Latest / Minds, Bodies, and Terawatts / Cerebras AI Chipmaker Goes Public with $95B Market Cap
Transcript
- 0:00[Host] This is Minds, Bodies, and Terawatts, where silicon meets the stock ticker. Cerebras went public on Nasdaq Thursday and closed at a ninety-five billion dollar market cap.
- 0:14[Guest] Priced the offering at a hundred eighty-five a share, closed at three-eleven. Twenty times oversubscribed. Wall Street isn't betting on AI software anymore, it's betting on the physical layer underneath it.
- 0:30[Host] For listeners who don't follow chips, what does Cerebras actually make?
- 0:36[Guest] CEO Andrew Feldman stood on the Nasdaq floor Thursday and put it simply: we built a chip the size of a dinner plate, fifty-eight times larger than anything previously made. Instead of cutting a silicon wafer into hundreds of small dies, Cerebras keeps the entire wafer intact. One processor, purpose-built for AI inference.
- 1:04[Host] So why did investors pile in at this scale?
- 1:08[Guest] The hyperscaler spending wave. Amazon, Google, Meta, and Microsoft are projected to spend over six hundred billion on AI infrastructure this year, according to Morgan Stanley. Cerebras is positioning as the alternative for customers who need supply chain diversification away from Nvidia.
- 1:30[Host] But here's what I don't understand. Cerebras pitches itself as a US sovereignty play, yet TSMC still fabricates their wafers in Taiwan on the same fabs Nvidia uses. Where's the domestic supply chain?
- 1:48[Guest] That's the honest gap in the story right now. The chip design is American, system assembly is domestic, but fabrication still routes through Taiwan. The bet investors are making is forward-looking: TSMC's Arizona fabs come online over the next two years, and Cerebras is first in line to shift production stateside. They're buying the trajectory, not today's reality.
- 2:18[Host] Nvidia controls roughly eighty-five percent of the AI training chip market. How does a challenger justify this valuation without matching Nvidia's software ecosystem?
- 2:30[Guest] It doesn't have to match it today. Twenty-times oversubscription tells you about institutional anxiety, not current capabilities. When four companies spend six hundred billion a year and the primary chip supplier sits a hundred miles from mainland China, hedge funds price in contingencies.
- 2:53[Host] DeepSeek's Liang Wenfeng said it bluntly: money was never our problem, chip bans are. Export controls create guaranteed demand for any alternative that Washington won't restrict.
- 3:07[Guest] Which is why sovereign wealth funds piled into this IPO. Compute sovereignty is becoming national security doctrine dressed in a stock ticker. The two-ecosystem split between American and Chinese AI stacks is no longer a scenario. It's priced into the cap table.
- 3:29[Host] Patrick Deglon maps this as the physical layer of what he calls The Brain in Unscarcity, the compute hardware that makes artificial intelligence possible. When one island can throttle that entire layer, the IPO isn't a stock pick. It's a geopolitical instrument.
- 3:51[Guest] For more on the compute arms race and where the next bottleneck forms, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.