04%-0.20% range, while Coinone persisted with a zero-fee strategy. Affected by this, its market share climbed to 11.5% in March 2025, reaching 30.5% in December, surpassing Upbit's 29.7% for the first time. Trading demand highly sensitive to fees flowed heavily to Coinone, mainly including two categories: demand for capital outflow to participate in overseas derivatives trading, and demand for earning USD exchange rate returns.

This fully illustrates that stablecoin trading demand possesses extremely strong price elasticity. Regardless of which exchange users buy from, stablecoins themselves are homogeneous assets, and a large number of users withdraw coins to external wallets after purchasing. Therefore, the core elements distinguishing the competitiveness of each exchange are left to only fees and liquidity. In fact, a fee difference of merely 0.05 percentage points is enough to rewrite industry rankings.

It is worth noting that this landscape reshuffle occurred only in the stablecoin sector. In June 2026, in all-category cryptocurrency trading, Upbit's share was 60.0%, Bithumb was 32.0%, the two combined accounted for over 90%, while Coinone was only 6.2%. In other words, in the market where Upbit holds an absolute overall advantage, stablecoins are the segment where its shortcomings are most prominent.

At the same time, the overall market size is shrinking rapidly. In July 2026, the total daily average stablecoin trading volume of the five major Korean exchanges was $466.69 million, compared to $2.37 billion in January, a decrease of up to 80.3%. Dunamu (Upbit's parent company) had revenue of 234.6 billion KRW in the first quarter of 2026, down 55% year-over-year; operating profit was 88 billion KRW, down 78% year-over-year. In just six months, the market size shrank to one-fifth of its original size. In such an environment, sectors unaffected by bull/bear cycles and with stronger demand resilience have strategic value far higher than during market boom periods.

Stablecoins, the Core Carrier of Cross-Border Capital Flow in Korea

To understand the essential demand for stablecoins in Korea, one must track where the funds go after users buy stablecoins.

In June 2026 alone, stablecoin withdrawals from the five major Korean exchanges to overseas exchanges totaled 2.7625 trillion KRW, while stablecoins flowing back from overseas totaled 2.2022 trillion KRW, resulting in a net outflow of 560.3 billion KRW. Since statistics on this data began in January 2025, there has been a net outflow every month for 18 consecutive months, with the cumulative net outflow reaching approximately 14.9 trillion KRW.

This continuous net outflow forms a sharp contrast with the stock market. In the second quarter of 2026, net sales of Korean overseas stock investments were 1.6185 trillion KRW, while the stablecoin net outflow reached 1.6872 trillion KRW. Overseas stock capital flows will reverse and flow back to the local market with market conditions; but even during downward market cycles, stablecoins maintain a net capital outflow, sufficiently reflecting the special positioning of stablecoins in the Korean market.

The role of stablecoins as a channel for transferring capital to overseas exchanges and DeFi was not inherent, which makes this phenomenon even more worthy of study. The launch of stablecoins in the Korean KRW trading market was very late; among the five major exchanges, Upbit was the last to open the USDT KRW trading pair, officially launching only in 2024. Before this, users wanting to transfer funds to overseas exchanges could only buy volatile assets like Bitcoin or XRP and then transfer them, needing to bear the price fluctuation risk during the transfer process.

After the KRW trading pairs were listed, this cross-border function was quickly undertaken by stablecoins. The Financial Supervisory Service of Korea began specifically statistics on stablecoin cross-border transfer data from January 2025, which itself indicates that regulatory agencies already view stablecoins as the main tool for cross-border capital flow. As early as 2019, USDT's global trading volume had already surpassed Bitcoin; but because Korea lagged in listing KRW trading pairs, it equivalently completed this development process in a compressed manner after 2024.

Actual Effects of the Stablecoin Trading Expansion Strategy

Returning to Upbit's decision, this fee waiver is a limited-time event, ending on August 9; whereas Coinone, which grabbed a large market share, has implemented a permanent zero-fee policy since October 2025.

The effects of limited-time promotions have precedents for reference. Korbit once launched a USDC zero-fee and reward event from January to April 13, 2026; during the event, stablecoin market share reached 3.48%, but after the event ended, trading volume flowed back to Upbit and Bithumb. Referring to this historical case, after Upbit's promotional event ends, the stimulated trading volume is likely to dissipate as well. Below, combining data from the promotional period and normal times, we assess the subsequent trend.

Below is the comparison of daily average stablecoin trading volume during Upbit's promotion versus the month before the promotion: daily average trading volume 30 days before promotion was 46.96 billion KRW; after the promotional event started, daily average trading volume rose 162.0%, reaching 123.06 billion KRW. Excluding weekend factors and counting only weekday data, the increase was 170%, growing from 55.03 billion KRW to 148.69 billion KRW.

The policy took effect very quickly. On July 25, the day before the event, trading volume was 29.95 billion KRW; on the day the event launched (Sunday), it reached 72.19 billion KRW, doubling compared to the previous weekend; on the first working day, July 27, it surged to 162.27 billion KRW, touching a peak of 203.29 billion KRW on July 29.

Three key phenomena are exposed in the data:

• Almost all new trading volume came from USDT. USDT daily average trading volume rose from 46.06 billion KRW to 120.71 billion KRW, maintaining a 98.1% share in Upbit's total stablecoin trading volume, with almost no change before and after the event. During the same period, newly listed RLUSD (daily average 710 million KRW) and USDG (daily average 340 million KRW) accompanying the event contributed only about 1% to the overall increment, with popularity fleeting. RLUSD was 5.6 billion KRW on the listing day, then quickly fell back to around 100 million KRW daily average; USDG was 2.24 billion KRW on the listing day, then similarly shrank quickly. The heat brought by new coin listings lasted only one day. Meanwhile, long-tail stablecoins such as USD1, USDS, USDE, and gold-pegged assets XAUT, even enjoying full fee waivers, saw trading volume flat or even decline. Conclusion: This round of Upbit's stablecoin expansion strategy did not divert USDT trading traffic to other types of stablecoins.
• The event has not yet ended, but the stimulation effect has already started to decay. The first week's daily average trading volume was 148.65 billion KRW, falling back to 84.68 billion KRW in the second week. Excluding weekends and looking only at weekdays, trading volume decreased 33%, dropping from 163.944 billion KRW to 110.58 billion KRW. This belongs to the typical trend of rapid ebbing of heat in the early stage of an event.
• Exchange rates brought interfering variables. In July, the KRW strengthened; Upbit platform USDT price fell from 1,517 KRW on June 26 to 1,423 KRW on August 4, with the end of July exactly falling within the decline interval, coinciding with the promotion time. Exchange rate fluctuations themselves generate arbitrage and bottom-fishing demand, so part of the trading volume growth comes from exchange rate market conditions, not entirely the effect of fee waivers.

Comprehensively speaking, once the promotion ends and fees are restored, Upbit will find it difficult to retain the market share temporarily gained during the event. As long as Coinone's permanent zero-fee policy remains in effect, Upbit will face a dilemma in the future: either follow suit and permanently waive stablecoin fees, or sacrifice market share to preserve fee income.

Regulatory Institutionalization Becomes a Key Variable

I believe Upbit itself is also clear that the trading volume growth brought by fee waivers is only a short-term effect, and listing multiple new stablecoins cannot truly divert USDT trading traffic.

Calculations show that in the past 15 days, Upbit gave up approximately 1 billion KRW in fee income. Why invest resources in a low-margin sector and voluntarily waive fees? The changes happening in the Korean regulatory environment are an important clue.

First, the landing of KRW stablecoins is imminent. In the 2026 Economic Growth Strategy, the Korean government formally plans to promote the legislation of the "Digital Asset Basic Act" in the second half of the year. The bill is expected to include clauses such as the KRW stablecoin issuance licensing system, reserve asset requirements, and user redemption rights. The US GENIUS Act is expected to be fully implemented between late 2026 and early 2027, at which time USD stablecoins will usher in a new round of global popularization.

Second, changes in Dunamu's equity structure. In November 2025, Naver Financial passed a share swap plan, resolving to make Dunamu a wholly-owned subsidiary; both parties proposed building a payment ecosystem centered on stablecoins and digital wallets as an important synergy direction. From this perspective, Upbit's stablecoin trading volume, liquidity, and user base are not just sources of fee income, but also distribution channels for future payment businesses. Therefore, expanding the business base as much as possible at present has strategic significance.

Third, the contradiction lies in: regulatory laws may make it difficult for Dunamu to fully enjoy the dividends of KRW stablecoins. The current "Specific Financial Transaction Information Act" and "Virtual Asset User Protection Act" stipulate that virtual asset service providers shall not trade assets issued by related parties. After Dunamu is incorporated into the Naver Group, if a consortium led by Naver issues KRW stablecoins, there is a regulatory interpretation: Upbit may be restricted from listing that currency. Whether local KRW stablecoins can be listed in the future is still doubtful, so the opportunity that can be firmly grasped at present is to become the distribution hub for USD stablecoins in Korea.

On August 7, South Korea's largest cryptocurrency exchange Upbit officially announced that it will terminate trading support for BONK on September 7 at 15:00 (KST), involving the BONK/KRW and BONK/USDT trading pairs, with withdrawal services retained until October 7. Upbit stated that after evaluating BONK, it found security incidents such as hacker attacks of unknown cause or unresolved issues occurring in the distributed ledger used by the operators for issuing, transmitting, and storing virtual assets, as well as behaviors where the issuer or operator failed to disclose important matters regarding virtual assets in a timely manner through appropriate electronic transmission media. After comprehensive consideration, it was confirmed that there were many deficiencies, which could lead to user losses.

After the news was announced, the BONK price fell from $0.0000028 to $0.0000025, a drop of about 10%, and its current market cap is $222.26 million.

Foresight News previously reported specifically on this event in "4.4 Million Pries Away 20 Million: BONK Suffers a Legal Open Robbery". One month ago, BonkDAO just experienced a governance attack that shocked the industry, with about $20 million in treasury assets being "legally" transferred away.

BONK Attacker Cashed Out Approximately $13.58 Million

On June 30. The attacker submitted proposal BIP #76 via the Realms governance platform in the Solana ecosystem, titled "Sowellian BonkDAO". The proposal was superficially packaged as a governance optimization plan, but the core instruction was to directly transfer about 4.426 trillion BONK tokens from the BonkDAO treasury to an address controlled by the attacker. At that time, the BONK circulating supply was about 88 trillion, and the 1% voting threshold was about 880 billion. Between July 4 and 5, the attacker bought through exchanges such as Binance and Bybit, supplemented by some DeFi lending, accumulating about 882.285 billion BONK tokens, at a cost of about $4.4 million, just meeting the quorum requirement.

On July 6, the proposal entered voting. Only 7 addresses participated in the entire process, and the address controlled by the attacker contributed 99.878% of the affirmative votes. After the vote passed, the smart contract immediately executed the transfer automatically, and about 4.426 trillion BONK tokens (worth about $20 million at the time) were transferred from the treasury to the attacker's wallet. The entire process did not trigger any timelock, nor were there additional multisignature or manual review steps.

After obtaining the funds, the attacker acted quickly. Within about 9 hours after the transfer was completed, about $190,000 worth of BONK was transferred to OKX. The remaining about $19 million was transferred to a newly created multisignature wallet, which Chainalysis described as a "BONK 2.0" shadow DAO, jointly controlled by the malicious voting wallet, the fund receiving wallet, and a third-party address financially linked to the voting address.

At the same time, the attacker began to sell off the portion of BONK used to obtain voting rights. About 1 hour after the voting was completed, the relevant addresses began to liquidate positions worth about $5.3 million. In the following weeks, on-chain monitoring showed that the attacker continued to transfer funds to platforms such as Coinbase.

On July 17, the attacker transferred 1.186 trillion BONK tokens (worth about $4.11 million) into Binance.

On July 19, according to on-chain analyst Yu Jin's monitoring, the BONK treasury attacker transferred another 400 billion BONK tokens to Coinbase, worth about $1.11 million. Within 12 days since the relevant tokens were transferred out of the treasury, the BONK price fell from $0.0000047 to $0.0000027, a cumulative drop of about 41%.

On July 20, that attacker had completed the liquidation; 30 minutes prior, the last 400 billion BONK tokens in hand ($1.17 million) were also deposited into Coinbase.

Data statistics show that the attacker's cumulative cash-out scale was about $13.58 million.

High Difficulty, Fund Recovery Still Without Result

Bonk officials responded quickly after the incident, confirming that "BonkDAO encountered a malicious governance proposal, resulting in about $20 million worth of BONK being transferred out of the treasury". The team stated that they had identified the exchange wallet addresses used by the attacker to build positions in advance and notified law enforcement departments, while maintaining communication with exchanges, cross-chain bridges, and the Solana Foundation, attempting to recover funds and lock down the responsible parties.

On July 13, BonkDAO released a follow-up update: relevant wallets have been marked and are under continuous monitoring, and the team is exploring all possible recovery paths; emphasizing that the BONK token itself and user personal assets were not affected, and the token contract is secure; planning to release a formal post-incident analysis report subsequently, and the project team also called on the community to focus on governance mechanism improvements.

On July 23, officials further stated, confirming that recovery work is still ongoing.

As of now, no official announcement has been seen on public channels regarding large amounts of funds successfully returned to the treasury, nor has there been formal confirmation of large-scale funds being successfully frozen or recovered.

On July 7, some exchanges almost simultaneously listed BONK as a trading caution target. One month later, Upbit determined that the issues were not eliminated and formally decided to terminate trading support first. Since South Korea implemented the "Virtual Asset Investor Protection Act" in July 2024, DAXA (Korea Digital Asset Exchange Association) has strict statutory obligations for projects involving significant governance loopholes and security hazards. Upbit delisted BONK out of compliance risk avoidance and mandatory legal review.

Currently, other exchanges have not yet taken action temporarily.

It needs to be pointed out that since the transfer was executed automatically in full compliance with on-chain governance rules, the difficulty of recovery is significantly higher than traditional hacker incidents.

The attacker purchased sufficient voting rights (about 1% of supply), submitted a proposal, passed the vote, and then the smart contract automatically executed the transfer. The entire process was fully compliant with the governance rules of BonkDAO on Realms at the time. There was no private key leakage, no contract vulnerability, and no unauthorized call. Most jurisdictions remain conservative in their attitude towards pure governance attacks; courts are more inclined to view "code is law + vote passed" as valid internal decisions rather than theft in the traditional sense, which significantly weakens the strength of criminal filing and civil freezing.

From on-chain data, the attacker has also completed the sell-off and transfer of part of the assets, further compressing the scale of freezeable assets. In addition, the attacker's identity is not public, and factors such as high cross-border enforcement costs and long time periods make the probability of recovering most of the losses relatively low.

Governance Issues Urgently Need to Be Resolved

The core controversy of this event lies in: when the voting threshold is too low, the participation rate is extremely low (only about 2.9% of members participated), and there is a lack of timelock and abnormal proposal interception mechanisms, the attacker used $4.4 million to leverage $20 million in assets, the return on investment was extremely high, and the entire process was compliant.

Any DAO holding a large treasury, if the long-term voting participation rate is extremely low and the quorum is set very low, is equivalent to exposing the control of the treasury to the open market. Whoever can concentrate-buy the minimum threshold tokens in a short time may take away the treasury. This is not a theoretical risk, but an executable path that has been verified.

If the treasury sets an execution delay of 3-7 days or even longer, the community and project party at least have time to discover anomalies, initiate an emergency vote to veto, or intervene through multisig. In reality, many mature DAOs (especially those with large treasuries) will mandate timelocks for proposals involving the treasury.

When community activity is sluggish for a long time, relying solely on token weighting is essentially using "money" to decide governance, rather than "people". For projects with large treasury scales, the project party should consider introducing higher quorums and dual thresholds (requiring both a certain proportion of voting rights and a certain number of independent addresses). In addition, additional approval mechanisms can be added for large proposals, such as a multisig committee.

Governance security can never be completely solved by post-incident remediation; it must be considered based on the "worst-case scenario" at the design stage.

At 20:30 Beijing Time on Friday, the U.S. Bureau of Labor Statistics will release the July Non-Farm Payrolls report, which is expected to further confirm the "low hiring, low layoffs" pattern in the labor market.

The market's focus is no longer just on the number of new jobs, but on the labor force participation rate, wage growth, and which industries are mainly driving employment growth. These data will provide an important basis for the Federal Reserve to judge economic resilience and the future interest rate path.

July Employment May Continue to Grow Moderately, Paving the Way for Fed to Focus on Inflation

Economists expect that U.S. non-farm payrolls in July will increase by about 80,000, slightly higher than June's 57,000, but still far below normal levels in the past.

For comparison, in the decade before the pandemic and during the period from 2022 to 2024, the average monthly new jobs in the United States was usually close to 200,000.

Nowadays, U.S. companies are significantly slowing down their hiring pace.

On one hand, the Iran war has led to rising energy prices and pushed up corporate operating costs. Facing higher uncertainty, companies are more inclined to cut controllable costs, and hiring is precisely the part that is easiest to adjust.

On the other hand, the Trump administration's tightening of immigration policies has also reduced labor supply, making companies face more limited talent sources even if they wish to expand hiring.

This spring, U.S. employment briefly warmed up, with new jobs reaching 214,000 in March, but this momentum subsequently weakened significantly due to geopolitical conflicts, rising energy prices, and declining corporate confidence.

However, low hiring does not mean a collapse of the employment market. The unemployment rate in June dropped slightly to 4.2%, and market consensus expects this level to be maintained in July.

Data on Thursday showed that the scale of layoffs in the United States fell to a two-year low, and initial jobless claims were below 200,000 for the third consecutive week, one of the lowest levels in nearly 60 years.

Federal Reserve's Lisa Cook once stated that the current employment market has formed a balanced state of "low hiring, low layoffs." "Although the hiring rate is low, the unemployment rate remains stable because layoffs are also few," Cook said.

Compared to the unemployment rate, the change in the labor force participation rate is more worthy of attention. The labor force participation rate in July is expected to rebound slightly to 61.6%. Previous June data showed that the U.S. labor force participation rate dropped significantly to 61.5%, the lowest level since March 2021. If the impact of the pandemic is excluded, this is the lowest level since June 1976.

Among them, the participation rate of the core working-age population aged 25 to 54 also declined significantly, hitting the lowest level since December 2023 and recording the largest single-month drop since the pandemic.

Economists will closely observe whether this change is caused by seasonal factors and statistical fluctuations, or whether it means that deeper problems are emerging in the employment market.

In terms of wages, the market expects average hourly earnings in July to rise 0.3% month-on-month, with the year-on-year growth rate remaining around 3.5%. This growth rate is close to pre-pandemic levels and basically aligns with the Federal Reserve's 2% inflation target, showing no indication that wages are driving inflation up again.

Nela Richardson, Chief Economist at Automatic Data Processing (ADP), the largest wage data service provider in the United States, stated that labor costs are currently not the main source of U.S. inflation pressure.

At the same time, new jobs in the United States are becoming increasingly concentrated in a few industries. The healthcare industry has become the main pillar of employment growth this year, with hospitals, clinics, and related medical service institutions contributing more than half of the new jobs since 2026.

However, in a healthy expansion cycle, almost all industries will increase hiring, but the United States has not seen such a broad recovery for nearly two years currently.

Citibank economist Veronica Clark wrote in a letter to clients: "Moderate employment data, plus a high probability of significant downward revisions to previous data, should 'further refute the claim that the labor market is tightening and may become a source of inflationary pressure.'"

She added that a slowing labor market paves the way for Federal Reserve officials to focus on inflation data when making interest rate decisions.

Heather Long, Chief Economist at Navy Federal Credit Union, pointed out that it is reasonable for the Federal Reserve to focus on inflation currently, but it still needs to pay attention to whether the economy can create enough development opportunities for young people.

Fed to Keep Close Watch on Inflation, Gold Holding Firm at $4000 Awaiting Rebound?

Although the employment report is an important reference for the Federal Reserve to observe the economy, it may not change the policy direction in the short term.

Federal Reserve Chair Kevin Warsh previously stated that the U.S. employment market is currently in a "stable" state, and future policy focus will still depend on inflation trends.

Some officials even believe that if inflation remains high, the possibility of future interest rate hikes cannot be ruled out.

However, some market institutions have begun to bet that if the employment market weakens further, the Federal Reserve may turn to cutting interest rates again.

Citi economists believe that although current employment data can still be described as "stable", changes may occur in the coming months, with the unemployment rate expected to rise above 4.5% and push the Federal Reserve to reconsider cutting interest rates.

Citi expects that the Federal Reserve may resume cutting interest rates starting from the fourth quarter of this year and cut rates a cumulative three times before January 2027.

Vanguard Group believes that its 401(k) retirement account data suggests that new jobs in the United States in July may be only 18,000, showing significant weakness in the summer employment market, and warns that this weakness may continue into autumn.

Vanguard economists stated that the increase in labor force exit reflects weak hiring, especially affecting young workers. As some people who exited the labor market look for work again, while corporate hiring speed remains slow, the unemployment rate may face upward pressure in the future.

According to data from the CME FedWatch Tool, traders currently expect the probability of a Federal Reserve interest rate hike in September to be 55%, while this probability was 63% a week ago.

StoneX Senior Analyst Matt Simpson believes that regardless of how the non-farm data performs, $4000 has proven to be a solid bottom for gold, "I suspect longs are waiting for a price pullback to take the opportunity to adjust upwards to near $4600. In the short term, non-farm may bring some noise, but price trends have shown that gold seems eager to rebound."

Marex pointed out in a monthly report, "Entering August, our attitude towards gold has improved, and we expect the trading range will expand⁠."

Growth was mainly concentrated in yield-bearing products, including tokenized government bonds and multi-strategy funds (such as JTRSY, BUIDL, sUSDS), with Aave, Morpho, and Kamino providing the deepest liquidity. Meanwhile, on-chain RWA spot trading volume increased by approximately 220% year-on-year, while native crypto spot trading volume on decentralized exchanges fell by approximately 70%.

Bitcoin has been oscillating around 64K for over two months since May. The area above 65K is a dense trading zone from the beginning of the year's rally. Currently, hotspots are not in the crypto sector, making it difficult for speculative funds to flow in, and the trend may lean downward. The cryptocurrency market has undergone a 9-month bear market, and overall Implied Volatility (IV) has remained low for over a quarter. If there are no incremental capital flows in Q3, there are concerns regarding potential significant risk exposure in the market.

On August 7, OpenAI announced that ChatGPT's weekly active users have reached 1 billion, while making GPT-5.6 Luna available to free users with unlimited text conversations.

The day before, multiple DeepSeek API users received notice: The company plans to generally raise API service prices in the near future, "expected increase is significant," specific prices and execution time await further announcement.

This seems somewhat counterintuitive.

For the past two years, the narrative Chinese large model vendors are best at is extreme cost-performance ratio.

Previously, DeepSeek provided performance close to GPT-4 levels at less than one-thirtieth of OpenAI's price, known in the industry as the "Price Butcher," Liang Wenfeng became "Saint Liang," the consensus was: The core advantage of Chinese large models is to drive the usage cost of AI to the floor.

Now the floor has loosened, OpenAI has become the one giving things away for free to the world, while Chinese vendors are starting to talk about payment and price hikes. Between offense and defense, has the position really changed?

OpenAI: Free Is Not Charity

Let's clarify one thing first: GPT-5.6 Luna is not OpenAI's strongest model.

It is positioned in the medium capability range, more than enough for daily conversations, simple writing, and basic translation, but complex reasoning and multi-step code analysis still rely on the more advanced SOL series.

OpenAI takes a "good enough" model to cover the widest user scenarios.

This strategy has its backing, and also its costs.

The backing comes from the cost side.

Over the past 18 months, the unit Token cost curve for large model inference has been astonishingly steep: model architecture optimization, quantization technology maturity, and inference engine upgrades combined together, the same computing power cluster can serve dozens of times the request volume today compared to two years ago.

When marginal costs are low enough, free approaches the logic of Google Search: free entry, monetize the ecosystem.

The costs are equally clear.

In Q1 2026, OpenAI revenue was $5.7 billion, non-GAAP operating loss rate -122%, for every $1 earned it loses an additional $1.22, full year estimated net loss $14 billion.

Of the 1 billion weekly active users, 50 million are paid subscribers, a payment rate of about 5%.

Subscription fees obviously cannot support this company, money comes from elsewhere: advertising business launched 6 weeks ago reached $100 million annualized revenue; enterprise API continues to expand, Codex 5 million weekly users, enterprise customers currently contribute over 40% of revenue...

In other words, ChatGPT's business model is switching from "selling model subscriptions" to "collecting platform tax": the model itself is free, advertising, enterprise services, and developer ecosystem growing on top of the model are the revenue sources.

1 billion weekly active users is the core of this strategy, it doesn't need every user to pay, only needs users to open it every day, then charge from a minority of high-value demands.

This is very classical internet platform economics, previously often compared to Claude's "enterprise market, paid programming" route, now looking at it, OpenAI is determined to become the universal entry point of the AI era.

DeepSeek: Servers Can't Hold Up

DeepSeek's situation is completely different from OpenAI.

V4 Flash topped the OpenRouter global call volume weekly chart, processing 7.22 trillion Tokens in a single week.

According to OpenCode data, on August 1 alone, V4 Flash single-day processing volume reached 8 trillion Tokens.

Interface timeouts and lag occurred frequently during peak hours on weekdays, mid-July first implemented peak-valley pricing mechanism (peak doubled), August 6 directly announced comprehensive significant price increase.

In other words: Too many users, insufficient computing power.

Too low pricing attracted a large number of low-frequency, low-payment willingness calls, server resources were occupied by invalid requests, enterprises and developers who truly need deep reasoning instead did not get stable experience.

DeepSeek needs to keep users who treat AI as a toy outside the door, let those willing to pay for high-quality reasoning stay.

Doubao launched paid version on June 24 (68/200/500 yuan three tiers monthly fee, basic functions remain free) is an action on the same logic line, 345 million monthly active users, daily consumption of tens of millions of yuan inference costs, e-commerce commissions cannot make up for it.

Nowadays, domestic large model competition has become white-hot, financing environment is also cooling, insufficient computing power is always a stranglehold, past strategies of relying on capital to extend life and unlimited subsidies are unsustainable, proving profitability has become a more urgent task than proving technological leadership.

Of course, Chinese large model price increases do not mean giving up price advantages.

More accurately, the cold start phase of the price war has ended.

The task of the previous phase was "let users be able to use it", now the task has become "let users be willing to pay for good things".

One Sinks Down, One Goes Up

When GPT-4 was just released, everyone compared model capabilities, who read more books, who had higher exam scores.

By the GPT-5.6 Luna stage, the gap between top models is visibly shrinking, continuing to compete on benchmark test scores yields diminishing marginal returns.

Competition is shifting from "who is smarter" to "who is more indispensable".

OpenAI chose to push free at this node, betting on user time and entry status.

It wants to make itself the default way for most people to access AI, just like Google used to be the default way for search.

Chinese large model vendors choosing to raise prices or launch paid versions, is under the dual pressure of capital ebb and computing power control, seeking certainty of a business closed loop, no longer satisfied with "many users but not making money" situation, need to prove the large model business can become self-sustaining.

Directions are opposite, anxiety is symmetrical.

OpenAI needs to prove platform tax can cover inference costs before the financing window closes; Chinese companies need to turn user scale into sustainable revenue before computing power peaks.

This differentiation may eventually grow into a three-layer structure.

The bottom layer is daily conversation and general Q&A.

Model capabilities have already overflowed, costs are low enough to be negligible, free will become the norm.

Whoever charges here will be abandoned by users, OpenAI making Luna free is to pull as many people as possible into this layer.

The middle layer is professional reasoning.

Code generation, data analysis, legal assistance, medical diagnosis, these scenarios have rigid requirements for accuracy and depth, users are also willing to pay for results, DeepSeek and Doubao's paid versions, target precisely this layer.

The top layer is Agent execution.

AI no longer just answers questions, but directly completes operations for users: negotiating business, fixing code, passing audits... At this layer, charging model will change from "charge by Token" to "charge by result".

Each has troubles, each has dreams.

OpenAI sinks down, betting on entry and habits, wants 1 billion people to be inseparable from it; Chinese large models go up, want users to actively pay... Both dreams are not cheap, the only thing certain is: The era of free regardless of cost has ended, the free that remains consists of businesses that can balance the books.

Seized digital assets will be held in custody through Upbit Custody, operated via a 365-day × 24-hour real-time response monitoring system, and utilizing security technologies such as 100% cold wallet isolation environment, Multi-Party Computation, Distributed Key Generation, and Multi-signature to enhance asset custody and management security.

For a fund to change its name sounds ordinary enough.

But changing "2x Long Hynix" to "At Most 2x Long Hynix"—adding the two words "At Most"—could mean the hope of breaking even for tens of thousands of unitholders is extinguished right there.

07709 is a CSOP 2x Leveraged ETF tracking South Korean semiconductor giant SK Hynix.

When it listed in October 2025, the issue price was only 7.8 HKD. Coinciding with the AI wave sweeping the globe, SK Hynix, as a core supplier of HBM chips, saw its stock price soar. This product went crazy along with it; by June 2026, the price surged to 193.65 HKD, a gain of over 10 times, with size breaking through 130 billion HKD, becoming a hot favorite in the market.

However, leverage has always been a double-edged sword.

In late June, SK Hynix's underlying stock fell from its highs, and the 2x long 07709 plummeted accordingly. The maximum decline was nearly 87%, falling from 193.65 HKD to around 25 HKD, with billions in market value evaporating. As of press time, the latest price of 07709 is 28.6 HKD.

Just as unitholders waited anxiously, hoping a rebound in the underlying stock would help them break even, the fund company CSOP made its move.

An announcement on July 27 declared that starting August 3, the product would switch to a "flexible leverage structure"—the leverage multiple would no longer be fixed at 2x, but could be dynamically adjusted between 1.1x and 2x.

In plain language, this means: when the market is good, they'll try to give you 2x; when the market is bad, they'll quietly drop it to 1.1x.

Objectively speaking, during a market crash, reducing the leverage multiple can reduce the ETF's decline, serving to protect investors. But the problem is, if the market bottoms out and rebounds, reducing leverage will make it slower for losing investors to break even, consuming more time costs, capital costs, etc. This is especially pertinent given that the Korean stock index, SK Hynix, and Samsung Electronics stock prices have already retraced so much, and Morgan Stanley has issued a report stating that the leverage clearing in Korean stocks is nearing the end and valuations are becoming attractive again.

Of course, this change has a specific background: the Securities and Futures Commission (SFC) of Hong Kong released new regulations regarding leveraged products on July 24, allowing products to adjust target leverage multiples under extreme market conditions. Procedurally, the fund company may not have violated any rules.

But does compliance necessarily mean reasonableness?

When rising, "2x" is the selling point to attract capital; when falling, "2x" becomes "At Most 2x", shrinkable at any time.

For the same product, the same batch of unitholders, the treatment between rises and falls is like night and day.

Although this operation is not illegal, it effectively rewrites the rules of the game, touching upon the most core and sensitive point of the fund industry, even this financial market, and the business society—contractual spirit.

For fund companies in Hong Kong, changes such as investment objectives, performance benchmarks, and diversification restrictions that damage investors' original expectations belong to major changes to the fund constitution and cannot be changed unilaterally by the manager.

The formal process for modifying rules is as follows:

1) The fund company and trustee submit filing to the Hong Kong SFC in advance to obtain regulatory pre-examination opinions;

2) Send a circular to all unitholders and convene a unitholder meeting;

3) A special resolution needs to be passed: more than 75% of the voting shares present must approve for it to take effect;

4) The notice period must be at least 30 days in advance, giving unitholders buffer time to redeem and exit;

5) Only after final approval by the SFC does the constitution revision officially take effect.

Did this company go through these procedures for such a major modification to the product, and did it obtain the consent of the fund unitholder meeting? If there are procedural issues, should the fund unitholders join together to claim compensation for the losses suffered?

What is more thought-provoking is that the fund charges a 1.60% management fee annually, accumulating approximately 356 million HKD in revenue from listing to present.

When the NAV surges, management fees rise with it; when the NAV plummets, management fees are collected without error.

Now that the rules have changed, the fund company has successfully avoided liquidation risk and continues to make money steadily, while unitholders who entered at high positions may not even have this last hope of waiting for the underlying stock to rebound and turning the tables via leverage.

To put it plainly, this name change is essentially sacrificing the unitholders' potential to break even in exchange for the fund company's own survival and continuity.

This structural misalignment where the manager earns steadily while unitholders lose alone is the most terrifying part.

Theoretically, if the fund company predicts the stock price will bottom out and rebound, it can adjust the leverage back to 2x, but this requires very high trading capability and precise prediction to achieve. If the fund company truly had such capability, why did it fail to successfully predict the significant retracement over the past month?

The chill here is not about the K-line chart, but a cruel story about rules, contracts, interests, and trust.

Compiled by: Jiahuan, ChainCatcher

Crypto assets are no longer a niche market. Stablecoins carry trillions of dollars in transactions annually, and major banks and payment companies are developing on-chain businesses. However, U.S. federal rules regulating these activities remain incomplete.

The CLARITY Act is designed to address this issue. The bill proposes establishing a federal regulatory framework for the crypto market, dividing responsibilities between the Securities and Exchange Commission (SEC) and the Commodity Futures Trading Commission (CFTC), requiring project parties to disclose information and restrict insider behavior, while bringing intermediaries such as trading platforms into a regulatory system similar to traditional financial markets. If the bill passes, blockchain systems will gain clear basic rules, ending the state of uncertainty that has hindered innovation and exposed consumers to risk for years.

This article is a Q&A piece compiled and published by a16z crypto based on a recent video conversation. The guests were a16z co-founder Marc Andreessen and a16z crypto founder Chris Dixon.

The two discussed why the crypto industry needs clear and lasting rules now, how the CLARITY Act will protect consumers, and why regulatory ambiguity benefits violators. The conversation also covered illegal financial activities, privacy, and government ethics; what happens if the bill fails to pass; why this matters for U.S. technological leadership; and why maintaining the status quo might be the biggest risk.

Why Does the Crypto Industry Need Rules Now?

Since the release of the Bitcoin whitepaper, the crypto industry has changed significantly. Initially used mainly by amateur players and tech enthusiasts, it has now developed into an industry with gradually maturing infrastructure and increasing institutional participation.

This technology has grown into an industry. Stablecoins process trillions of dollars in transactions annually, comparable in volume to the Visa network. Large financial institutions such as banks, asset management companies, card networks, and fintech companies are developing products around stablecoins, tokenized stocks, tokenized deposits, and other digital assets. Underlying networks have also become faster and cheaper: transactions that previously cost several dollars can now be settled in less than a second on widely used blockchains, at a cost of less than one cent.

For various reasons, U.S. regulation of crypto assets has been split into two parts: stablecoins, and the market beyond stablecoins. The GENIUS Act, effective July 2025, established a federal framework for stablecoins, but the blockchain networks and trading markets on which stablecoins rely still lack a complete federal regulatory system. This is like regulating only mobile phones while leaving communication base stations in a legal gray area.

"We are not trying to get something for nothing, nor are we asking for subsidies, protectionist policies, or other forms of support. We just want a long-term stable framework that allows everyone to conduct business responsibly. In my view, this is a perfectly natural request on many levels."——Marc Andreessen

Guidance issued by regulatory agencies can fill some gaps but cannot replace legislation. Changes in agency leadership or a new administration taking office can cause this guidance to change. When companies decide whether to invest in a business that takes five or even ten years to yield results, they must know what the rules are, which regulatory agency has jurisdiction, and whether products developed today will remain legal tomorrow.

The CLARITY Act will provide a long-term framework for businesses to operate responsibly.

How Will the CLARITY Act Protect Consumers?

The most basic consumer protection issue in the current crypto market is that crypto trading platforms are not subject to a complete federal system, whereas major securities and commodities trading venues such as the New York Stock Exchange and Nasdaq have long been under such regulation.

The New York Stock Exchange and Nasdaq have clear federal regulatory agencies. In contrast, crypto trading platforms lack registration, supervision, audit, information disclosure, trading monitoring, and customer asset protection systems covering the entire market. The CLARITY Act will provide a clear path for digital assets to transfer from SEC regulation to CFTC regulation.

Crypto trading platforms registered at the federal level will be subject to audit and financial control requirements. Platforms must safely custody customer assets, comply with anti-fraud and insider trading prohibitions, and provide operational information to regulators. Companies that refuse to meet these standards will not be able to operate legally in the United States.

These requirements help prevent scenarios that caused the FTX collapse from happening again. It was alleged that FTX transferred funds between affiliated entities, had insufficient internal controls, and the customer assets actually held did not match the amounts claimed. Federal regulation cannot guarantee that fraud will never occur, but it can significantly increase the difficulty of concealing fraud and allow regulators to intervene before problems turn into disasters.

"There must be a system first. Enterprises must have risk control, comply with regulations, and undergo audits... We also need it to prevent disasters from happening and avoid more FTXs."——Marc Andreessen

The same principles apply to products sold under the name "stablecoins." Terra-Luna was once promoted as a stable asset, but it had neither U.S. dollar reserves nor other stable reserve asset support. Under the stablecoin regulatory framework, compliant U.S. dollar stablecoins must be fully backed by corresponding reserves and audited. The CLARITY Act will bring similar constraints to other parts of the crypto market.

How Does the CLARITY Act Prevent Regulatory Ambiguity from Rewarding Violators?

Ambiguous regulatory rules can trigger a race to the bottom.

A U.S. company that takes compliance seriously may need to invest significant funds in lawyers, internal controls, audits, sanctions screening, and customer protection. These tasks are costly and may slow down product development. Offshore competitors can save on these expenses, replicate products, provide services at lower prices, and move faster, with their speed coming precisely from not doing compliance.

The result is that uncertainty penalizes responsible enterprises while benefiting offshore competitors. Law-abiding U.S. trading platforms bear all compliance costs, while non-compliant offshore platforms continue to provide services to U.S. users when they should not.

"Currently, which rules apply to which institutions is extremely unclear. I recognize that as long as there are gray areas in regulation, the market will basically move towards a race to the bottom... This ambiguity ultimately allows bad actors to take advantage."——Chris Dixon

The CLARITY Act will draw regulatory boundaries: which enterprises belong to intermediaries, which rules apply to them, which agency regulates them, and what the consequences of refusing compliance will be. Any company that custodies customer funds or assists in completing financial transactions must comply with the same types of anti-money laundering, sanctions, and Treasury regulations as similar financial institutions such as payment service providers and fintech companies.

Clear rules benefit enterprises willing to meet standards; gray areas benefit those looking for loopholes.

How Will the CLARITY Act Strengthen Sanctions Enforcement?

Privacy does not equal anonymity. People often describe public blockchains as anonymous systems, but in actual use, many public chains are highly transparent.

Transactions are permanently recorded on public ledgers. Wallet addresses do not directly display legal names, but investigators can track fund flows and link these activities to trading platforms, accounts, devices, or other identity information. Years later, the records still exist; therefore, law enforcement agencies may find evidence from them that was not available when the transactions occurred.

Some payment methods do not leave public traces, but blockchains leave traceable paths. For this reason, some national security officials describe crypto transactions as "leaving traces for future prosecution": records left today may help investigators identify and prosecute criminals in the future.

"It applies the same anti-money laundering and Treasury rules applicable to other market intermediaries to crypto intermediaries."——Chris Dixon

But traceability and privacy are two different issues. One should not be forced to disclose every medical expense or transfer to the world just to use a blockchain. The existing financial system also recognizes that ordinary people need privacy, while regulated institutions must still fulfill sanctions and anti-money laundering obligations.

Early debates surrounding internet encryption technology provide a useful reference. Strong encryption technology was once seen as a threat because criminals could also use it; in export controls, it was even classified alongside military technology. But it is precisely encryption technology that made secure banking, e-commerce, and confidential communication possible.

"Just because bad actors use encryption technology to do bad things, does that make encryption bad? Or is encryption the foundation for building trust, conducting business, and allowing law-abiding citizens at home and abroad to cooperate and do business with each other, making it valuable in itself?"——Marc Andreessen

Blockchain privacy faces the same line. Privacy protects legal activities; concealment aimed at evading the law is still subject to law enforcement pursuit.

How Does the CLARITY Act Address Stablecoin Reward Controversies While Allowing Banks to Continue Developing On-Chain Businesses?

Banks believe that stablecoin issuers and wallet service providers should not rebuild deposit accounts in disguise outside the banking system by paying interest on balances. They worry that consumers may transfer deposits from banks into stablecoin products, thereby reducing the source of funds banks use to issue loans.

The CLARITY Act addresses this concern: the bill prohibits paying interest on stablecoin balances and also prohibits products that are equivalent to interest-bearing accounts in function or economic effect.

However, the bill still allows rewards based on transaction behavior. Wallet service providers or retailers can reward customers who use stablecoins for shopping, just as credit cards offer points and retailers operate membership reward programs. The difference between the two is: the former receives rewards due to consumption behavior, while the latter collects interest merely by holding a balance.

This compromise basically meets the main demands of banks, while not going so far as to ban ordinary reward programs. Many reward programs currently offered by card networks, payment apps, and retailers also adopt similar models.

It is worth noting that the banks making these demands are themselves adopting blockchain technology. Large financial institutions such as Goldman Sachs, Fidelity, BlackRock, Stripe, Wells Fargo, and JPMorgan Chase have all developed or supported blockchain products.

"One change blockchain brings to the financial industry is providing a unified framework that allows everyone to say: 'Okay, let's enter the 21st century together.' So, it solves not only technical problems but also coordination problems."——Chris Dixon

Banks see the same opportunities as the crypto industry: existing financial infrastructure is fragmented and difficult to transform. Blockchain provides a shared framework that allows financial institutions to reduce intermediary layers, settle assets on common infrastructure, and collaboratively advance modernization, without requiring each bank to separately rebuild a system connected to each other.

Under What Circumstances Do Software Developers Need to Bear Responsibility?

The CLARITY Act distinguishes between two types of behavior: knowingly helping others commit crimes, and publishing general-purpose software. Developers who develop tools for criminal purposes, market tools to criminals, or directly assist illegal activities will still bear responsibility.

What the bill does not accept is another practice: making developers bear unlimited liability for all downstream uses that are unforeseeable and uncontrollable. Open-source code can be copied, modified, and deployed by people the developer has never seen, used in scenarios the original author never envisioned. If developers are required to be responsible for all these uses, open-source software will be almost impossible to continue developing or obtain funding support.

"This is simply impossible, and it would make software development impossible, because no developer can predict how software will be used in the future. You don't even have to look at it from a software perspective; it's the same for any product. If I run a hotel, and a criminal checks in and plans a crime in the hotel, does that make me an accomplice?"——Marc Andreessen

Its impact is not limited to the crypto industry. Academic research, startups, venture capital, and open AI models all rely on open-source software. A feasible liability boundary should be subjective intent and actual participation: when a person knowingly assists in a crime, they should be held accountable; if neutral tools are abused by others later, it cannot automatically make the tool developers bear responsibility.

How Exactly Will the CLARITY Act Handle Securities Law?

A security does not automatically become a non-security just because it is put on a blockchain. Tokenized stocks are still stocks, also still belong to securities, and continue to be regulated by the SEC. Enterprises cannot evade information disclosure, registration, and investor protection requirements just by moving assets on-chain or calling them "tokens."

"What the CLARITY Act does is just write this into law and provide clear definitions. This way, everyone can know exactly where they stand without having to file a lawsuit every time to find the answer."——Chris Dixon

What the bill aims to solve is another issue: how to regulate digital assets related to blockchain networks whose nature changes as the network develops.

Briefly, the CLARITY Act establishes a risk-based framework. A new blockchain network usually starts with a centralized entity: founders, companies, or small teams may control the network, possess information unknown to the public, and make decisions that affect token value. At this stage, related assets will be regulated by the SEC, subject to requirements similar to securities, including information disclosure, insider restrictions, and lock-up periods for founders and early investors.

As the network develops, control may gradually decentralize. If the network reaches the decentralization threshold specified in the bill, the nature of related assets may be closer to commodities rather than corporate securities. At this point, regulatory responsibility will transfer to the CFTC.

This does not mean the asset is unregulated from then on. Commodity regulation also deals with abuse behaviors such as fraud, market manipulation, and cornering the market. The regulatory agency changes because the nature of the asset itself has changed.

The bill will also introduce some restrictions that are not currently clearly defined. While the network is still controlled by a centralized entity, founders, venture capital firms, and other insiders may face longer lock-up periods and stricter information disclosure obligations. These restrictions aim to prevent insiders from dumping assets on the market before ordinary participants have obtained the same information, or before the product has developed into a sufficiently decentralized network.

What Happens If the CLARITY Act Fails to Pass?

Crypto regulation will not disappear because of this. Agencies such as the SEC, CFTC, and the U.S. Treasury have been issuing regulatory guidance and using existing authority to formulate rules within their respective jurisdictions; if the bill does not pass, they will most likely continue to do so.

The problem is that after a change in administration, regulatory agencies' interpretation of the law may change accordingly. Companies may invest years developing products based on one set of expectations, only to suddenly face a completely different interpretation after an election or change in agency leadership.

This uncertainty affects not only investment but also consumer protection. A long-term framework can clarify the powers of regulatory agencies, while requiring enterprises to complete registration, disclose information, protect customer assets, and comply with market rules. Without legislation, these responsibilities will remain scattered across different systems and may trigger disputes at any time.

"If the rules underfoot are constantly changing, enterprises are naturally less willing to invest significant time and money into development."——Chris Dixon

This industry has already experienced years of tough enforcement and political hostility; the more likely result is not the disappearance of the industry, but enterprises continuing to shift development to other regions. In this way, the supervision the U.S. can implement will instead decrease. U.S. regulatory agencies will find it harder to supervise offshore enterprises, law enforcement agencies will find it harder to reach these companies; their willingness to build products around U.S. standards will also be lower.

Why Does the CLARITY Act Continue the Tradition of U.S. Technological Leadership?

Once a technology is invented, it usually does not disappear. The real question is: where will it develop, which companies will become the leaders, and whose rules will shape it.

For more than a century, the United States has benefited from the advantage of major technologies being born and developing within the country. Technological leadership brings enterprises, employment, tax revenue, and professional capabilities, also providing economic resources for national key affairs and bringing security advantages.

"Regardless of political stance, every U.S. citizen should hope that the United States becomes the global technology leader."——Marc Andreessen

The history of encryption technology development illustrates the stakes involved. When the United States restricted the export of strong encryption technology, foreign competitors did not stop development, but placed products outside the United States, and users turned to use these products. Only after restrictions were adjusted could U.S. companies participate in building a secure internet economy.

Blockchain technology brings the same issue. Future financial systems, technical standards, and leading enterprises will emerge somewhere. If they develop mainly overseas, the United States will lose both economic opportunities and regulatory influence.

The CLARITY Act will give responsible enterprises a reason to build according to U.S. law. a16z believes this will benefit consumers, law enforcement, and national security, and also help the United States participate in setting standards for the next generation of financial infrastructure.

Which Other Institutions Support the CLARITY Act?

Supporters of the CLARITY Act include legislators, law enforcement organizations, financial institutions, and technology companies.

This legislation is the result of years of promotion by both parties in the U.S. Congress. Legislators from both parties have been working hard to establish a federal framework for the digital asset market. The largest law enforcement organization in the United States, the Fraternal Order of Police (Fraternal Order of Police), has also expressed support for the bill and refuted the claim that "the bill will weaken sanctions or anti-money laundering enforcement."

"The Fraternal Order of Police just announced support for the CLARITY Act. It is the largest law enforcement organization in the United States."——Chris Dixon

Support also comes from the financial industry. Goldman Sachs CEO David Solomon has already endorsed the CLARITY Act, and other financial institutions and fintech companies are already developing blockchain products. a16z believes that support from different fields indicates that all parties are gradually forming a consensus: the United States needs a set of clear, enforceable digital asset market rules.

When market rules are vague, consumers cannot determine what protections they enjoy; responsible enterprises must bear high compliance costs, while offshore competitors can bypass these requirements. The CLARITY Act attempts to replace this state of uncertainty with a clear system.

What really needs to be compared is not the CLARITY Act with another hypothetical law, but the system after the bill passes versus the current state. a16z believes that as long as a clear development path is provided for responsible enterprises, the bill can strengthen consumer protection, support law enforcement, and increase the possibility of next-generation financial technology developing in the United States.

In the winter of 2017, because high-speed rail did not allow large-capacity lithium batteries, Wang Xingxing held a robot dog and took a train for over ten hours from Hangzhou to Beijing to pitch to Sequoia China. At that time, his company's account was almost unable to pay wages.

9 years later, Unitree Robotics is about to become the "first humanoid robot stock" in China's A-share market. The latest news is that on August 6, Unitree Robotics announced an issue price of 150.80 yuan/share, online roadshow on the 7th, and online subscription will open on the 10th.

Many people expect its market value to break through 100 billion yuan. Multiple secondary market practitioners expressed the same view to Tencent Technology: although the stock market has fluctuated greatly recently, Unitree's IPO subscription will still be very scarce, "because everyone believes this belongs to an industry supported by policies, and the leader must rise."

According to the prospectus, the company is publicly issuing 40.4464 million new shares this time, accounting for 10% of the total share capital after issuance. On August 6, the final issue price was determined at 150.80 yuan/share, actually raising about 6.1 billion yuan, corresponding to a post-issuance valuation of about 61 billion yuan. The initial online issuance was only 6.471 million shares. Calculated at 500 shares per signature, there are only 12,942 winning numbers available for lottery in the whole market, less than 13,000. This means that for every 10,000 valid subscription allocation numbers, there are only about 2 winning spots.

According to incomplete statistics, in the past two years, more than 300 startup companies have been born in China's embodied intelligence industry. By August this year, at least 5 companies had valuations over 20 billion yuan, and nearly 50 companies were preparing for IPOs on the HK Stock Exchange or A-shares. For these companies, Unitree's stock price will constitute a valuation anchor for A-shares and a valuation reference for HK stocks.

This is a critical moment—however, in this capital feast supported by the concept of embodied intelligence, a fault line is appearing between the wealth creation in the primary market and the reality of the secondary market.

The People Who Made the Most Money from Unitree

Wang Xingxing does not fit the typical profile of a hard-tech entrepreneur—this became the origin of "contra-consensus" for Unitree's early bettors.

He graduated from Shanghai University with an ordinary resume. During early financing, he encountered setbacks everywhere. In the golden decade of internet model innovation, VC had templates for identifying people: prestigious school background, big tech executives, overseas returnee elites, or serial entrepreneurs. These to some extent guaranteed the lower limit of startup projects, but also screened out entrepreneurs like Wang Xingxing to some extent.

Chuxin Capital Partner Tian Jiangchuan once publicly reflected on this experience. At the end of 2017, Tian Jiangchuan met Wang Xingxing for the first time in a coffee shop in Hangzhou. At that time, Unitree's products had already shown extreme cost-reduction thinking and differentiated technical paths, but Tian Jiangchuan finally gave up investing. "Afterwards I reviewed, the problem mainly lay in my 'arrogance of elitism': Xingxing graduated from Shanghai University, and I thought the robot industry needed a top school background." Tian Jiangchuan later admitted. It was not until 2020 that Chuxin Capital chased Unitree back at a price more than 4 times higher.

An investor who has watched the domestic robot sector for more than ten years told Tencent Technology that when Unitree was founded, the quadruped robot sector did not receive much attention domestically, and very few institutions contacted it early on.

Time rewarded the "contra-consensus" players who entered the earliest. In 2016, Yin Fangming, who had worked at MediaTek, Sogou, and Qihoo 360, made an angel investment of 2 million yuan and obtained 15% equity in Unitree Robotics. The post-money valuation corresponding to this investment was only 13.33 million yuan. Nowadays, this investment indirectly holds equity in Unitree Robotics through the shareholding platform Tianjin Junwan Hongyi. Tianjin Junwan Hongyi holds 3.0699% equity in Unitree overall, ranking as the tenth largest shareholder. On a look-through basis, Yin Fangming actually indirectly holds about 0.46% equity in Unitree. If calculated based on the initial issue valuation of 42 billion yuan, the book value of shares indirectly held by Yin Fangming is about 200 million yuan, with an overall return rate of about 100 times. He cashed out 58 million yuan in advance by transferring part of the existing shares in 2025.

In terms of return multiples, the institution that earned the most multiples is Variable Capital. This early-stage fund invested only 2.09 million yuan in Unitree Robotics' angel round in 2018, and the return multiple so far reaches 174.62 times. Plus the exited part, the total return is about 364 million yuan.

Sequoia China's return multiple is also significant. The pitch session Wang Xingxing exchanged for by taking that train made the Sequoia Seed Fund issue an investment intention letter immediately. This investment of 15 million yuan corresponded to a post-money valuation of only 150 million yuan. After multiple rounds of additional investment, Sequoia China cumulatively invested about 102 million yuan and currently holds a combined 7.11% shares. Calculated based on the issue valuation of 42 billion, its corresponding book market value is about 2.98 billion yuan, with an absolute amount exceeding 2.5 billion yuan.

Meituan is the institution that earned the most in absolute amount. Through entities such as Hanhai Information and Chengdu Longzhu, Meituan holds a combined 9.65% shares in Unitree Robotics, becoming the largest external institutional shareholder. Calculated based on the issue valuation, the market value corresponding to Meituan's shareholding is about 4.05 billion yuan. Considering its cumulative investment of about 400 million yuan in rounds such as the 2024 B2 round, Meituan's book return exceeds 3.6 billion yuan.

Meituan Longzhu Partner Wang Xinyu met Wang Xingxing in the first week of formal work in 2016, but did not invest in him until 2024. At the end of 2023, Wang Xinyu went to the United States to research robot laboratories at top universities such as Harvard, MIT, and Stanford, and found that these institutions representing the world's most cutting-edge research forces were all using Unitree's robot dogs for secondary development. "If the best PhD students in the world are using Unitree's robots for the most cutting-edge research, won't its AI capabilities be solved?" Wang Xinyu once said in a media interview.

Other investors who entered early or mid-stage also benefited richly. Matrix Partners China entered around the 2022 Series B valuation of about 1.12 billion yuan, currently holds a combined 5.45% shares, corresponding to a market value of about 2.29 billion yuan, with a book return multiple of about 45 times; Shunwei Capital first entered in January 2021 at a valuation of 380 million yuan, currently holds 3.98% shares, corresponding to a market value of about 1.67 billion yuan, with a book return multiple of about 26 times; the CITIC group entered in the 2024 B2 round, currently holds 4.49% shares, corresponding to a market value of about 1.88 billion yuan; the SCGC group cumulatively invested about 90 million yuan, currently holds about 2.55% shares, corresponding to a market value of about 1.07 billion yuan, with book returns all around 10 times. Shanghai Yuyi, as an employee shareholding platform, holds 10.94% equity, corresponding to a market value of about 4.59 billion yuan, among which 14 core employees hold about 5.92 million shares, with an average per capita book market value of nearly 48.9 million yuan.

In June 2025, Unitree Robotics confirmed the completion of Series C financing closing, led by funds under China Mobile, Tencent, Jinqiu, Alibaba, Ant, and Geely Capital, with the post-money valuation reaching 12.7 billion yuan. Calculated based on the 2016 angel round valuation, its valuation surged nearly 1000 times in 9 years. Before the IPO issuance, the combined shareholding ratio of the top ten shareholders reached 71.50%.

For early investors, a valuation of 61 billion is already enough for them to succeed and retire.

The Primary Market Expects Unitree to Surge

For the primary market, Unitree's listing is crucial. Currently, the valuations of several leading unlisted embodied intelligence companies have reached 20 to 30 billion yuan. If there is no "anchor" of Unitree's surge in the secondary market, subsequent high-valuation projects will be affected.

A CEO of a robot company with a valuation over 10 billion told Tencent Technology that just like NIO's stock price continued to fall for a while after listing, causing XPeng's subsequent financing to be hindered, "although each company looks different, investors think you are all robots." They are all expecting the stock price of the company that lists first in the industry to surge.

Multiple robot company management personnel told Tencent Technology that humanoid robots are eager to list, first for primary-secondary linkage fundraising, and second often due to urging from capital parties. "On the surface, these institutions have not yet reached the exit cycle, but as long as one company lists, subsequent companies will face huge pressure from shareholders." A CEO of an embodied intelligence component company currently preparing for listing told Tencent Technology.

Some views compare the current embodied intelligence to new energy in 2021, expecting that more than 80% of companies will be eliminated in the future. Concerns about future uncertainty in the capital market and the actual slow landing speed of the industry constitute universal anxiety in the industry.

Capital is willing to pay an issue P/E ratio of 219.23 times for Unitree—far exceeding the industry average P/E ratio of 38.56 times—betting on a future where humanoid robots can fully replace human labor. But currently, Unitree's profits mainly come from robot dogs. Unitree admitted in the prospectus, "During the reporting period, the company has not yet scaled the application of its self-developed general embodied large model to robot products. If brain technology fails to make important progress, there is uncertainty in the large-scale application of general robots."

In other words, capital pays for its market value with the "brain", but Unitree can currently only make money by selling the "cerebellum".

Unitree is trying to make up for this key lesson. In the proposed fundraising of 6.099 billion yuan, there is a special fund for intelligent robot model research and development projects. This is the necessary path for its leap from a "hardware manufacturer" to an "embodied intelligence full-stack platform", and is also the key to supporting its market value.

No Consensus in Secondary Market, But Actions Are Honest

If there is consensus in the primary market, then the influence of this consensus cannot be called sustainable in the secondary market yet.

In the current A-share environment, funding is relatively tight, and old-brand value stocks and semiconductor sectors have diverted a lot of funds. The "Embodied Intelligence/Physical AI" favorite in the eyes of primary market investors may be just a hardware company facing high valuation pressure in the eyes of some secondary market funds. A public fund person told Tencent Technology that the gap between the two logics is one of the sources of uncertainty Unitree faces after listing.

Unitree's circulating cap on the first day of listing is extremely small, amplifying this emotional gaming. The total public issuance of Unitree Robotics this time is 40.4464 million shares, but the initial issuance volume facing online investors is only 6.4710 million shares, accounting for 16% of the total issuance. The remaining 84% of chips were allocated to institutional investors through strategic placement (8.0893 million shares, accounting for 20%, locked for 12 to 24 months) and offline placement (25.8861 million shares, accounting for 64%). In the total share capital of 404.4643 million shares after issuance, only about 29.77 million shares are tradable on the first day, accounting for about 7.36% of the total share capital. More than 90% of shares are in a locked state on the first day of listing. Under conditions of limited supply, once market sentiment is high, stock price elasticity will be significantly amplified.

The aforementioned public fund person told Tencent Technology that the speculation on the Unitree chain in the secondary market has always shown typical "event-driven, surge and fall" characteristics. He believes that this attention sustainability is very limited, "often where it rises, it returns where it was, cannot be maintained long-term."

Taking the 2026 Spring Festival as an example. Funds gambled in advance before the festival on the expectation of humanoid robots appearing on CCTV Spring Festival Gala, completing a round of speculation. On the night of February 16, New Year's Eve, humanoid robots from companies such as Unitree Robotics appeared on stage. The good news was quickly realized and evolved into funds leaving. On February 20, the first trading day of the Year of the Horse in HK stocks, targets such as Dobot and UBTECH surged significantly and then fell back quickly; only a few trading days later on February 24, the A-share robot concept sector suffered a heavy blow, Wzhou Xinchun had a maximum intraday drop of over 9%, closing down 6.9%, and core component companies such as Leader Harmonious Drive and Wanxiang Qianchao led the sector decline.

The process of Unitree Robotics sprinting for the STAR Market was also extremely volatile. From acceptance on March 20 to registration approval landing on July 2, it took only 104 days, creating the fastest audit record on the STAR Market. On July 2, the CSRC approved Unitree Robotics' IPO registration. The next day, the A-share robot concept exploded across the board, with over 50 stocks hitting the limit up or rising more than 10%. However, just in the first three weeks of July when Unitree's IPO registration became effective and the World Artificial Intelligence Conference was held intensively, the CSI Robot Index fell 12.77% in a single week, and the STAR 50 plummeted 10.5% cumulatively in three days. When expectations were realized in advance, the market lost incremental funds, and capital flight under high crowding led to a tragic stock price plunge.

In the whole of July, A-share market value evaporated over 12 trillion yuan, the Shanghai Composite Index fell cumulatively by 6.4%, the Shenzhen Component Index fell drastically by 16.21%, the ChiNext Index fell drastically by 23%, and the STAR 50 Index fell drastically by 25.90%, creating the largest single-month drop in history. AI concept stocks welcomed the most tragic month, the ChiNext and STAR 50 indices fell 23% and 25.90% respectively, and the semiconductor index fell over 33%.

In other words, Unitree's valuation anchor actually does not depend on Unitree itself.

Multiple analysts focusing on the robot sector pointed out that what can still drive the entire humanoid robot industry is Tesla, the reason being analogy with the electric vehicle industry: truly pure electric vehicles began to walk into people's lives only after Tesla Model 3 scaleization, which triggered changes in domestic cognition of pure electric vehicles, and then the rise of other brands. And even the Tesla chain regarded as a weather vane faces great uncertainty.

Tesla Optimus third generation release expectation is delayed to the first quarter of 2026, with core upgrade points focusing on hand flexibility and body structure. This continuous change at the design level means that the supply chain previously carefully dismantled and repeatedly speculated by the market faces restarting at any time. "From the release of the first generation Optimus to now, except for general contractors like Sanhua and Tuopu, the design material usage, corresponding suppliers, and value amount of every joint have changed several rounds. The design we see today may be completely different from the design when future actual humanoid robots begin to popularize. An analyst said.

Against this background, the aforementioned public fund person believes that the key for the entire humanoid robot sector still depends on whether Tesla can open expectations. "If Tesla cannot open this expectation, Tesla's stock price itself is also falling, also in a slump, industrial progress continues to be lower than expectations, Unitree will also find it difficult to walk out an independent upward trend."

Despite this, he said, not a single person around him watching Unitree's IPO subscription has decreased.

编译：Luffy，Foresight News

早在 1503 年，罗马就出现了针对下一任教皇人选的下注活动。到 1916 年，按照 2012 年美元币值换算，美国人单单在美国总统大选这一件事上的投注规模就达到约 2.11 亿美元，全部发生在纽约的博彩市场。交易繁忙的日子里，选举相关投注的交易量甚至超过华尔街场外交易所的股票交易。

自文艺复兴时期起，人们就开始用金钱为不确定的结果定价，而这类市场在这件事上表现得相当出色。那么为什么成熟可行的预测市场迟迟没有出现？Kalshi 与 Polymarket 又是如何最终实现突破？本文将梳理预测市场的发展历程，并探讨它未来的发展方向。

什么是预测市场？

预测市场允许用户针对未来事件结果交易份额，交易价格介于 0‑1 美元之间，该价格实时反映市场对事件发生概率的判断。和体育博彩不同，你不用持仓至事件结算。随着市场波动、概率发生变化，你可以随时开仓或者平仓。

理论上，预测市场可以成为包罗万象的信息市场：可以交易美联储加息概率、泰勒・斯威夫特能斩获多少个格莱美、2 月 18 日巴黎的气温等等。

Kalshi 的文化主题预测市场

理论基础与早期困境

1988 年是现代预测市场的奠基之年。该领域公认的鼻祖 Robin Hanson 写下了信息市场与理念期货的首批学术理论。同年，爱荷华大学三位教授搭建了爱荷华电子市场（IEM）。历经五届选举周期，爱荷华电子市场给出的概率预测，有 74% 的情况准确度超过民调，印证了弗里德里希・哈耶克 1945 年提出的观点：市场是汇集群体智慧最高效的方式。

爱荷华电子市场创始人 George Neumann、Forrest Nelson、Robert Forsythe，来源：NBC News

尽管早期已经得到概念验证，但 2000‑2010 年代充斥着大量失败项目。 2003 年 7 月，美国国防高级研究计划局（DARPA）推出的政策分析市场上线仅一天便被叫停，两名参议员指责这个由 Robin Hanson 设计的项目相当于开设刺杀事件博彩市场。美国国会禁止好莱坞证券交易所转型为真正的电影期货交易所。 Intrade 在都柏林运营十余年，2012 年遭到美国商品期货交易委员会（CFTC）起诉，理由是其向美国用户提供未注册期权，该平台于 2013 年 3 月倒闭。

加密行业曾被视作解决方案。以太坊主网上线为开发者提供了可编程底层基础设施，去中心化与抗审查特性，看上去恰好契合预测市场的发展需求。但新的问题接踵而至。2018 年上线的 Augur，用户需要承担高昂的以太坊 Gas 费，产品体验糟糕。平台用户峰值仅 265 人，一个月内就暴跌至 37 人。

预测市场面临的问题

2024 年之前大量项目走向消亡，常见解释是监管环境严苛、产品执行力糟糕。团队低估了监管审查力度，同时忽视产品界面与用户体验。但这些还不足以解释预测市场深层次的结构性矛盾。Nick Whitaker 与 J. Zachary Mazlish 在 2024 年一篇广为流传的《Works in Progress》文章中，提出了更透彻的分析。

一个能够持续运转的市场，需要三类核心参与者：

• 储蓄者：追求长期收益，实现财富增值
• 赌徒：追求刺激与快感
• 专业交易者：依靠深度分析，从错误定价中套利

基础形态的预测市场，对这三类群体都缺乏吸引力。 预测市场属于零和博弈，扣除手续费之后更是负和博弈，储蓄者因此完全不愿参与，他们需要正和市场实现财富增长。 绝大多数现实事件结算周期漫长、题材小众，很难吸引赌徒，赌徒普遍偏爱快速出结果的标的。

缺少储蓄者与赌徒带来的交易对手盘，专业交易者也找不到值得入场的流动性。市场最后就只剩下专业交易者之间互相博弈，这就是现实版本的无交易定理：如果所有人都足够理性，没有人愿意充当对手方。

抛开市场结构，绝大多数题材对普通大众吸引力有限。没有交易量，专业交易者就没有动力入场，去争夺微薄的潜在收益。当然也存在例外，例如体育、政治题材。Whitaker 和 Mazlish 得出结论，如果没有外部补贴，「万物皆可预测」 的预测市场模式无法规模化。

预测市场如何终于实现突围

尽管存在上述种种现实缺陷，预测市场还是成长为成熟的产品品类。2024 年美国总统大选期间迎来爆发，其给出的概率被广泛引用，成为一类事实参考依据。《纽约时报》引用预测市场数据，CNBC 进行播报，彭博终端更是直接内置相关数据。 该赛道企业总融资规模已经超过 50 亿美元，过去 18 个月融资还在加速。

即便你没有关注这个赛道，大概率也听过推动行业突破的两家平台：Polymarket 与 Kalshi。两家合计占据行业 90% 以上交易量，行业月度总交易量突破 580 亿美元。

Polymarket

Polymarket 由 Shayne Coplan 在 2020 年创立。他是纽约大学肄业生，2014 年参与以太坊 ICO，2019 年曾写信给 Robin Hanson，希望把预测市场变为现实。新冠疫情期间，他在纽约下东区的公寓里推出了这款产品。

加密基础设施的迭代，例如低成本二层网络、稳定币，帮助 Polymarket 避开早期加密预测市场的陷阱。平台运行在 Polygon（以太坊二层网络），Gas 费被压至几美分；使用平台自有稳定币 PUSD 完成结算，1 美元的兑付金额就是实打实的 1 美元，持仓周期内不存在价格波动风险。 交易采用混合订单簿，链外撮合保障速度，链上完成结算保障可信，兼具中心化交易所的流畅体验与非托管结算的特性。

Polymarket 创始人兼 CEO Shayne Coplan，来源：Forbes

Polymarket 采取 「先上线，后续解决监管问题」 的思路，相比同期竞品获得更高自由度与更快的迭代速度。2020 美国总统大选期间平台获得早期增长，月交易量达到约 2600 万美元，之后又依靠疫情、流行文化相关市场持续扩张。

监管缺失的隐患最终找上门。2022 年 1 月，CFTC 对 Polymarket 处以 140 万美元罚款，并且要求平台封禁美国用户。合规自此成为优先级最高的事项：平台对美国地区做地理屏蔽，聘请前 CFTC 主席担任顾问，面向全球其余地区继续运营。2023 年平台交易量约 7300 万美元，与目前的交易量相比微不足道，但足以使其熬过加密寒冬。

接下来就是 2024 年美国总统大选，属于行业的高光时刻。虽然正式禁止美国用户访问，Polymarket 却成为本次大选文化层面的代表平台。大选相关市场累计交易量约 36 亿美元，对特朗普胜选的概率判断，准确度超过民调与专家评论。这次爆火把预测市场推到全球视野之下。

大选结束一周后，联邦调查局 FBI 突击搜查 Coplan 的公寓，调查平台是否存在美国用户绕过 2022 年禁令在国际版网站交易的行为。2025 年 7 月，美国司法部与 CFTC 结束调查，没有提起任何指控。 数日之后，Polymarket 斥资 1.12 亿美元收购拥有 CFTC 牌照的交易所 QCEX。同年 10 月，纽约证券交易所母公司洲际交易所（ICE）同意向 Polymarket 最高投资 20 亿美元，投资前估值 80 亿美元；ICE 同时成为 Polymarket 事件数据的全球分销商。依靠收购 QCEX，Polymarket 在 2025 年 12 月重新进入美国市场。

大选热度退去之后，体育、国际地缘事件带动平台持续增长。截至 2026 年 7 月，Polymarket 累计完成 7.077 亿笔交易，总交易量突破 1119 亿美元。在 2024 年它毫无疑问是行业龙头，之后被 Kalshi 反超。

Kalshi

Kalshi 从成立之初就做出了截然相反的选择。创始人 Tarek Mansour 与 Luana Lopes Lara 均为麻省理工学院毕业生（Lara 曾经是职业芭蕾舞演员，出演过《天鹅湖》，之后转行进入金融市场），2018 年创立公司。他们赌的是：从一开始就坚守合规，比发展速度更加重要。

Kalshi 联合创始人 Tarek Mansour、Luana Lopes Lara，来源：Forbes

两位创始人等待了近两年时间才拿到上线批准。2020 年 11 月，CFTC 批准 Kalshi 成为指定合约市场，这是全美首家获得联邦监管许可、可以挂牌事件合约衍生品的交易所。这张牌照构成 Kalshi 核心法律依据，联邦法律优先级高于各州博彩相关法律。Kalshi 于 2021 年 7 月正式上线。

但拿到牌照也带来一系列麻烦。平台上新热门事件流程缓慢，严格的 KYC 要求进一步拖累增长。最沉重的打击是：平台申请开设 2024 美国总统大选市场，遭到 CFTC 驳回。面对行业史上最大的催化剂，Kalshi 却无法参与。

Kalshi 提起诉讼并且胜诉。2024 年 9 月，一名联邦法官裁定 CFTC 越权，选举相关合约既不违法，也不属于博彩。距离大选投票仅剩 32 天，Kalshi 才重新开放大选交易。这次延误，叠加严格 KYC 导致缺少国际用户，让它在用户心智与市场份额上输给 Polymarket。Kalshi 大选相关交易量仅约 5 亿美元，而 Polymarket 达到 36 亿美元。

大选结束之后，合规方面的投入开始获得回报。Robinhood 和 Kalshi 合作，推出首款预测市场产品；彭博终端直接接入 Kalshi 的数据。这两笔合作的基础正是 Kalshi 的监管资质。

2025‑2026 年，Kalshi 大力拓展政治之外的赛道。体育成为第一大品类，出圈营销起到很大作用。NBA 总决赛街头采访中出现的 「尼克斯四场横扫」 口号，相关片段播放量达数千万。团队借势世界杯展开营销，投放带有提莫西・查拉梅、利昂内尔・梅西、卢卡・东契奇的广告。世界杯期间平台 300 万用户创造 270 亿美元交易量。

截至 2026 年 7 月，Kalshi 累计完成 9.826 亿笔交易，总交易量 1557 亿美元，规模超越 Polymarket，成为行业新龙头。

渠道全面铺开

两家头部平台的成功，带动大量交易所、券商推出自家预测市场业务。许多公司采取了快速上线策略，通过将现有平台集成到自己的应用程序中来测试市场需求。

Coinbase 与盈透证券另辟蹊径，做聚合模式，汇集多个平台的流动性与市场。CME 集团从零搭建自有产品。Robinhood 采取系统化路径，先路由订单至 Kalshi，需求得到验证之后再自建底层系统，2026 年 6 月推出受 CFTC 监管的交易所 Rothera。这次押注很快得到回报，Robinhood 事件合约业务 2026 年二季度营收 1.56 亿美元，同比增长超 10 倍，已经超过加密交易业务的 1 亿美元营收。

结构性难题已经被解决了吗？

Kalshi 与 Polymarket 解决了执行层面的失败：监管策略、低廉手续费、面向普通消费者的产品体验。但它们是否破解 Whitaker 和 Mazlish 提出的深层需求难题，要另当别论。

• 赌徒群体：部分解决，但远未做到万物皆可预测。体育题材天然适配预测市场，核心问题只是能否从传统体育博彩手里抢夺交易量，而它们已经做到。今年至今，大部分交易量来自体育。组合投注玩法是重要增长动力：2025 年 9 月在 Kalshi 上线时仅占总交易量 3%，到 2026 年 7 月已经攀升至 38%。紧随其后的是加密货币价格预测。政治题材不再局限大选，军事与地缘冲突市场交易量 27.6 亿美元，略高于美国大选市场的 27.3 亿美元，加上海外选举，广义政治赛道规模继续走高。文化市场（音乐、影视、名人）规模持续增长。美联储利率决议、通胀这类经济题材同样有量。多个赛道交易量都在上涨。虽然距离 「所有题材都能形成活跃市场」 的宏大愿景还有很远，但过去 12 个月已经取得实质性进展。
• 专业交易者：依靠激励得到部分解决。专业交易者需要充足交易量，以及除其他专业交易者之外的对手盘。体育市场 1200 亿美元、加密市场 220 亿美元的体量，已经可以满足这一点。Susquehanna 2024 年作为做市商接入 Kalshi；之后和 Robinhood 成立预测市场合资公司。Jump Trading 向两家平台都进行股权投资，以此换取提供流动性；Citadel 也在评估入局可能性。
• 储蓄者群体：问题仍未解决。预测市场仍然是零和博弈，投入赌博的资金放弃了投资国债或其他地方可能获得的收益。

预测市场的下一个时代

下一阶段，预测市场将从小众平台演变为给全球信息定价的基础设施。新市场也会催生出全新的机制。

定制化对冲

企业可以对冲传统金融、保险无法覆盖的特殊风险。举个例子，一家冰淇淋店可以针对夏季气温偏低做风险对冲。这类需求过去无法被满足，传统保险公司不会承接，因为小众标的很难实现盈利承保。

突破 0‑1 美元二元定价模式

永续市场：针对任意事件的连续交易市场。以通胀举例：二元市场只能赌固定结果，例如 「通胀会不会高于 3.1%？」。永续市场允许直接做多、做空通胀率本身。

组合市场与理念治理：不再只对单一事件定价，而是定价两件事件之间的关联。例如 「如果埃隆・马斯克辞职，特斯拉股价会是多少？」，「如果美国发动入侵，油价会到什么位置？」。多个市场分别对各个变量定价，可以更加精准评估资产价值。理念治理更进一步，借助条件市场辅助治理与政策决策：推行市场预判会带来更好结果的政策。

AI 智能体充当真相挖掘者

未来市场会大量运行 AI 智能体，全天候完成调研与交易。这类智能体可以充当自动化真相探测器，扫描电报社群、社交媒体，比人类评论员更快挖掘事实依据，针对错误定价进行交易，进一步提升预测市场准确度。

媒体的数据底层

新闻深度融合：CNN、CNBC 的合作预示着和传统媒体更深的共生关系。媒体报道不再只聚焦已经发生的事实，同时覆盖未来将会发生什么。

长尾题材拓展：市场拓展至长尾领域，给社区本地事件、小众文化趋势定价，例如科技行业裁员、泰勒・斯威夫特专辑表现。但想要激发需求，需要改变用户习惯，用户不只是看新闻，还要参与对结果的押注。

解决储蓄者缺失的难题

带收益的抵押品：交易者不再质押闲置的 USDC，可以使用生息资产，例如 sUSDe、代币化美债充当保证金。仅 Ethena 的 sUSDe 在过去两年就通过获取永久资金利率实现了 4% 至 30% 的年化收益率 ，并且已经可以作为抵押品接入 Aave、Pendle 和 Morpho 等平台。如果预测市场支持同类抵押品，储蓄者的资本在充当仓位担保的同时持续赚取收益，而不是在事件结算之前完全闲置。

金融与 DeFi 模块化组合：可以在预测市场之上搭建结构化产品，既可以和其他资产打包，也可以当做抵押品进行借贷。当事件合约可以与生息资产打包，或是用于抵押借贷，那么零和博弈就不再是该资本的唯一回报。

Trump and Warsh established direct communication, which may reduce the policy shock caused by the White House publicly attacking the Fed in the short term, but at the same time increased market skepticism regarding central bank independence and the policy reaction function. Consequently, the market has shown a "bear steepener" structure with relatively stable short-end yields and higher long-end yields: investors believe Warsh may not raise rates immediately, but demand higher long-term inflation and institutional risk premiums. The 10-year US Treasury yield is currently around 4.64%, and the 30-year remains above 5%.

US stock indices remain near historical highs, but the internal structure has weakened. For chip, memory, and software companies, even if giving strong revenue guidance, stock prices may still fall, indicating that the AI trade is shifting from "revenue growth" to testing capital return rates, profit margins, and valuation realization.

In the baseline scenario, the market in August will maintain volatility at high levels and sector rotation, rather than immediately entering a full-scale bear market. AI leaders with stable cash flows, financials, industrials, and some energy stocks may relatively outperform, while high-valuation semiconductors, memory, and unprofitable software stocks face further valuation compression. If oil prices remain around $80 and employment continues to cool, the Fed may continue to pause rate hikes in September, and tech stocks are expected to rebound; if oil prices break through $90 again and wages and inflation remain sticky, the Fed may raise rates, long-term bond yields continue to rise, and the Nasdaq will face a more significant risk of a 10% level correction.

Note: The content of this article is not investment advice, nor does it constitute an offer, solicitation, or recommendation for any investment product

MiniMax stated on platform X that this is precisely why they chose to open-source; by sharing model weights, they can spark community creativity and rapidly iterate engineering-level optimization results. This indicates that open-source strategies can effectively accelerate the optimization iteration cycle of AI models and are a paradigm of technical community collaboration.

The data communication market is expected to expand at a CAGR of 28% over the next five years, exceeding $70 billion by 2030. 1.6T products are the largest source of incremental growth, with a CAGR of approximately 120%, contributing about $40 billion by 2030. JPMorgan pointed out in a research report on August 6 that after including the NPO/CPO market in the addressable market calculation, the scale will surpass $18 billion by 2030, accounting for more than 25% of the entire data center communication market. The optical module market is undergoing a long-term structural inflection point from pluggable to co-packaged. Google is expected to become the largest optical device purchaser by 2030, and Meta will lead in NPO/CPO deployment. In terms of short-term earnings, COHR has the highest earnings certainty, LITE has significant controversy but valuation provides a margin of safety, and FN needs September quarter guidance to boost confidence.

COHR Fundamentals Most Solid, LITE Controversy Overpriced

COHR has the highest earnings certainty. Product capacity is gradually increasing, and revenue and profit margins are expected to continue improving in the second half of the year. Telecommunications and data center infrastructure growth momentum is strong, and optical module and CPO business expansion is proceeding as expected. The demand environment in the industrial sector is improving, and most covered companies have confirmed this trend, which will help demand growth and product structure optimization in the second half of the year. COHR's current stock price corresponds to 23 times the expected earnings per share for fiscal year 2028, and accelerated growth and product structure optimization are expected to drive earnings beyond expectations.

Market concerns about LITE are somewhat excessive. The main controversies focus on three aspects: increased revenue from lower-margin TRx business may drag down gross margins, long-term risks brought by Chinese competitors increasing indium phosphide (InP) supply, and potential delays in CPO capacity expansion. JPMorgan believes these concerns will not be significantly reflected in the upcoming earnings report. The pressure on gross margins from accelerated growth in Google's TRx business will be offset by growth in higher-margin businesses such as telecommunications/DCI and EML. Guidance for capacity expansion of CPO and optical circuit switching systems will meet expectations. LITE's current valuation is only 22 times the expected earnings per share for fiscal year 2028. Investor sentiment is cautious before the earnings report, which instead provides room for upside potential after earnings exceed expectations.

FN is the hardest to predict among the three. June quarter performance may be flat, and there is likely no significant improvement in EML supply. The key lies in the September quarter earnings guidance; management needs to emphasize accelerated business growth, especially in the optical module and high-performance computing fields. Nvidia optical module-related EML capacity will increase, AWS optical module business continues to grow, and Trainium/HPC business is expected to achieve $150 million in revenue. However, FN's current stock price corresponds to 23 times the expected earnings per share for fiscal year 2028, comparable to COHR and LITE, while earnings growth potential is relatively lower, and the risk-reward ratio is not as good as the former two.

1.6T Drives Market Explosion, Pluggable Dominates Growth Before 2028

LightCounting's latest forecast raises the overall CAGR of the data communication market to 28%, with the scale growing from $20 billion in 2025 to over $70 billion by 2030. 1.6T is the core of growth, with a CAGR of approximately 120%, contributing about $40 billion by 2030. 3.2T will start contributing from 2027 and is expected to increase to $14 billion by 2030. 800G will continue to expand before 2028, followed by slowed growth.

The latest forecast is raised by about 15% overall, of which the 1.6T market contributed about 35% of the increment, while 800G and other markets remained basically flat.

Pluggable transceivers dominate growth before 2028. From 2026 to 2028, the pluggable market will increase by $14 billion, while the increments for LRO/LPO, NPO, and CPO are each less than $3 billion. From 2028 to 2030, the pattern shifts; NPO and CPO will grow by $9 billion and $6 billion respectively, while the increments for pluggable and LRO/LPO are both less than $2 billion.

After NPO was included in the addressable market calculation, NPO/CPO forecasts were significantly raised. Compared with April, the average forecast for 2026 and beyond was raised by more than 50%. By 2030, NPO and CPO combined will exceed $18 billion, of which about 40% of deployments are for scale-out scenarios. CPO distribution between scale-out and horizontal expansion is more uniform; NPO introduces additional latency because signals need to be transmitted through longer copper wires, so its application in scale-out scenarios is relatively limited. It is expected that by 2030, only 33% of NPO deployments will be related to scale-out.

Google Becomes Largest Buyer, Meta Leads NPO/CPO

By 2030, Google is expected to become the largest customer for data center communication optical devices. Meta and Google will also become top customers in the NPO/CPO field. The top five U.S. cloud service providers will expand at a CAGR of 29%, in line with the overall market growth, collectively accounting for more than 60% of the market share. Google and Meta's growth rates are expected to exceed the industry average. By 2030, Meta will lead in NPO/CPO deployment, followed by Google, Microsoft, Amazon, and Oracle.

Nvidia will regain share in the short term due to 1.6T dominance, with overall share in the data center market expected to rebound to 25% in 2026, accounting for more than 70% of the 1.6T market. However, in the long run, as cloud service providers increasingly procure products from third parties, Nvidia's market share is expected to drop from 18% in 2025 to 12% in 2030.

The telecommunications and data center interconnect market is expected to grow at a CAGR of 18% to 2030, with market size increasing from $4 billion in 2025 to $9 billion. 1.6T is the main driver, with a CAGR exceeding 100%, contributing over $3 billion by 2030. 800G CAGR during the same period is 40%, contributing about $3 billion. The latest forecast is raised by 11% on average, mainly from the high-speed optical module market in ZR form factor.

The long-term narrative of the optical module market is switching; pluggable dominates growth before 2028, and NPO and CPO take over the increment after 2028. Google and Meta are redefining the customer landscape, and Nvidia's short-term 1.6T share rebound will not reverse the long-term trend.

Disclaimer

This article is a compilation and interpretation by TechFlow Research of a third-party broker research report (JPMorgan, August 6, 2026), combined with public market information. The ratings, target prices, earnings forecasts, and related judgments cited in the article are the views of the broker's analysts, represent only their institution's position, do not represent the views of TechFlow Research, and do not constitute any investment advice.

The market involves risks, decisions need to be independent. This article should not be used as a basis for buying or selling any securities.

In terms of short-term financial reports, JPMorgan Chase believes COHR has the highest performance certainty, with revenue and profit margins expected to continue improving; the market is overly concerned about LITE, with valuation corresponding to only 22x 2028 EPS, leaving room for outperformance; FN needs September quarterly guidance to boost confidence. In terms of customer landscape, Google is expected to become the largest optical component purchaser by 2030, and Meta will lead NPO/CPO deployment. Nvidia's short-term share rebounded to 25% due to 1.6T, but long-term share is expected to drop from 18% to 12%.