At New York Climate Week, NVIDIA is showcasing five companies using its AI platforms to accelerate clean energy deployment. The projects span grid digital twins, nuclear operations, recycled battery storage, advanced fission, and fusion energy. Each is aiming to compress development timelines from years to months.
The common thread is NVIDIA's compute stack: CUDA for grid simulation, Blackwell for battery intelligence, and Omniverse with OpenUSD for reactor and fusion digital twins. All five companies are members of NVIDIA's Inception startup program, and the pitch is consistent: replace slow physical iteration with simulation, then let AI agents run the operations.
Confirmed
According to NVIDIA's blog post, the five companies are:
- ThinkLabs AI uses CUDA-powered digital twins and agents to cut grid interconnection study times from 30–45 days to two minutes for Southern California Edison.
- Atomic Canyon deploys its Neutron AI workbench and NIVA virtual assistant on NVIDIA accelerated compute to streamline nuclear plant operations at PG&E's Diablo Canyon and across the national fleet.
- Redwood Materials integrates repurposed EV batteries with an AI intelligence layer running on the Blackwell platform, delivering off-grid power for AI factories in months instead of years.
- TerraPower connects an Omniverse-powered digital twin to accelerate siting and delivery of its sodium-cooled Natrium reactors from years to months.
- Commonwealth Fusion Systems (CFS) uses Omniverse libraries and OpenUSD to compress SPARC tokamak experimentation from years to weeks, targeting grid connection in the 2030s at a Chesterfield County, Virginia site.
Unknown
- Specific NVIDIA hardware configurations (GPU counts, cluster sizes) for each project.
- Whether the Blackwell platform deployment at Redwood Materials is in production or pilot stage.
- Independent verification of the two-minute interconnection claim beyond the Southern California Edison case study.
- Regulatory approval status and timelines for TerraPower's Natrium and CFS's ARC plants.
- Total capital deployed or revenue tied to these NVIDIA-enabled projects.
Our take
NVIDIA is positioning its full stack — CUDA, Blackwell, Omniverse — as the compute substrate for the energy transition. The pattern is consistent: each startup replaces slow, physics-limited iteration with simulation and AI-assisted operations. That framing is compelling, but the hard part is what happens after the pilot. Whether these results scale past single-site deployments, survive regulatory review, and hold up without NVIDIA's direct involvement will determine if this is a real infrastructure shift or a well-packaged demo reel.
