OpenAI Doubles Cerebras Chip Deal to $20B, Takes Equity Stake

OpenAI Commits Over $20 Billion to Cerebras Chips

OpenAI has agreed to spend more than $20 billion on servers powered by Cerebras Systems chips over the next three years, according to a report from The Information published April 17. The deal doubles the size of a January agreement valued at more than $10 billion and represents the largest non-Nvidia AI infrastructure contract ever signed.

Under the expanded arrangement, OpenAI will receive equity warrants in Cerebras that could give it up to a 10% ownership stake in the chipmaker as spending commitments increase. OpenAI has also agreed to provide Cerebras approximately $1 billion in funding to help develop data centers that will run OpenAI's AI products, further tightening the relationship between the two companies.

Cerebras Files for IPO at $35 Billion Valuation

The expanded deal comes as Cerebras moved to go public, filing its S-1 registration statement on April 17 with plans to list on the Nasdaq under the ticker symbol CBRS. The company is targeting a valuation of approximately $35 billion and plans to raise around $3 billion in the offering, with Morgan Stanley as lead underwriter. Cerebras was last privately valued at $23.1 billion during its Series H round in February 2026.

The OpenAI partnership is central to the IPO narrative. A $10 billion multi-year compute agreement was already the largest non-Nvidia AI infrastructure contract, and doubling it to $20 billion-plus fundamentally changes the revenue diversification story that investors will scrutinize. Oracle also namedropped Cerebras alongside Nvidia and AMD during its March 2026 earnings call, confirming that Oracle Cloud Infrastructure runs Cerebras hardware for customer workloads.

The Wafer-Scale Chip Challenging Nvidia

Cerebras builds what CEO Andrew Feldman calls "the fastest AI hardware for training and inference." The company's flagship WSE-3 wafer-scale chip contains 4 trillion transistors and 900,000 cores, occupying an entire silicon wafer rather than being cut into individual chips. Cerebras claims the WSE-3 delivers 21 times the performance of Nvidia's DGX B200 system at one-third the cost and power consumption.

"We're building the infrastructure layer that AI needs to scale to the next level. This partnership with OpenAI validates the wafer-scale approach." — Andrew Feldman, CEO, Cerebras Systems

The technology appeals to companies seeking alternatives to Nvidia's dominant GPU ecosystem, particularly for inference workloads — the process of running trained AI models in production. While Nvidia's GPUs were effectively repurposed from gaming for AI training, newer architectures like Cerebras' promise faster responses, lower energy use, and significantly lower costs for the inference stage that drives the majority of production AI compute.

Broader AI Chip Startup Boom

The deal lands amid a record year for AI chip investment. Startups in the sector have raised $8.3 billion globally in 2026, according to Dealroom. MatX, Ayar Labs, and Etched have each secured $500 million rounds, while European challengers including Axelera and Olix have raised over $200 million each. CNBC reported on April 17 that Dutch startup Euclyd, backed by the former CEO of ASML, is seeking at least 100 million euros for chips it claims deliver 100 times the power efficiency of Nvidia's latest Vera Rubin GPUs.

Nvidia itself has responded aggressively to the competitive threat, acquiring assets from inference startup Groq in a $20 billion deal in December and committing $4 billion to companies working on photonics. The AI chip race is intensifying precisely because the industry's focus is shifting from training — where Nvidia remains dominant — to inference, where architectural alternatives have a genuine shot at disrupting the status quo.

What This Means for Engineers and Job Seekers

The proliferation of AI chip architectures beyond Nvidia is creating new career opportunities in hardware-software co-design, compiler optimization, and systems engineering. Engineers with experience in custom silicon, ASIC design, or inference optimization are among the most sought-after hires in the industry. For software engineers, the growing diversity of AI hardware means that skills in model optimization, quantization, and multi-platform deployment are becoming increasingly valuable as companies evaluate alternatives to the Nvidia-CUDA ecosystem.