Cerebras Startup Files IPO Application: AWS and OpenAI Deals Worth $10 Billion

American startup Cerebras Systems has filed documents for an IPO — its debut on the public market could become one of the most significant in the AI-semiconductor sector in recent years. The company has chosen the timing against the backdrop of two strategic deals that have confirmed its position in the industry. Cerebras was founded in 2016 in Silicon Valley and specializes in developing chips for accelerating neural networks.

Its flagship product is the Wafer Scale Engine (WSE), a microchip that occupies the area of an entire silicon wafer. In terms of size and transistor count, the WSE vastly exceeds standard GPUs: the WSE-3 version contains 4 trillion transistors compared to 80 billion in Nvidia's flagship H100 chip. The WSE architecture enables data processing with minimal latency — all computations occur on a single substrate without the need to transfer data between separate chips via slow buses.

The company positions its solutions as an alternative to Nvidia's GPU clusters for training and inference of large language models. Instead of thousands of individual GPUs connected via high-speed networks, Cerebras offers several WSE chips that deliver comparable performance with reduced communication overhead. The company claims its systems can reduce training time for some models by tens of times.

Before filing for IPO, Cerebras concluded two strategic agreements. The first is with Amazon Web Services: the company's chips will be integrated into AWS infrastructure, giving the startup access to enterprise clients worldwide through the platform of the world's largest cloud provider. This significantly shifts the balance of power: until now, AWS had primarily relied on its own chips — Trainium and Inferentia — as well as GPUs from Nvidia.

The inclusion of Cerebras in the AWS ecosystem means the company has passed rigorous technical and commercial vetting — such partnerships are concluded rarely. The second agreement is with OpenAI. According to sources, the contract value exceeds 10 billion dollars.

Deal details are not disclosed, but the arrangement may involve supplying computational infrastructure for training and inference of next-generation models. For Cerebras, this is not merely a large order but a strategic validation: one of the most computationally demanding companies in the world has chosen their chips. OpenAI traditionally worked with Microsoft Azure infrastructure based on Nvidia hardware — the diversification of suppliers appears as a significant shift.

The IPO comes on the wave of excitement around AI infrastructure. Investors are actively seeking ways to participate in the sector's growth without betting on a single model or service. AI semiconductors represent one of the few segments with nearly guaranteed demand: whatever models win the race, they will need chips.

Cerebras claims the role of a second significant player after Nvidia, although questions about production scalability and software stack maturity remain open. The main signal for the industry: the deals with AWS and OpenAI show that major players are ready to diversify chip suppliers and not limit themselves to Nvidia. In conditions of GPU scarcity and geopolitical risks around Taiwanese manufacturers, this creates a real niche.

The key question is whether Cerebras can scale production and support a software stack at the level to which engineers working with the CUDA ecosystem have become accustomed.