Breaking NVIDIA(NVDA.US)'s "exclusive supply"! OpenAI's first integration of Cerebras chips, competing in the programming assistant race with a lightweight version of Codex

According to the Tonghuashun Finance APP, OpenAI officially released its first AI model based on chips from semiconductor startup Cerebras Systems Inc., marking a move by ChatGPT developer to actively expand its chip supplier network to break Nvidia’s (NVDA.US) monopoly.

The model is named GPT-5.3-Codex-Spark, the latest member of the Codex family of programming automation tools, positioned as a lightweight version with lower computational consumption but faster response speed. This version is scheduled to launch on Thursday, supporting software engineers in quickly completing tasks such as code snippet editing and testing. Users can interrupt processes or issue new commands at any time without waiting for lengthy computation cycles.

Last month, OpenAI and Cerebras finalized a hardware procurement agreement worth over $10 billion to achieve faster model inference responses. For Cerebras, this deal is a key move to carve out a niche in the market long dominated by Nvidia. For OpenAI, it is part of its ongoing effort to expand its supplier base to meet exponential growth in computational demand.

In October last year, OpenAI reached a major partnership with AMD (AMD.US), another competitor of Nvidia, planning to deploy a total capacity of 6 gigawatts of AMD graphics processors over the coming years. That same month, OpenAI also finalized a deal to customize chips and network components from Broadcom (AVGO.US). Recently, OpenAI’s relationship with Nvidia has attracted attention due to reports of tension between the two companies, though their CEOs subsequently publicly stated they will continue close collaboration.

An OpenAI spokesperson emphasized in a statement on Thursday that the partnership with Nvidia is “foundational,” and its most powerful AI models are the result of “years of hardware and software engineering collaboration.” “For this reason, we consider Nvidia as the core architecture for training and inference, while consciously expanding our ecosystem beyond the core through collaborations with Cerebras, AMD, and Broadcom.”

The new Codex model is also part of OpenAI’s latest effort to compete for dominance in the high-growth AI coding assistant sector against rivals like Google and Anthropic. OpenAI revealed that the current weekly active users of Codex have surpassed 1 million.

The company stated that the new version will initially be available for research preview to ChatGPT Pro subscribers, with broader user access gradually expanding over the coming weeks.