Reports do not confirm Cloudflare as a customer of new Cerebras hardware

In short: Cerebras is linked to major AI companies through partnerships, but public reporting does not back up a claim that Meta, OpenAI, Cerebras, and Cloudflare are all early customers of entirely new Cerebras hardware.

What happened

Cerebras Systems builds very large AI chips that are used in data centers. Think of a data center as a warehouse full of computers that companies rent to run apps.

Recent reports say Cerebras has relationships with OpenAI and Meta. OpenAI signed a multi-year agreement in early 2026 for about 750 megawatts of Cerebras computing capacity, reported to be worth more than $10 billion, with deployment planned from 2026 to 2028 in phases.

This deal is aimed at AI “inference,” which is the part where an AI model answers your question, like when ChatGPT writes a reply. Cerebras says its design can reduce waiting time, because it packs a lot of computing and memory onto one chip (like putting a whole kitchen on one countertop instead of spreading tools across rooms).

Meta also uses Cerebras systems for inference for its Llama models. In Meta’s Llama API, developers can choose Cerebras for fast text generation, alongside other providers such as Nvidia and Groq.

However, the specific claim that Meta, OpenAI, Cerebras, and Cloudflare are among the first customers of Cerebras’ new hardware is not clearly supported by public information. In particular, there are no confirmed reports that Cloudflare is a Cerebras customer for this new hardware.

Why it matters

If these deals scale up, they could make AI tools feel more responsive for everyday users. At the same time, the gap between what is claimed in marketing and what is confirmed in public reporting can make it hard to know who is actually using which hardware.

Source: Wired