Cerebras Stock Rebounds as Sam Altman Reaffirms OpenAI Partnership Amid AI Inference Battle

Sam Altman Reassures Investors on Cerebras Partnership

Cerebras Systems stock climbed more than 6% in premarket trading Monday after OpenAI CEO Sam Altman publicly addressed speculation about the companies' relationship1

1

2

2

. Altman posted on X Friday that Cerebras is "a close partner" and emphasized the two companies have "a deep engagement pushing on the frontiers of speed"3

3

. The statement came after Cerebras stock lost 20% of its value last week and touched its weakest level since the AI chipmaker listed on the Nasdaq, following reports that OpenAI had selected Nvidia graphics processing units to run the Ultrafast mode in GPT-6.1 Sol1

1

. Cerebras shares rose almost 3% in extended trading Friday following Altman's comments and were last up 6.3% in premarket Monday at $177 per share1

1

.

Fast Inference Emerges as Key Battleground in AI Computing Market

The market turbulence highlights a fundamental shift in how AI hardware competition is evolving. The next big AI chip battle may not be about who trains the biggest model but about who can make AI respond fastest once the model is already built2

2

. This is putting Cerebras Systems and Nvidia on opposite sides of an increasingly valuable fight over AI inference, the computing required every time a user asks a question, generates code, or interacts with an AI agent2

2

. Mizuho analyst Vijay Rakesh estimates the fast-inference market could reach $550 billion by 2030, representing roughly 20% of AI workloads2

2

5

5

. Fast inference has become a key revenue focus, with industry token pricing running 6 to 10 times higher than standard tokens5

5

.

Cerebras Builds Specialized Position Against Nvidia's Broad Ecosystem

Cerebras builds specialized chips and AI computing systems engineered to process AI workloads more quickly than conventional graphics processing units1

1

. Under a $10 billion agreement struck with OpenAI in January, Cerebras is contracted to deliver 750 megawatts of computing capacity over the period ending in 20281

1

. OpenAI also gave Cerebras a $1 billion working-capital loan, and the deal has a value above $20 billion4

4

. In August, the company launched its CS-4 system, which it says generates answers up to 30 times faster than GPU-based options from rivals like Nvidia3

3

. It reached over 4,400 tokens per second per user on GPT-OSS-120B, about twice its predecessor's speed3

3

.

Analysts See Complementary Roles for Cerebras and Nvidia

Barclays analyst Tom O'Malley said OpenAI's "relationship remains strong" with Cerebras, pointing out that while Nvidia is running OpenAI's new GPT-6.1 Sol Ultrafast model, Cerebras remains the preferred provider but is constrained by supply3

3

. O'Malley called the multi-day sell-off in Cerebras shares "overdone," saying investors had overreacted to concerns3

3

. Analyst Paul Meeks noted that as Cerebras specializes in fast AI inference, Nvidia chips may handle the initial "prefill" phase while Cerebras chips could still be needed for the decode stage3

3

. Cerebras has formed ties with AMD and AWS, with AMD handling the prefill stage which processes the prompt and prepares the model state, while Cerebras handles the decode stage which creates output tokens4

4

.

Revenue Growth and Market Challenges for the AI Chipmaker

Cerebras posted $210 million in core revenue for the second quarter of 2026, more than twice the prior-year level, with cloud business growing 281% year over year4

4

. The company raised its 2026 core revenue forecast to $880 million to $890 million and reported $25.4 billion in future performance obligations4

4

. More than 600 megawatts of data-center capacity now sit live or under contract4

4

. Despite strong growth metrics, the stock is now down nearly half from its post-IPO high, and the company's market cap has fallen to just over $39 billion from $95 billion at its May debut1

1

. Shares had also faced a supply-related overhang, with a 19.4 million-share unlock adding stock to the market last week3

3

. Cerebras' IPO lockup is structured as 11 staggered releases, including three larger waves, with others due October 14 and October 283

3

.

What Investors Should Watch

Mizuho reiterated an Outperform rating and $300 price target on Cerebras Systems, projecting AI fast inference data center deployments could drive company revenue to $13.5 billion by 20295

5

. According to analysts at Citi, projections for Cerebras' revenue through 2028 have not shifted, noting that frontier-AI labs' latest models would initially roll out on internal chips before running on third-party or Cerebras cloud1

1

. The analysts indicated that any additional slippage before gross margins bottom out would likely dampen investor enthusiasm, especially considering how richly Cerebras' shares are valued, and that a sustained recovery in the stock depends on concrete signs that margins are finding a floor1

1

. The next proof points for Cerebras are capacity deployment, new customers beyond OpenAI, and whether the company's speed advantage translates into improving margins2

2

. Andrew Feldman's strategy could reshape how companies handle AI inference, response speed, and computing costs, potentially affecting Nvidia's market position, cloud infrastructure, chip design, and the way major AI companies build future systems4

4

.