Cerebras says it will work with everyone in AI hardware except NVIDIA

Cerebras CEO Andrew Feldman said Wednesday at the Bloomberg Tech conference that the AI chipmaker will work with every major hardware maker except NVIDIA, positioning the company as the connective layer for buyers seeking an alternative to a single dominant supplier. Feldman argued that concentration on one vendor is a vulnerability, and that Cerebras aims to be the hedge for cloud providers and enterprise customers assembling AI stacks from multiple vendors. The remarks come as Cerebras, which went public last month in the biggest US tech IPO in years, seeks to replicate its partnership with Amazon by cooperating with a broad set of data-center component suppliers.

The strategy is defined by a single exclusion. Speaking at the Bloomberg Tech conference on Wednesday, Cerebras chief executive Andrew Feldman said the AI chipmaker is working with every major hardware maker in the industry apart from one: NVIDIA. The line was not a complaint. It was the pitch. Feldman’s argument is that the buyers who matter want an alternative to a single dominant supplier, and that Cerebras can be the connective layer for everyone assembling an AI stack out of other parts. If a customer is building with AMD, with cloud providers, with custom silicon, with open models or specialised accelerators, Cerebras wants to be in the conversation. The company is positioning itself as the option for buyers who have decided that depending on one vendor for the most expensive line in the budget is itself a risk.T hat framing is careful about what it does not say. Feldman’s case is not that Nvidia is weak; Nvidia remains the dominant supplier of the GPUs that train most of the industry’s models. The case is that concentration is a vulnerability, and that cloud providers, model labs, and enterprise buyers want a credible second source rather than a wholesale replacement, a wariness sharpened as Nvidia’s $40bn in AI equity bets https://thenextweb.com/news/nvidia-40bn-ai-equity-investments-2026 tightened its grip on the field. Cerebras is selling itself as the hedge, not the substitute. The remarks come as Cerebras moves to work with a wide range of AI data-centre component suppliers, opening the door to more partnerships of the kind it already has with Amazon. The “everyone but Nvidia” posture is, in effect, a go-to-market plan: define the field as Nvidia and not-Nvidia, then offer to be the organising partner for the second camp. The Amazon arrangement is the template Cerebras wants to repeat. By cooperating with a broad set of data-centre component suppliers rather than tying itself to one platform, the company is trying to make itself a default ingredient in stacks assembled from many vendors, the inference engine that slots in wherever a buyer has chosen not to standardise on Nvidia. Each new supplier relationship widens the set of configurations Cerebras can sit inside, which is the practical mechanism behind the rhetoric. Cerebras has the timing to press the point. The company went public last month in the biggest US tech IPO in years https://thenextweb.com/news/cerebras-ipo-5-55-billion-biggest-tech-2026 , giving it both capital and the visibility that makes enterprise buyers comfortable signing on. A freshly listed chipmaker with a clear story about not being Nvidia is a more credible counterweight than a private one making the same claim. Whether the strategy converts into durable revenue is the part Wednesday’s remarks did not settle. Being the alternative to Nvidia is a crowded ambition, shared by AMD and a field of custom-silicon efforts, and willingness to partner is not the same as winning the workloads. What Feldman established in San Francisco is the framing Cerebras intends to sell against. The customers will decide whether not-Nvidia is a product or just a position. Get the TNW newsletter Get the most important tech news in your inbox each week.