Cerebras Systems (CBRS) shares climbed approximately 6% in premarket trading on Monday, recovering part of last week's steep losses after OpenAI CEO Sam Altman moved to reassure investors that the AI chipmaker remains a key partner. The rebound follows a 20% plunge last week, triggered by reports that OpenAI would deploy Nvidia graphics processing units for the "Ultrafast" mode of its GPT-6.1 Sol model, rather than Cerebras chips.
In a post on X on Friday, Altman addressed the speculation directly: "There is some speculation about our partnership with Cerebras. Cerebras is a close partner, and we have a deep engagement pushing on the frontiers of speed." The statement helped calm investor fears, though Cerebras shares remain more than 40% below their May Nasdaq debut price.
Analysts See Selloff as Overdone
Barclays analyst Tom O'Malley said the relationship between OpenAI and Cerebras "remains strong," arguing that concerns about the partnership may have been exaggerated. While Nvidia is powering GPT-6.1 Ultrafast, Cerebras remains the preferred provider for certain workloads, though its ability to meet demand is constrained by supply. O'Malley described the recent selloff as "overdone as highlighted by Sam Altman's post."
Freedom Capital upgraded Cerebras to Buy from Hold last week, calling the market's reaction an overreaction. Analyst Paul Meeks set a price target of $209, implying more than 25% upside from Friday's close. Citi analysts also maintained their revenue outlook for Cerebras for 2026 through 2028, stating it remains "unchanged." They noted that frontier-AI labs' latest models would initially run on internal chips before moving to third-party or Cerebras cloud, so it is too early to draw conclusions.
However, Citi cautioned that the stock's next leg higher could depend on improvements in profitability. "We believe the stock's ability to outperform is increasingly tied to evidence that gross margins are stabilizing. Any further delay in the gross margin trough would likely weigh on sentiment, particularly given Cerebras' premium valuation," they wrote.
Cerebras Bets on Fast AI Inference
Cerebras has positioned itself as an alternative to Nvidia by developing large-scale processors and AI systems designed to run models faster than traditional GPUs. The company claims its flagship Wafer Scale Engine 3 delivers faster performance than Nvidia GPUs for certain AI workloads. Its relationship with OpenAI is particularly significant, underscored by a $10 billion agreement signed in January to provide 750 megawatts of computing power through 2028.
Mizuho reiterated its Outperform rating and $300 price target, pointing to the growing market for fast AI inference workloads. The firm estimates the fast-inference total addressable market could reach $550 billion by 2030, representing a 291% compound annual growth rate and about 20% of all AI workloads. Fast inference has become an important revenue opportunity for Cerebras because industry token pricing is reportedly six to ten times higher than standard tokens.
Potential applications range from trading and defense to software development, where faster AI responses could reduce coding time. OpenAI has reported inference speeds up to 14 times faster using Cerebras technology, reinforcing the company's position as a partner focused on AI response speed. Mizuho also sees Meta Platforms as a potential future inference customer.
Despite potential share-unlock pressures, Mizuho projects that AI fast-inference data-center deployments could help Cerebras generate $13.5 billion in revenue by 2029.
This article is for informational purposes only and does not constitute financial advice.
