AMD Secures 6 GW OpenAI AI Compute Deal, Fuels $34 B Data Center Growth

AMDAMD

Advanced Micro Devices secured a transformative multi-year AI partnership with OpenAI to supply 6 GW of AI compute capacity, underpinning its full-stack, rack-scale AI systems strategy. This collaboration supports AMD’s data center segment growth, which has scaled revenue from $2 bn to over $16 bn and targets $34 bn total.

1. Riot Platforms Secures 10-Year Data Center Lease with AMD at Rockdale Site

Riot Platforms has entered into a landmark Data Center Lease and Services Agreement with Advanced Micro Devices, under which AMD will occupy an initial 25 megawatts of critical IT load capacity at Riot’s Rockdale, Texas, facility. The ten-year lease carries approximately $311 million in contract revenue, with three five-year extension options that could lift total revenue to around $1 billion. Riot funded its $96 million fee-simple acquisition of the 200-acre site by selling 1,080 bitcoins, and plans to retrofit existing infrastructure to deliver the first phase in January 2026 and complete it by May 2026. Should AMD exercise its expansion and right-of-first-refusal options, its commitment could grow to 200 megawatts, positioning Riot’s 1.7-gigawatt approved power portfolio as a strategic AI/HPC hub.

2. OpenAI Partnership Anchors AMD’s AI Roadmap

Advanced Micro Devices has secured a multi-year, 6-gigawatt AI compute partnership with OpenAI that extends its data center narrative from standalone accelerators to full-stack, rack-scale systems. Under this agreement, AMD will deliver not only its latest MI300-series processors but also networking, power distribution and software orchestration, addressing hyperscalers’ bottlenecks in scaling large language models. This deal underpins AMD’s projection to grow its data center segment from approximately $2 billion in annual revenue three years ago to a run rate exceeding $16 billion today, with analysts forecasting total data center revenue could reach $34 billion by the end of 2024.

3. High-Bandwidth Memory Strategy Poised to Capture Market Share

Industry analysts highlight AMD’s emphasis on high-bandwidth memory (HBM) in its latest MI325X accelerators as a differentiator for AI inference workloads, where memory bandwidth can be more critical than raw compute horsepower. Following recent quarterly results showing data center revenue growth accelerating beyond 30 percent year-over-year, AMD’s HBM-centric architecture is expected to resonate with hyperscale customers prioritizing memory-intensive tasks. The firm’s server segment has reported sequential margin expansion and improved utilization rates in its new 5-nanometer products, reinforcing forecasts that the data center division could outpace overall company growth in 2026.

Sources

SFFSB
+6 more