Productivity 19-Year-Old Founder Emerges From Stealth With $11 Million to Sell You a $3,499 'Brain in a Box' That Runs Your AI Agents at Home AI Agents Half a Million Interviews In: HackerRank's AI Interviewer Chakra Goes GA, and It Wants to Replace Three Hiring Rounds With One Security After Claude Agents Escaped Its Sandbox 3 Times, Anthropic Deploys Real-Time Classifiers to Stop the Next Escape Before It Happens Business Meta Halves Its Internal Claude Users to 30,000 and Microsoft Slashes a $1 Billion Anthropic Budget by More Than a Third Business Sony Innovation Fund Backs Primitive Labs, a Startup That Builds Simulated Crowds to Stress-Test Products Before Launch Apple Intelligence Apple Removed the Apple Intelligence Off Switch in macOS 27 — So a Developer Built a CLI That Deletes It Anyway Security OpenAI Turns On Invisible Text Watermarks for ChatGPT in the EU — and Publishes Exactly How Weak They Are News A Mystery 'Space Bunny Alpha' Model Just Topped OpenRouter's Leaderboard With 38.7 Trillion Tokens a Week Productivity 19-Year-Old Founder Emerges From Stealth With $11 Million to Sell You a $3,499 'Brain in a Box' That Runs Your AI Agents at Home AI Agents Half a Million Interviews In: HackerRank's AI Interviewer Chakra Goes GA, and It Wants to Replace Three Hiring Rounds With One Security After Claude Agents Escaped Its Sandbox 3 Times, Anthropic Deploys Real-Time Classifiers to Stop the Next Escape Before It Happens Business Meta Halves Its Internal Claude Users to 30,000 and Microsoft Slashes a $1 Billion Anthropic Budget by More Than a Third Business Sony Innovation Fund Backs Primitive Labs, a Startup That Builds Simulated Crowds to Stress-Test Products Before Launch Apple Intelligence Apple Removed the Apple Intelligence Off Switch in macOS 27 — So a Developer Built a CLI That Deletes It Anyway Security OpenAI Turns On Invisible Text Watermarks for ChatGPT in the EU — and Publishes Exactly How Weak They Are News A Mystery 'Space Bunny Alpha' Model Just Topped OpenRouter's Leaderboard With 38.7 Trillion Tokens a Week

Cerebras Sinks 20% to a Post-IPO Low as Report Says OpenAI's Ultrafast Mode Runs on Nvidia Instead

Cerebras Sinks 20% to a Post-IPO Low as Report Says OpenAI's Ultrafast Mode Runs on Nvidia Instead

Cerebras Systems closed Friday at $166.43, down about 20% for the week and at its lowest level since its May IPO, after research firm SemiAnalysis reported that OpenAI's Ultrafast mode for GPT-6.1 Sol will run on Nvidia GPUs rather than Cerebras hardware. The stock also absorbed 19.4 million newly unlocked insider shares, and is now worth just over $39 billion — less than half its first-day market cap.

Cerebras Systems, the maker of dinner plate-sized AI inference chips that staged one of 2026's most explosive public offerings, just got a brutal reminder of how fast Wall Street can turn. The stock closed Friday at $166.43, down roughly 20% for the week and sitting at its lowest level since the company's Nasdaq debut on May 14.

The immediate trigger was a report from the research firm SemiAnalysis, which posted on X on Wednesday that OpenAI will power the "Ultrafast" mode of its GPT-6.1 Sol model with Nvidia graphics processing units instead of Cerebras hardware. Ultrafast, announced at OpenAI's DevDay as a roughly 300-token-per-second speed tier, had been one of the marquee showcases for Cerebras' speed-first architecture — and investors had treated the company's relationship with OpenAI as its core asset.

That relationship runs deep. In January, Cerebras struck a deal worth more than $10 billion with OpenAI to supply 750 megawatts of computing power through 2028, leasing capacity from its own data centers as a cloud service. Losing a visible piece of OpenAI's inference stack to Nvidia, even while the broader contract stands, has put a damper on the thesis that Cerebras could become the credible non-Nvidia option for frontier labs.

OpenAI's CEO moved quickly to calm the waters. Responding to what he called "speculation about our partnership with Cerebras," Sam Altman wrote on X: "Cerebras is a close partner, and we have a deep engagement pushing on the frontiers of speed." The stock rose almost 3% in extended trading on Friday after the post, but that only partly claws back the week's damage. Shares are now down by more than half from their initial post-IPO pop, when the company closed its first day at a $95 billion market cap — it now sits at just over $39 billion.

Timing made everything worse. Post-IPO restrictions on insider shares expired this week: according to the prospectus, up to 19.4 million shares held by directors, officers, employees and other early holders unlocked on Wednesday, equal to 8% of shares outstanding. That follows a schedule of up to 14.6 million shares unlocking every two weeks since August 19. CEO Andrew Feldman and CTO Sean Lie, both minted billionaires by the IPO, sold over $240 million of Class A shares between August 20 and September 25 under trading plans adopted shortly after listing.

There is a structural question underneath the price action. McKinsey projects that inference will surpass training as the dominant workload in AI data centers by the end of the decade — the exact market Cerebras was built to win with wafer-scale ASICs that lease speed rather than sell chips. But Nvidia has been defending that same territory, and the SemiAnalysis report suggests that when OpenAI needed its fastest mode, it chose the incumbent.

Cerebras still has its $10 billion OpenAI commitment, a differentiated architecture, and a market cap that has already corrected by more than half. What it no longer has is the benefit of the doubt. The next test will be whether the company can point to production inference wins beyond OpenAI — because the market has just shown what happens to "Nvidia alternative" stories when a headline says the alternative wasn't used.

Comments (0)

Log in to join the discussion

Log In

No comments yet