NVIDIA, OpenAI, Meta Fuel AI Infrastructure Race
NVIDIA and OpenAI announced a partnership to deploy 10 gigawatts of AI systems, aiming to power superintelligence development. Concurrently, Meta plans to spend billions on AMD chips and has secured a multi-billion dollar deal with NVIDIA to expand its AI compute capacity. NVIDIA also unveiled its 'Rubin' platform and plans to build a Windows PC System-on-Chip, competing directly with Qualcomm and Intel.
- The NVIDIA 'Rubin' platform, named after astronomer Vera Rubin, is a comprehensive system featuring six different chip types, including the Rubin GPU and the 'Vera' CPU. This integrated approach aims to reduce AI inference token costs by up to 10 times compared to the prior 'Blackwell' platform. - NVIDIA's 'Vera' CPU component of the Rubin platform will feature 88 custom-designed cores based on Arm architecture. - Meta's deal with AMD is valued at over $100 billion and includes an option for Meta to acquire up to a 10% equity stake in AMD, signaling a deep, long-term strategic partnership beyond a simple supplier relationship. This move is part of Meta's strategy to diversify its chip suppliers beyond NVIDIA. - The 10 gigawatt OpenAI-NVIDIA initiative is a massive undertaking, as a single gigawatt is roughly the output of a nuclear reactor. The deployment will involve tens of thousands of server racks and millions of GPUs. - Meta's total capital expenditure for 2026 is projected to be between $115 billion and $135 billion, a significant increase driven by AI infrastructure investments. - NVIDIA's planned Windows PC System-on-Chip (SoC) is a collaboration with MediaTek and is expected to feature Arm-based CPU cores alongside NVIDIA's GPU technology. This positions NVIDIA to compete directly with Qualcomm and Intel in the "AI PC" market. - The Rubin platform is slated for release in the second half of 2026, with an enhanced "Rubin Ultra" version planned for 2027. The subsequent generation of NVIDIA's data center architecture will be named after physicist Richard Feynman. - NVIDIA's data center division, which includes its AI chips, reported revenue of $51.22 billion in the third quarter of fiscal 2026, a 66% increase year-over-year, showcasing the immense demand fueling this infrastructure race.
