OpenAI has signed a major multi-year deal with Cerebras worth over $10 billion to boost the speed and scale of its artificial intelligence services.
Quick Summary – TLDR:
- OpenAI will purchase up to 750 megawatts of AI compute power from Cerebras through 2028.
- The deal focuses on speeding up real-time inference and scaling AI workloads efficiently.
- Cerebras’ chip systems offer up to 15x faster performance than traditional GPUs.
- The agreement marks a strategic shift away from reliance on Nvidia and fuels IPO ambitions for both companies.
What Happened?
OpenAI and Cerebras have entered into a massive multi-year partnership that will see OpenAI purchase up to 750 megawatts of computing power from the chipmaker. The agreement is valued at over $10 billion and is designed to support the growing demand for high-speed AI inference as OpenAI continues scaling its products like ChatGPT. The partnership will begin rolling out in 2026 and gradually expand over several years.
OpenAI🤝Cerebrashttps://t.co/zvVIdIsw2u pic.twitter.com/cKUL5ZSTE3
— Cerebras (@cerebras) January 14, 2026
OpenAI’s Bet on Speed and Scale
This new collaboration underscores OpenAI’s commitment to building a diverse and resilient AI infrastructure. Rather than relying solely on Nvidia GPUs, OpenAI is integrating Cerebras’ wafer-scale chip systems, which are built for ultra low-latency performance. These chips bundle compute, memory, and bandwidth onto a single massive processor, eliminating many of the bottlenecks seen in conventional systems.
OpenAI’s Sachin Katti stated:
Cerebras’ systems are reported to offer up to 15 times faster inference speeds than GPU-based setups. This improvement is especially valuable for tasks that require quick response times like coding agents, AI voice chat, and image generation.
Multi-Billion Dollar Deal with Strategic Impact
The financial structure of the agreement includes OpenAI purchasing cloud services powered by Cerebras chips. Cerebras will be responsible for building or leasing data centers to support the deployment, which will go live in multiple phases through 2028.
Cerebras CEO Andrew Feldman commented:
He also noted the long-standing relationship between the two companies, dating back to their early technical collaborations in 2017.
The timing of this deal is especially crucial for Cerebras, which is currently preparing for an initial public offering (IPO). This partnership not only diversifies its customer base, reducing dependency on UAE tech firm G42, but also showcases the performance of its technology in a major commercial context.
Meanwhile, OpenAI is also reportedly laying the groundwork for its own IPO, potentially targeting a staggering $1 trillion valuation. CEO Sam Altman previously stated that the company aims to invest $1.4 trillion in building 30 gigawatts of compute infrastructure, a figure that highlights the intensity of the current AI race.
The Bigger Picture in AI Infrastructure
The deal is one of the largest in the AI hardware space and signals a shift in the industry’s power dynamics. Nvidia’s dominance is facing new challenges from rising chipmakers like Cerebras. The focus on real-time inference performance could set a new standard for how AI applications are deployed and experienced.
In a landscape where speed, scale, and cost-efficiency are everything, OpenAI’s latest move shows its determination to lead not just in software and models, but also in hardware strategy.
SQ Magazine Takeaway
I think this is one of the smartest infrastructure bets OpenAI has made. If Cerebras can really deliver faster inference at this scale, it means ChatGPT and other AI tools can become way more responsive, smoother, and powerful in real time. That changes the entire user experience. Also, OpenAI choosing someone other than Nvidia is a big deal. It opens up the playing field and gives other chipmakers a real shot. And let’s be honest, spending $10 billion is no small thing. They are clearly gearing up for something massive.
