AMD to supply 6GW of compute capacity to OpenAI in chip deal worth tens of billions

Core Insights - AMD has signed a multi-year chip supply agreement with OpenAI, potentially generating tens of billions in revenue for AMD and enhancing its position in the AI sector [1] - The deal involves AMD supplying 6 gigawatts of compute capacity, sufficient to power approximately 4.5 million homes, starting with the Instinct MI450 GPU [1] - OpenAI will receive the first gigawatt of capacity in the second half of 2026 [1] AMD's Product Performance - AMD claims that the MI450 series will outperform Nvidia's comparable offerings through hardware and software enhancements, many of which will be developed with OpenAI's collaboration [2] - Current MI355X and MI300X series GPUs are already effective for AI inference in large language models due to their substantial memory capacity and bandwidth [2] Stock Agreement Details - As part of the agreement, OpenAI has the option to purchase up to 160 million shares of AMD stock, representing a 10% stake [3] - The first tranche of shares will vest with the initial 1 gigawatt deployment, with additional tranches vesting as OpenAI acquires the total 6 gigawatts [3] Stock Price Milestones - OpenAI's stake is linked to AMD's stock price milestones, with the final tranche vesting when AMD's stock reaches $600 per share [4] - AMD shares experienced a significant increase, closing at $164.67 before opening at $222.24, a nearly 35% rise following the announcement of the deal [4] OpenAI's Strategic Moves - OpenAI is actively pursuing multiple chip partnerships to build its AI infrastructure, including plans for five new Stargate data centers with a total capacity of 7 gigawatts [4] - OpenAI CEO Sam Altman described the partnership with AMD as a significant advancement in building the necessary compute capacity for AI [7] Recent Partnerships - In the previous month, OpenAI secured a $100 billion investment from Nvidia for at least 10 gigawatts of compute capacity and signed a $10 billion deal with Broadcom for custom AI chips [8] - OpenAI is also collaborating with Samsung Electronics and SK Hynix to source DRAM memory chips for its Stargate project and to establish data centers in South Korea [8]