NVIDIA and Oracle to Build DOE’s Largest AI Supercomputer with 100,000 Blackwell GPUs
NVIDIA and Oracle have unveiled a landmark collaboration to build the U.S. Department of Energy’s largest AI supercomputer at Argonne National Laboratory, anchored by the new “Solstice” system with 100,000 Blackwell‑architecture GPUs. A companion “Equinox” system with 10,000 GPUs arrives first, combining for an expected 2,200 exaflops of AI performance to accelerate open‑science breakthroughs.
The Breakthrough
- “Solstice” to deploy 100,000 NVIDIA Blackwell GPUs, the largest DOE AI system announced to date.
- “Equinox” to feature 10,000 Blackwell GPUs, targeted for availability in the first half of 2026.
- Combined AI performance of roughly 2,200 exaflops across the two systems.
- Systems to be installed at Argonne National Laboratory for public researchers.
Technical Details
- Built on NVIDIA’s Blackwell GPU architecture with NVIDIA high‑performance networking.
- Software stack includes Megatron‑Core for training frontier models and TensorRT for high‑throughput inference.
- Designed to power emerging agentic AI workflows in open science.
- Oracle’s role provides hyperscale infrastructure expertise for DOE deployments.
Impact and Applications
- Expected to accelerate research in climate modeling, materials, biology, and energy systems, while expanding access for U.S. scientists.
- Part of a broader U.S. effort to ramp AI compute capacity; NVIDIA has also outlined plans for multiple government AI supercomputers, signaling sustained investment in sovereign AI infrastructure.
Future Outlook
- Timeline: Equinox in 1H 2026; Solstice schedule to follow DOE deployment planning.
- Additional systems for federal research programs are anticipated, with configurations exceeding 100,000 GPUs across sites as projects progress.
- Watch for software optimizations (Megatron‑Core, TensorRT) to lift utilization and shorten training cycles on trillion‑parameter models.
In a year defined by escalating AI infrastructure bets, the Solstice‑Equinox build stands out for sheer scale and public‑science focus—positioning Argonne, DOE, and U.S. researchers to push the frontier of AI‑driven discovery.