SemiAnalysis (@SemiAnalysis

SemiAnalysis

2,446 posts

SemiAnalysis

@SemiAnalysis_

Joined January 2024

SemiAnalysis
@SemiAnalysis_
4h
Pretraining fundamentally does not make sense anymore for anyone other than frontier labs. Although there are a lot of people at enterprises & startups who have "Pretrainitis" to show “impact” and get promotions, fundamentally, it doesn’t make sense. There is probably higher ROI
32K
SemiAnalysis
@SemiAnalysis_
8h
GPU Racks hitting 400kW? Legacy data centers wont be able to handle it and the grid WILL get throttled. Radiant's 12 month, dirt to AI production, was made possible by bypassing the grid. Head of Infrastructure, Patrick Wohlschlegel tells @JordanNanos
94K
SemiAnalysis
@SemiAnalysis_
12h
Intel Should Raise Capital Intel's woes are behind them. The heavy spending is ahead of them. Why an equity issuance in a hot equity market could make Intel so much better sooner.
Intel Should Raise Capital
From newsletter.semianalysis.com
128K
SemiAnalysis
@SemiAnalysis_
Jun 11
SLOP ALERT: Claude Code UI is complete slop. In the in-app file tree, when u click on a .png, it opens it as a base64-encoded file instead of rendering the image. We’d rather Anthropic not release the desktop app than release an L desktop App. Tons of bugs.
47K
SemiAnalysis
@SemiAnalysis_
Jun 10
Replying to @SemiAnalysis_
Obviously this is way worse than API overall. However, explicitly nerfing subscriptions leads to huge public backlash, and the rapidly falling cost of intelligence means you'll be able to profitably serve Opus 4.8 level models for $20/month in the near future. We therefore think
143K
SemiAnalysis
@SemiAnalysis_
Jun 10
Replying to @SemiAnalysis_
The margin on a subscription plan is a function of the average utilization. If we assume both companies have 75% API gross margins, this results in the following subscription margins. (3/4)
222K
SemiAnalysis
@SemiAnalysis_
Jun 10
Replying to @SemiAnalysis_
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found
2.7M
SemiAnalysis
@SemiAnalysis_
Jun 10
What's the better business model for an AI lab, subscription or API? (1/4)🧵
132K
SemiAnalysis
@SemiAnalysis_
Jun 10
DeepSeek is going heavy-asset. On June 9, the company posted an opening for IDC planning engineers, a role explicitly scoped to the design and delivery of MW-to-GW scale infrastructure. It follows April's hiring of data center O&M engineers in Ulanqab, Inner Mongolia. Taken
102K
SemiAnalysis
@SemiAnalysis_
Jun 10
Replying to @SemiAnalysis_
Full breakdown with Jordan Nanos (@JordanNanos), Jeremie Eliahou Ontiveros (@JeremieEO), Joey Brookhart (@SaasquatchC), and Crystal Huang (@egg1459) on SemiAnalysis Weekly:
18K
SemiAnalysis
@SemiAnalysis_
Jun 10
AI market beat 2025 expectations by massive margins. But here's what's wild about the composition of this beat.
00:00
24K
SemiAnalysis
@SemiAnalysis_
Jun 10
HISTORY LESSON: In 1968 the US, USSR, UK, France, and China signed the Nuclear Non-Proliferation Treaty, declaring nuclear weapons too dangerous for any more countries to build. All five already had them. Everyone else had to submit to inspections while the cohort pinky-promised
NomoreID
@Hangsiin
Jun 9
When Fable 5 is used for frontier LLM development, it does not notify the user and instead limits the model’s capabilities through methods such as prompt modification, steering vectors, and PEFT. Anthropic estimated that this would affect approximately 0.03% of traffic.
194K
SemiAnalysis
@SemiAnalysis_
Jun 10
Replying to @SemiAnalysis_
And every next-gen win in inference is a datacenter win. CPO, copper backplanes, NVL scale-up domains, better pJ/bit, better perf/watt — none of that ships in a laptop chassis. The steel mill gets cheaper per ton every year. The village furnace can only get so hot! (4/4)
18K