OpenAI has partnered with AMD to deploy up to six gigawatts of AI compute power, starting in 2026 with AMD’s next-generation MI450 chips, to meet the rapidly growing demand for AI inference workloads and accelerate AI innovation. This collaboration highlights the critical role of diverse, large-scale compute infrastructure and robust energy solutions in advancing AI technology, positioning both companies for significant growth in the expanding AI accelerator market.
OpenAI has entered into a significant partnership with AMD to support a massive buildout of AI compute capacity, marking a major milestone for both companies and the AI industry. AMD will provide up to six gigawatts of AI compute power, starting with an initial deployment of one gigawatt in the second half of 2026 using their next-generation MI450 chips. This collaboration reflects the urgent need for increased computational power to support the rapidly growing demand for AI applications, such as ChatGPT, which now boasts 800 million weekly active users. Both companies emphasize that compute is foundational to AI progress and economic transformation.
The focus of this compute expansion is primarily on AI inference workloads, which are critical for deploying AI models at scale. OpenAI acknowledges that the demand for AI compute is vastly underestimated and that current limitations in computational power are hindering the launch of new features and products. The partnership aims to accelerate the availability of compute resources to meet this demand, with plans to deploy the capacity across multiple locations and cloud providers, including Oracle. This distributed approach is necessary given the sheer scale of compute required.
Financially, OpenAI is exploring various funding options to support this ambitious buildout, including equity, debt, and creative financing strategies. AMD expressed confidence in OpenAI’s ability to meet its commitments, highlighting the mutually beneficial nature of the deal for both companies and their shareholders. The partnership also underscores the importance of a robust supply chain and energy infrastructure, with mentions of nuclear power as a potential energy source to support the massive power requirements of AI data centers.
From a technology perspective, AMD’s MI450 series chips are seen as highly competitive, especially for inference tasks, complementing existing NVIDIA-based training infrastructure. OpenAI values having a diversity of chip options to optimize for different workloads and use cases. The collaboration is described as an “all-in” partnership, with both companies working closely on hardware, software, and supply chain logistics to ensure the successful deployment of this large-scale compute capacity.
Looking ahead, AMD views the AI accelerator market as enormous, with a total addressable market potentially exceeding $500 billion in the coming years. The partnership with OpenAI serves as a strong validation of AMD’s technology and positions the company well within the competitive landscape dominated by NVIDIA. Both AMD and OpenAI see this deal as a catalyst for faster industry-wide growth in AI compute infrastructure, which is essential for powering the next generation of AI-driven economic and technological advancements.