In October 2025, Advanced Micro Devices secured a landmark multi-year commitment from OpenAI to deploy 6 gigawatts of AMD Instinct AI accelerators across the ChatGPT developer’s global data centers. The agreement calls for roughly 6 million high-performance accelerator cards over five years, representing an estimated $120 billion in hardware purchases and stock-warrant incentives. Shipments will begin in the second half of 2026 and ramp meaningfully in 2027 with next-generation Instinct MI450 series chips, providing AMD with a highly visible revenue backlog and cementing its role as a strategic supplier to the fastest-growing AI applications.

Beyond individual GPU accelerators, AMD is promoting integrated rack-scale AI solutions that address hyperscaler bottlenecks in power delivery, networking fabric and software management. The company’s new ‘Project Helios’ modular rack design combines EPYC CPUs, Instinct GPUs, in-house Infinity Fabric networking and optimized software libraries into a turnkey AI cluster. Early reference configurations delivered to Microsoft and Oracle in late 2025 achieved 25% higher compute efficiency per kilowatt and 40% lower interconnect latency versus incumbent alternatives, positioning AMD to capture a growing share of large-scale AI deployments.

AMD’s data center revenue has surged from $2 billion in fiscal 2020 to over $16 billion in 2023, driven by EPYC server CPU adoption and the initial rollout of MI300-series accelerators. Management projects data center sales will exceed $34 billion in 2024, underpinned by the OpenAI deal, Helios systems and a pipeline of AI-optimized CPUs and GPUs through 2028. With gross margins expanding by 500 basis points over the past two years and a consistent target of 60%+ gross margin on next-generation AI products, AMD is positioned for a multi-decade growth trajectory as cloud providers and enterprise customers accelerate AI infrastructure spending.