Thursday, November 5, 2026

Back to the front page

From X

@convequity

Convequity on X

OpenAI’s TPU lesson makes me more cautious on $CBRS. Richard Ho, OpenAI’s hardware chief and a former TPU designer, says one lesson from Google was that “you have to make it programmable.” Models change too quickly to overfit the chip to one architecture. That caught my attention because TPU is already programmable. Groq statically schedules execution in its compiler, while Cerebras makes a different bet on wafer-scale compute and distributed local memory. The question is how much of the performance advantage survives when models change, including MoE routing and memory demands. Ho also wants hardware that can move across inference workloads without locking the fleet into fixed ratios of specialized machines. That matters when buyers are committing years of capex and power. If OpenAI and other large buyers increasingly favor that flexibility, I expect CBRS to remain a specialist over the next few years. The absolute market can still be large enough for Cerebras to build a very good business, but I’m more skeptical that its architecture becomes the mainstream choice. "As long as you’re willing to live with those limitations, they’re good. But if you’re able to make it programmable, then you have a lot of flexibility for adapting to change. And we had to adapt to change. TPU also adapted to different changes in the algorithm, even at that time with CNNs. Clearly now, in the world of LLMs and agents, the mix is changing pretty rapidly. So that was the pitfall we really wanted to avoid: to overdesign and overfit to one particular model. We looked at it in that context." ---[54] The Man Who Built OpenAI's First Chip TechTechPotato

Trading Post November 5, 2026

MercadoLibre’s logistics expansion cut margins by 60 basis points, while Mercado Pago grew credit 62% with 7% NPLs.

Intel: The CPU Comeback, Discounted

Intel's Q3 2026 results were hit by memory costs and an Israel shutdown, but enterprise AI inference could lift CPU demand as server…

Problems at Twist Bioscience (TWST)

Twist Bioscience’s $200 rebound followed an Anthropic-linked short squeeze, yet its DNA and protein services face broad competition while…

VSH Earnings Review

Vishay reported record Q3 revenue of $979 million, up 24%, and guided Q4 to about $990 million as rising backlog and margins contrast with…

Thursday Tailwinds #13: Nov 5, 2026

Florida marinas are 92% full as boat counts rise and permits constrain new supply, creating pricing power for existing operators such as…

eToro: Expanding Beyond Crypto

EToro trades at 2.89x EV/FCF with $304 million in trailing free cash flow, expanding from crypto into equities, commodities, currencies and…

SILC · long

Silicom LTD

Silicom (SILC) is pitched as inexpensive on its existing hardware business, with upside framed as a cheap call option.

Research Note

Nine Mile Metals is the subject of a research note.

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

X accounts we follow

Every account on this site, hand-picked.

91 accounts

Investor letters we follow

359 firms

Sectors

Ordered by how often our editor reaches for them.