⚡
Brief IA
›

Cerebras Targets 600 MW and a 750 MW Deal with OpenAI

🛠️ AI Tools·Tom Levy·

Cerebras Targets 600 MW and a 750 MW Deal with OpenAI

Cerebras Targets 600 MW and a 750 MW Deal with OpenAI
⚡
Key Takeaways
1Cerebras announces over 600 MW of data center capacity in service or under contract by the end of 2027
2The company raised $5.5 billion in May, launched the CS-4 in August, and signed with OpenAI to deploy 750 MW from 2026 to 2028
3Andrew Feldman is set to detail this strategy at Disrupt, where he will address the constraints of computing, energy, and infrastructure
💡Why it matters — The increase in AI power now depends as much on industrial infrastructure as on advancements in chips.
⚡Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

The race for computing power puts infrastructure in the spotlight. Cerebras announces a significant increase in energy and industrial capabilities, a new wafer-scale system, and a major agreement with OpenAI. Andrew Feldman is set to detail this strategy at Disrupt, against a backdrop of growing demand for computing, energy, and data centers.

Energy Capabilities and Factories: Milestones Announced by Cerebras

Cerebras reports that it has over 600 megawatts of operational or contracted data center capacity, with deliveries expected by the end of 2027. The company also states that it will increase its production capacity by more than ten times in 2026. It announces the launch of its first data center capacity in Europe this year and aims to reach 200 megawatts on the continent by the end of 2027. Beyond processors, the growth of AI requires data centers, energy, cooling, and manufacturing, and Cerebras declares that it is strengthening its computing and production resources as it expands.

A Technical Bet: Wafer-Scale Computing and Its Offering

Cerebras challenges the reliance on conventional chip architectures to scale AI. Its approach involves building a processor directly on the wafer for demanding workloads and offering it as deployable systems on-site as well as through a cloud platform. This direction targets increasingly compute, energy, and infrastructure-intensive models.

Roadmap and Partnerships: IPO, CS-4, and Multi-Year Contract

The company raised $5.5 billion during its IPO in May. In August, it unveiled the CS-4, its new generation of wafer-based AI infrastructure. Cerebras has signed a multi-year agreement with OpenAI for the deployment of 750 megawatts of systems between 2026 and 2028, with a deployment timeline aligned with this period.

Announced Speaking Engagement and Executive Context

Andrew Feldman will speak at TechCrunch Disrupt 2026 in a session titled "Can AI Keep Scaling?". He is expected to address the rising demand for computing, energy, and infrastructure, and discuss scenarios if conventional hardware reaches its limits. A co-founder of Cerebras in 2015, Feldman previously co-founded and led SeaMicro, which was acquired by AMD in 2012, and held positions at Force10 Networks and Riverstone Networks. His session is part of a program featuring over 200 sessions across six stages, including panel discussions and workshops, scheduled from October 13 to 15 at Moscone West in San Francisco. More than 10,000 attendees are expected, along with over 250 speakers and 300 exhibiting startups.

⚡

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.