TL;DR
OpenAI has signed a multi-year agreement with Cerebras for 750 megawatts of compute capacity to be delivered through 2028, a deal reported to be worth over $10 billion. The partners say the systems will speed up responses for tasks that currently take longer to process, adding a low-latency inference option to OpenAI's compute mix.
What happened
OpenAI and AI-chip maker Cerebras announced a multi-year partnership in which Cerebras will supply 750 megawatts of compute beginning this year and continuing through 2028. TechCrunch and Reuters report the agreement is worth in excess of $10 billion. According to statements from both companies, the primary purpose of the deployment is to accelerate responses for workloads that currently require more time to process, with OpenAI describing the systems as a way to speed outputs for customers. Cerebras, which has marketed its chips and systems as faster than traditional GPU-based infrastructure, framed the deal as adding a dedicated, low-latency inference capability to OpenAI’s platform. The coverage also notes background on Cerebras: it has operated for more than a decade, filed for an IPO in 2024 but delayed it multiple times, has continued fundraising, and was reported to be in talks to raise another $1 billion at a roughly $22 billion valuation. OpenAI’s CEO Sam Altman is identified as an investor in Cerebras, and OpenAI at one point considered acquiring the company.
Why it matters
- Adds a dedicated low-latency inference option to OpenAI’s compute portfolio, aimed at improving response speeds for longer or more complex queries.
- Represents a large, multi-year commercial commitment that could shape how major AI workloads are provisioned outside traditional GPU deployments.
- Signals continued investor and partner interest in alternative AI chip architectures alongside established GPU suppliers.
- Could affect how OpenAI scales real-time interactions if the promised latency and throughput advantages materialize.
Key facts
- Agreement covers 750 megawatts of compute to be delivered from 2026 through 2028.
- The deal is reported by TechCrunch and Reuters to be worth more than $10 billion.
- OpenAI says the systems will speed responses that currently require more time to process.
- Cerebras asserts its AI-designed chips and systems can be faster than GPU-based setups like those from Nvidia.
- Cerebras has been operating for over a decade and saw heightened attention after the 2022 launch of ChatGPT.
- The company filed for an IPO in 2024 but postponed that process multiple times.
- Reports indicate Cerebras was in talks to raise about $1 billion at an approximate $22 billion valuation.
- Sam Altman, OpenAI’s CEO, is an investor in Cerebras; OpenAI once considered acquiring the company.
What to watch next
- Delivery and deployment timeline: Cerebras is scheduled to start supplying compute this year and continue through 2028; monitoring real-world rollout and performance will be key.
- Whether OpenAI will use Cerebras systems for specific models or workloads and how those models perform compared with GPU-based deployments — not confirmed in the source.
- Potential commercial or supply-chain impacts on GPU vendors and other infrastructure partners as OpenAI diversifies its compute stack — not confirmed in the source.
Quick glossary
- Inference: The process of running a trained machine learning model to generate predictions or outputs from new input data.
- Compute (megawatts): A measure of the electrical power capacity allocated to run data center hardware; in this context, megawatts indicate the energy capacity supporting AI systems.
- AI chip: A processor designed specifically to accelerate machine learning tasks, often optimized for neural network operations.
- GPU: Graphics Processing Unit, a type of processor widely used in AI training and inference due to its parallel computation capabilities.
Reader FAQ
How large is the deal between OpenAI and Cerebras?
TechCrunch and Reuters report the multi-year agreement is worth over $10 billion.
What will Cerebras deliver to OpenAI?
Cerebras will provide 750 megawatts of compute capacity beginning this year and continuing through 2028.
Will this replace Nvidia GPUs at OpenAI?
Not confirmed in the source.
Is Sam Altman connected to Cerebras?
Yes. The coverage identifies OpenAI CEO Sam Altman as an investor in Cerebras.
IN BRIEF Posted: 2:25 PM PST · January 14, 2026 IMAGE CREDITS: JUSTIN SULLIVAN / GETTY IMAGES Lucas Ropek OpenAI signs deal, worth $10 billion, for compute from Cerebras OpenAI…
Sources
- OpenAI signs deal, worth $10 billion, for compute from Cerebras
- Market Today: Verizon outage, Tesla FSD shift, banks in …
- SalvatoreRa/ML-news-of-the-week
- Data Center Industry news & Info
Related posts
- How Online Conspiracies and Influencers Shape Policy in Trump’s 2nd Term
- Google Pixel’s comeback: strong 2025 growth and renewed momentum
- Anthropic says using Claude Code to build a rival must go through the API