OpenAI is developing internal software to manage AI workloads across various chip providers, as reported by The Information. Compute chief Sachin Katti noted that this initiative aims for a heterogeneous AI infrastructure and may lead to the public release of the tool. Currently, OpenAI is utilizing or planning to use chips from multiple companies, including Amazon, Cerebras, AMD, and its own custom silicon, alongside Nvidia’s hardware, which has a significant presence in the AI ecosystem. This move could challenge Nvidia’s competitive advantage, largely driven by its CUDA software ecosystem.

AMD: AMD produces GPUs and AI accelerators used in data centers and high-performance computing. The company’s chips feature in OpenAI’s expanding roster of hardware providers for training and inference.
Amazon: Amazon supplies cloud computing resources and custom AI silicon through its AWS services. OpenAI is incorporating or planning to use Amazon’s chips as part of its diversified hardware approach.
Nvidia: Nvidia develops graphics processing units and related software platforms that power much of the current AI infrastructure. Its CUDA ecosystem has served as a primary competitive advantage in AI hardware. The news underscores how OpenAI’s multi-vendor strategy could diminish reliance on this ecosystem.
OpenAI: OpenAI is an artificial intelligence research and deployment organization focused on developing advanced AI models and infrastructure. The company is building internal software to orchestrate AI workloads across chips from various hardware providers rather than depending solely on one supplier. This effort positions OpenAI to support heterogeneous computing environments while preparing for next-generation training clusters.
Cerebras: Cerebras designs specialized AI accelerators optimized for large-scale machine learning workloads. OpenAI has integrated or intends to integrate Cerebras chips into its infrastructure mix.
Vera Rubin: Vera Rubin is Nvidia’s upcoming GPU architecture designed for advanced AI training clusters. OpenAI is preparing deployments based on this architecture for later in the year as part of its multi-chip strategy.
Sachin Katti: Sachin Katti serves as compute chief at OpenAI, overseeing the company’s AI infrastructure strategy. He has outlined plans for heterogeneous AI hardware setups and suggested the internal orchestration tool might eventually be released publicly. His comments highlight OpenAI’s intent to operate across diverse chip architectures.

{“Hardware Roadmap”: “OpenAI is preparing clusters leveraging Nvidia’s Vera Rubin architecture for AI training later this year.”, “Software Tool Development”: “OpenAI is creating internal software to manage AI workloads across heterogeneous chip providers, with potential plans to make the tool publicly available.”, “Infrastructure Diversification”: “OpenAI is incorporating or planning chips from Amazon, Cerebras, AMD, and its own custom silicon in addition to Nvidia hardware.”}