Energy as the AI foundation
Jensen Huang’s GTC messaging framed energy as the base of a 'Five Layer Cake' — arguing power efficiency and grid integration are critical if AI scales — and highlighted Earth 2 for solar forecasting/renewables integration (youtube.com). GTC commentators flagged energy firms like GCL Energy Technology moving from power suppliers to operating 'AI factories' that bundle compute and grid services (youtube.com).
Jensen Huang expanded the concept in an official NVIDIA blog titled "AI Is a 5‑Layer Cake," published on March 10, 2026. (blogs.nvidia.com)) NVIDIA’s Earth‑2 claims it can produce forecasts from minute‑level nowcasting up to 15‑day global outlooks, with “Global Data Assimilation” slated to arrive later in 2026. (futura-sciences.com)) The Earth‑2 stack is demonstrated on NVIDIA DGX GB200, HGX B200 and OVX systems and leverages generative models such as CorrDiff to accelerate climate and weather simulations. (engtechnica.com)) NVIDIA’s Earth‑2 partners cited in demonstrations include G42, JBA Risk Management and Spire, highlighting industry pilots that feed observatory and commercial data into the platform. (engtechnica.com)) GCL Energy Technology has moved on partnerships and corporate launches: it agreed with Ant Digital Technologies to form Ant Smart Energy Technology Co. Ltd., announced June 13, 2025 at the SNEC PV conference in Shanghai. (global.chinadaily.com.cn)) GCL’s Smart Computing (Suzhou) Center bills itself as an energy‑AI supercomputing hub with an eventual target around 1,000P of compute and a first phase reported at roughly 500P. (markets.financialcontent.com)) GCL executives — including Deputy Chairman Sun Wei — discussed “energy + computing power” with NVIDIA’s Energy Business GM Marc Spieler at GTC 2025 as part of a push to integrate grid services with large‑scale compute. (gcl-power.com)) NVIDIA and partners have published work on “power‑flexible” AI factories that can shift or curtail load during grid peaks, citing collaborations and white papers involving Emerald AI, EPRI, National Grid and Nebius. (blogs.nvidia.com))
