Can Anyone Dethrone Nvidia? Cerebras Is About to Find Out
Cerebras files for IPO with a $20B OpenAI deal in hand. What does this mean for Nvidia's dominance, AI infrastructure investment, and the next wave of chip competition?
Every few years, someone declares Nvidia's AI chip monopoly is over. This time, they're bringing a $20 billion contract as evidence.
What's Happening
Cerebras Systems is filing its S-1 IPO paperwork on Friday, April 18 — its second attempt after pulling a previous filing in late 2024 to update financial disclosures. The Sunnyvale, California-based chip startup was last valued at $8.1 billion in a September 2025 funding round that raised $1.1 billion.
But the number that's turning heads is bigger. OpenAI has expanded its deal with Cerebras to over $20 billion, up from an initial agreement worth more than $10 billion announced in January. Under that earlier deal, Cerebras committed to providing OpenAI with up to 750 megawatts of computing capacity through 2028. The new arrangement also gives OpenAI warrants to purchase Cerebras shares — making the world's most prominent AI lab a prospective shareholder in its own chip supplier.
Oracle is circling too. CEO Clay Magouyrk name-dropped Cerebras on Oracle's March earnings call as part of its cloud chip lineup, though Cerebras hasn't yet appeared on Oracle's official price list. A formal partnership could be imminent.
Why Cerebras Is Different — and Why That Matters
Most of the AI world runs on Nvidia GPUs. AMD has carved out a growing slice. Cerebras is doing something structurally different: its Wafer Scale Engine (WSE) treats an entire silicon wafer as a single massive chip, rather than cutting wafers into smaller units. The result is dramatically faster inference — the speed at which an AI model generates responses to user queries.
That speed advantage is why OpenAI is already using Cerebras infrastructure to power a coding tool. And it's why Cerebras has pivoted from selling chips outright to running its own data centers and selling cloud-based compute as a service — a model that generates recurring revenue and ties customers in more deeply.
Founder and CEO Andrew Feldman has been here before. He sold server startup SeaMicro to AMD for $355 million in 2012. In 2018, he reportedly fielded an acquisition approach from Elon Musk. His current investors include Sam Altman, who wears two hats here as both OpenAI CEO and Cerebras backer.
The Bigger Play: AI IPO Season Is Opening
The U.S. tech IPO market has been largely quiet since 2022. Retail and institutional investors have been waiting for high-growth AI companies to go public. With Anthropic and OpenAI both reportedly weighing IPOs this year, Cerebras' S-1 filing could mark the beginning of a new wave.
For investors, the calculus is layered. The $20 billion OpenAI contract sounds transformative — but how much of Cerebras' revenue is OpenAI-dependent? A company whose fortunes are tied to a single customer (one that also holds warrants in it) carries concentration risk that any serious investor should price in. The S-1, once public, will be required reading.
For the broader chip industry, the implications are structural. Every dollar of compute that flows to Cerebras rather than Nvidia is a data point in the argument that AI infrastructure is diversifying. AMD, Intel, and a growing roster of custom silicon startups are all making the same bet. Whether any of them can build the software ecosystem and manufacturing scale to truly challenge Nvidia's moat is the question the market will be pricing for years.
This content is AI-generated based on source articles. While we strive for accuracy, errors may occur. We recommend verifying with the original source.
Related Articles
TSMC posted a 58% profit jump and its fourth consecutive record quarter. As AI chip demand reshapes the semiconductor industry, here's what it means for investors, competitors, and the global tech supply chain.
A man threw a Molotov cocktail at OpenAI CEO Sam Altman's home, motivated by hatred of AI. His document listed names and addresses of multiple AI executives. This isn't just a crime story.
Alibaba and China Telecom launched a 10,000-chip AI data center in Guangdong powered by Alibaba's homegrown Zhenwu semiconductors. What does China's accelerating chip self-sufficiency mean for Nvidia, global AI competition, and your portfolio?
OpenAI's 13-page policy blueprint proposes robot taxes, a public wealth fund, and a four-day workweek. Is this corporate responsibility — or regulatory capture in disguise?
Thoughts
Share your thoughts on this article
Sign in to join the conversation