Cerebras Files for IPO at $35B Valuation After Securing $20B OpenAI Chip Deal

Cerebras Charges Toward a Public Listing

On April 18, 2026, AI chip startup Cerebras Systems officially filed for an initial public offering, targeting a valuation of approximately $35 billion and a listing date in mid-May 2026. The announcement marks the company's second attempt at going public after withdrawing an earlier IPO in 2024 due to a federal review of an investment from Abu Dhabi-based G42. This time, however, Cerebras arrives at the market with substantially stronger financials and a transformative deal with OpenAI anchoring its growth story.

The OpenAI Deal at the Center of It All

The IPO filing comes days after reports confirmed that OpenAI has committed to spending more than $20 billion on Cerebras chips over the next three years — more than double the $10 billion compute agreement the two companies signed in January 2026. Under the expanded terms, OpenAI will gain warrants for a minority equity stake in Cerebras, with potential ownership rising to roughly 10 percent as spending increases. OpenAI has also agreed to provide approximately $1 billion to fund Cerebras data center development dedicated to running its products.

CEO Andrew Feldman described the relationship as Cerebras capturing "the fast inference business at OpenAI," positioning the company as the key provider for OpenAI's latency-sensitive workloads where Nvidia's GPUs have traditionally dominated.

Financial Profile: Profitable on a GAAP Basis

Cerebras's S-1 filing revealed a financially unusual profile for an AI hardware startup. The company reported 2025 revenue of $510 million and a GAAP net income of $237.8 million — a rare profit for a chip company competing with Nvidia. On a non-GAAP basis, the company logged a net loss of $75.7 million, reflecting stock-based compensation and other adjustments. The company has raised approximately $2.1 billion in private funding, including a $1.1 billion Series G in 2025 and a $1 billion Series H in February 2026, which last valued it at $23 billion.

In addition to the OpenAI deal, Cerebras has secured an agreement to supply chips for Amazon Web Services data centers, further broadening its customer base beyond a single hyperscaler dependency.

The CS-3 Chip: A Different Bet on AI Hardware

Cerebras differentiates itself through its wafer-scale engine approach. Rather than linking hundreds of discrete GPU chips with high-bandwidth interconnects, Cerebras manufactures a single chip the size of an entire silicon wafer. The result is a processor with far more on-chip memory and dramatically lower inter-chip communication overhead — properties that translate into significantly faster inference speeds for large language models.

For AI inference use cases — where a model responds to user queries in real time — Cerebras claims throughput advantages that are difficult for conventional GPU clusters to match at equivalent latency targets. This has made the company particularly attractive to customers building real-time AI products rather than batch training workloads.

Competitive Landscape and Risks

Cerebras enters the public markets amid intensifying competition in the AI chip sector. Nvidia remains the dominant provider for AI training and retains a massive software ecosystem advantage through CUDA. Newer challengers including Groq, Tenstorrent, and SambaNova are also targeting inference workloads. AMD continues to compete in both training and inference through its MI300X line.

The heavy concentration of revenue in a single customer relationship with OpenAI also represents a material risk factor. If OpenAI's spending ramps more slowly than projected, or if the company shifts its infrastructure strategy, Cerebras's growth trajectory could be materially affected. The federal scrutiny that derailed the company's 2024 IPO, while resolved, highlighted regulatory risk tied to foreign investment exposure.

Outlook: AI Infrastructure Investment Cycle Continues

The Cerebras IPO arrives at a moment when AI infrastructure spending is accelerating across hyperscalers and frontier AI labs. OpenAI's decision to anchor its fast-inference workloads with Cerebras rather than exclusively with Nvidia signals a meaningful shift in how leading AI companies are thinking about compute procurement — prioritizing inference speed and cost efficiency alongside raw training performance.

If the IPO prices as expected, Cerebras would join a small group of publicly traded pure-play AI semiconductor companies, giving investors direct exposure to the inference layer of the AI stack at a moment when inference is consuming an increasingly large share of total AI compute spend.

Conclusion

Cerebras's IPO filing is one of the most significant AI infrastructure finance events of 2026. For investors and industry observers, it represents both a validation of the inference-optimized chip thesis and a real-time gauge of how the market values AI hardware companies outside the Nvidia orbit. Engineers, AI product teams, and infrastructure architects tracking next-generation compute options will want to watch this listing closely.