The multi-year agreement, reportedly valued at over $10 billion, will see OpenAI deploy up to 750 megawatts of Cerebras wafer-scale compute infrastructure—one of the largest AI hardware expansions in history.
The idea is to leverage "unique speed" that comes from combining "massive compute, memory, and bandwidth" on a "single giant chip" to eliminate the "bottlenecks that slow inference on conventional hardware."
TL;DR
- OpenAI has signed a multi-billion-dollar deal with Cerebras Systems to secure high-performance AI compute.
- The partnership will provide 750 MW of wafer-scale systems for large-scale inference workloads.
- The move diversifies OpenAI’s hardware beyond GPUs, reducing reliance on NVIDIA.
- It follows another deal with SoftBank’s SB Energy to power data centers using 2.3 GW of solar capacity.
A New Era Of Compute Power
OpenAI’s deal with Cerebras Systems marks a major step toward reshaping the AI hardware ecosystem.
The partnership centers on the company’s Wafer Scale Engine (WSE).
As per the company, its WSE-3 is the world’s largest chip, featuring 900,000 AI cores, 4 trillion transistors, and 44GB of on-chip memory on a single wafer, which delivers 125 petaflops of compute power.
This architecture is designed for high-throughput workloads such as inference, fine-tuning, and model serving, areas where Cerebras claims to deliver up to 15 times faster processing than traditional GPU clusters.
The collaboration will allow OpenAI to offload inference from its primary GPU clusters, freeing those systems for large-scale model training while improving latency and cost-efficiency for services such as ChatGPT and API integrations.
“OpenAI’s compute strategy is to build a resilient portfolio that matches the right systems to the right workloads. Cerebras adds a dedicated low-latency inference solution to our platform. That means faster responses, more natural interactions, and a stronger foundation to scale real-time AI to many more people,” said Sachin Katti of OpenAI.
Deployment will roll out across multiple stages between 2026 and 2028.
Industry observers view the deal as both a technical and strategic maneuver. With global GPU shortages and soaring chip costs, diversifying into alternative architectures gives OpenAI resilience and cost control. It also aligns with OpenAI’s long-term vision to make “frontier AI models more accessible, efficient, and sustainable.”
Market Impact
The announcement immediately rattled the semiconductor market—NVIDIA and AMD shares dipped as investors priced in future competition from Cerebras. Analysts note that wafer-scale compute could evolve into a credible alternative to GPU-based clusters, especially for inference workloads that prioritize throughput over flexibility.
“Just as broadband transformed the internet, real-time inference will transform AI, enabling entirely new ways to build and interact with AI models,” said Andrew Feldman, co-founder and CEO of Cerebras.
OpenAI’s decision signals growing momentum behind specialized AI chips and vertically integrated compute infrastructure. By locking in capacity early, the company ensures it can sustain the rapid user growth of ChatGPT and future multimodal systems while minimizing operational bottlenecks.
On The Heels Of Another Massive Partnership
This deal follows another major collaboration announced just weeks earlier, where OpenAI and SoftBank Group partnered with SB Energy (a SoftBank Group company) to supply 2.3 gigawatts of renewable solar energy for powering OpenAI’s next-generation data centers.
Under the agreement, SB Energy will build large-scale solar facilities in Texas dedicated to supporting OpenAI’s Stargate data-center network.
Both, OpenAI and SoftBank Group, will invest $500 million into SB Energy.
The three have also formed a non-exclusive preferred partnership to develop a new model for data center builds.
Topics For More Insights
The Bigger Picture
Together, these partnerships demonstrate OpenAI’s twin strategy—scaling compute capacity with Cerebras’ wafer-scale chips while ensuring sustainable energy supply through SB Energy.
Once complete, the OpenAI–Cerebras network could rank among the world’s largest dedicated AI compute clusters. It represents a shift from merely renting cloud compute toward co-developing hardware infrastructure tailored to AI workloads.
The dual approach underscores OpenAI’s intent to build a robust, self-reliant AI ecosystem, from silicon to sustainability.
Join The Discussion