NVIDIA showcases AI-driven sustainable energy tools

NVIDIA is showcasing how accelerated computing supports the sustainable energy transition at Climate Week NYC, highlighting LLM inference energy-efficiency gains and publishing product carbon footprint summaries.

• Main announcement/action: NVIDIA is presenting its climate and sustainability work at Climate Week NYC (through Sept. 26 in New York City), emphasizing that LLM inference energy efficiency improved 100,000x over the past 10 years, releasing product carbon footprint comparisons that show a 24% reduction in embodied carbon intensity between HGX H100 and HGX B200, and stating that all offices and operational data centers run on 100% renewable energy (with carbon-free electricity purchases covering 100% of leased data-center footprint).

  • Event: Climate Week NYC
    • Date: through Sept. 26 (year referenced: 2025)
    • Location: New York City (Nest Climate Campus referenced for a fireside chat)
    • Agenda / subject: sessions on AI and energy, AI for climate and weather research, panels including “AI: Powering a More Productive Energy Future” and multiple panels featuring Earth-2 work (Columbia University, AI for Energy, Google, AWS).

• Partnerships, products, and implementation details: NVIDIA is collaborating with startups and internal programs—notably Emerald AI (an NVIDIA NVentures portfolio company and NVIDIA Inception member)—on an NVIDIA Omniverse Blueprint / reference design for energy-efficient, grid-friendly AI factories aimed at maximizing energy-use efficiency; Emerald AI cites the design as capable of unlocking 100 gigawatts of grid capacity. NVIDIA also promotes the NVIDIA Earth-2 platform for high-resolution AI weather and climate modeling and will continue publishing product carbon footprint summaries for new products.