Cerebras Systems, a $95 billion AI chip company, is strategically positioning itself as a partner for businesses seeking to build artificial intelligence infrastructure outside Nvidia's dominant ecosystem. Speaking at the Bloomberg Tech conference in San Francisco, CEO Andrew Feldman revealed that Cerebras is working with every major hardware manufacturer in the industry except Nvidia, as it expands its partnerships across the AI infrastructure landscape.
Cerebras' Strategy of Broad Collaboration
Feldman emphasized that Cerebras is engaged in using other companies' components for part of the problem and its own for another part, collaborating with all members of the community—except Nvidia. “Everybody but them,” he stated. This approach underscores Cerebras' aim to develop across various platforms, whether hardware or cloud, without being confined to a single ecosystem. The company positions itself as an IT solutions provider that can integrate into any infrastructure used by firms implementing AI.
Partnerships with Amazon and OpenAI
Cerebras is building on recent agreements with Amazon and OpenAI to broaden its role in supplying hardware for AI workloads. Under its partnership with Amazon, Cerebras products will be used alongside Amazon Web Services' custom-designed chips in AWS data centers. The deal with OpenAI has further elevated Cerebras' profile among AI infrastructure providers. Feldman described the period leading up to the company's public listing as a “pretty good 90-day run,” referring to these announcements.
Public Listing and Financial Success
The California-based company went public this year, raising $5.5 billion in the largest US initial public offering of 2026 so far. Shares climbed significantly during their first day of trading, reflecting strong investor confidence in its unique technology.
Wafer-Scale Processor Technology
Cerebras develops wafer-scale processors, a chip architecture distinct from conventional AI processors used in most server systems. The company claims its technology processes and stores large amounts of data on a single processor, reducing reliance on external components. Its products are primarily aimed at AI inference tasks, where trained models generate responses, predictions, and outputs from new data. Because the processors do not fit traditional server designs, Cerebras has developed its own computing systems and operates AI computing services through its data center network.
