Last week, Cerebras got hit twice at once. Research firm SemiAnalysis posted on X that OpenAI would power its “Ultrafast” mode for GPT-6.1 Sol with Nvidia graphics processing units instead of Cerebras hardware. That was a real, verifiable competitive loss, at least for that specific Ultrafast serving configuration. Then, almost simultaneously, a new wave of shares hit the market: Cerebras structured its post-IPO lockup as 11 separate releases rather than one 180-day freeze, and last week’s tranche freed up to about 19.4 million shares, the first of three similarly sized waves. Two bearish catalysts arriving together produced a predictable result.
CBRS fell nearly 20% and closed Friday at $166.43, its lowest close since the May 14, 2026 IPO and below the $185 offer price. Cerebras’ market cap fell to about $43.6 billion as of October 5, 2026, down from roughly $95 billion at its first-day close. For a stock that closed its first trading day at $311.07, a 68% gain from the IPO price, the descent had been relentless.
Then Monday arrived. Sam Altman posted on X that “there is some speculation about our partnership with Cerebras,” adding: “Cerebras is a close partner, and we have a deep engagement pushing on the frontiers of speed.” Cerebras stock climbed about 9% on Monday, rebounding from last week’s decline. Thirteen words from one CEO, and a week’s worth of anxiety evaporated in a session.
Here is the analytical problem that bounce creates. The Nvidia news was concrete. OpenAI chose a competitor’s silicon for a specific, named product tier. The Altman post was reassuring but structurally vague: it confirmed ongoing engagement without specifying what workloads Cerebras actually holds. Barclays analyst Tom O’Malley called the multi-day sell-off “overdone,” saying investors had overreacted to concerns that OpenAI is using Nvidia GPUs for its new model. That may be right. It may also be a rationalization written after the fact.
What deserves more weight is the contract underneath the noise. OpenAI and Cerebras have a confirmed multi-year deal under which OpenAI would purchase up to 750MW of compute capacity, with tranches coming online through 2028. The Wall Street Journal reported the agreement at around $10 billion, and Reuters reported it as a $10 billion deal. That deal is real and still running. Losing the Ultrafast inference workload to Nvidia does not cancel it. But it does raise a legitimate question about how many workloads Cerebras actually retains as OpenAI’s model portfolio expands.
Experienced traders in situations like this separate the two variables cleanly. The Nvidia win was fundamental news. The lockup selling was mechanical: additional tranches of similar size are scheduled for October 14 and October 28, with the full lockup expiring no later than November 9. Forced sellers do not care about valuation. They sell because the calendar tells them they can. Conflating their exits with a fundamental verdict on Cerebras is a mistake that turns into missed opportunity or premature re-entry, depending on which error you make.
The 9% bounce on Monday fits a well-worn pattern: a crowded short position meets a headline that does not confirm the bear thesis, and the covering is sharp. Altman’s comments may ease concerns over the OpenAI relationship, but investors are likely to remain focused on whether Cerebras can continue securing high-speed inference workloads as competition from Nvidia intensifies. That question has not been answered by a social media post.
The trader’s lesson here is about supply and signal. When a lockup release and a competitive setback land in the same week, the price action tells you almost nothing useful about intrinsic value. The stock had to go lower because sellers were forced in. Whether Monday’s bounce reflects genuine re-accumulation or short covering around the Altman comment is only legible if you watch how CBRS behaves in the sessions around October 14 and October 28. If buyers absorb those tranches on solid volume, that is evidence. One compliment is not.