Never really got to explain my reasoning for my $SIVB short - but suffice it to say now if you marked their holdings to market their book value would be negative
I believe hyperscalers will issue more than double current issuance expectations for 27-28. Beyond the potential for IG spreads to widen, I don’t really know what the implications will be.
So @jukan05 has been saying how glad he is that his income is in USD because KRW keeps going down, seems like a good opportunity to test his conviction on memory.
Pretty cool, maybe Jensen knows what he’s doing.
But idk, I saw a bunch of finance people on Twitter who didn’t know what a token was until last year say that Nvidia doing AI PCs was stupid.
Need to give a shoutout to Ascend for helping us plan Analyst #3’s travel to Oman at 3AM back in March (and saving us a ton of money on other travel as well).
If you travel a lot, you should absolutely check out joinascend.com. And mention Citrini referred you.
We don’t do paid promotion ever and Ascend hasn’t given us anything or asked us to do this, they’re just a really good operation that should be more well known.
Our North American World Cup beneficiaries basket has sneakily broken out to new highs amidst…well, the World Cup.
Lead this week by AirBnB who’s CEO was doing customer support on X today (bullish indication of locked in status imo)
This isn’t what min-maxing means. It’s a term from originally D&D where you have a set number of points to allocate to your attributes (like strength, dexterity, wisdom etc.)
Wanna be the world’s strongest warrior? Dump your points into strength by having the intelligence score
John: "The tokenmaxxing/ token-minimizing thing is so silly."
"Do you know where 'maxxing' comes from? It comes from gamer lingo — 'min-maxing.'"
"You shouldn't be tokenmaxxing or token-minning. You should be token min-maxing."
"The idea of min-maxing is, when you're in a
AI workloads are creating new memory objects that are too large for HBM but too valuable for cold storage. It’s simply to expensive to use DRAM for everything if it’s possible to use NAND for some of it. We’ve gotten three announcements focused on flash from three major players
AMD and AAPL both, in different ways, have announced a focus on flash as a way to reduce the AI memory tax. We wrote about this in our State of the Themes last week.
$AMD is acquiring MEXT, an AI-driven memory optimization company for data center infrastructure. MEXT’s tech helps make flash storage behave more like DRAM, expanding usable memory capacity while maintaining performance.
AI workloads are creating new memory objects that are too large for HBM but too valuable for cold storage. It’s simply to expensive to use DRAM for everything if it’s possible to use NAND for some of it. We’ve gotten three announcements focused on flash from three major players
Whether this is yet another jevons paradox situation that simply means the DRAM shortage remains as severe but we use it even more effectively, or something that lays the groundwork for DRAM margins to get hit as lower margin (and easier-to-produce, especially by China) flash
Market down on Iran: “I should sell stuff but I’m gonna buy more AI stuff because it’s not like higher oil prices impact AI”
Market up on Iran: “I’m gonna buy more AI stuff, risk on baby”
Undefeated theme tbh.