“I personally think there will be huge demand for faster inference across all parts of the AI economy. … Companies are paying disproportionately more for faster inference and this is good for Cerebras.”
Faster inference commands outsized pricing; Cerebras owns 1% of a huge market for fast inference at small model sizes
“they kind of committed to this dead-end process by having the biggest scale-up world as a wafer size … the models got much bigger than just a single wafer. And so they have really fast inference, but ”
Wafer-size ceiling caps Cerebras's fast-inference edge as models outgrow one wafer, pressuring a rich IPO.
“In a compute shortage demand overflows to 2nd/3rd best chips, but Nvidia's general-purpose GPU still wins most cases”
In a compute shortage demand overflows to 2nd/3rd best chips, but Nvidia's general-purpose GPU still wins most cases
“AMD is definitely good enough right now”
AMD is good enough now on rack scale-up; Lisa Su figures out MI450 and may ship a fast SRAM offload inference chip within 12 months
“TSMC is kind of a kingmaker in terms of supply … I'm going to expect supply to kind of lag over and over and over. But demand signals will continue to essentially command premiums, move up wafer prici”
TSMC is the supply kingmaker; cleanroom lead times keep supply lagging demand, sustaining premium wafer pricing for years
“Wix is down a bunch. This seems like a very logical company to suffer in the age of vibe coding.”
Wix is structurally exposed to vibe coding; down 45% since a $92 buyback, the website-builder model is collateral damage
“the notion that CUDA is sort of this grand lock-in is overplayed … the Gemini 3”
If CUDA lock-in is overplayed and TPU/Trainium/Cerebras are credible, NVIDIA's pricing power and premium multiple compress.
“46% year-over-year revenue growth accelerating for the second straight quarter.”
Figma revenue growth is accelerating, 46% YoY for a second straight quarter, with guidance raised
“You know, the folks that are attracted to both of those areas tend to be very technology-driven. They love distributed systems. They love the hard problems…”
AI and crypto share systems-thinking talent; high-throughput L1 demand as AI-native settlement lifts SOL.