AWS and Nvidia announced 2 million additional GPUs — Blackwell Ultra, Rubin and Rubin Ultra — for deployment across 2027–2028, taking the committed total from more than 1 million to more than 3 million. The earlier 1M commitment, made at GTC 2026 for deployment starting this year, was exhausted ahead of schedule. The deal also covers 100,000 GPUs of secure AI-factory capacity for the US government.
NVIDIA's AVO agent scored 100.00 RHAE on ARC-AGI-3's 25-environment public set using Claude Opus 5, which scores 30.2% on its own. The held-out semi-private and private sets were not run. Headlines called it a perfect benchmark score for the model; it is a measurement of the scaffold, and the tracker does not record it.
Our read — labelled opinion, not investment advice.
Nvidia will backstop up to $105B of OpenAI's 20-year lease on a Pike County, Ohio data center being built by SoftBank-owned SB Energy — covering part of the lease and power payments plus a floor on the site's residual value. Capacity reaches 8 GW, with the first 800 MW due in 2028. Nvidia is the exclusive chip supplier and is also investing $1.5B in SB Energy.
SpaceX and Nvidia said they are jointly developing the compute payload for Starmind AI1, a satellite carrying Nvidia Rubin GPUs and Vera CPUs in a Vera Rubin NVL72 rack-scale system. SpaceX committed to Nvidia exclusively for its next generation of AI infrastructure, with production at its Bastrop, Texas Gigasat factory and first deployments targeted as early as 2027.
Nvidia will invest $5B in Safe Superintelligence, Ilya Sutskever's secretive startup, giving it access to the Vera Rubin platform and shifting its compute from Google TPUs to Nvidia GPUs — increasing SSI's compute by an order of magnitude. The round lifts SSI's total to ~$7B at a ~$32B valuation; the investment is milestone-contingent.
Nvidia and SK Group announced a $500B+ initiative spanning AI data centers and next-gen memory. It pairs a long-term SK hynix HBM4 supply-and-co-development partnership with an SK Telecom 2GW data center (Vera Rubin chips, HBM4) online from 2027, plus Naver/Brookfield and Hyundai Genesis autonomy tie-ups. Samsung is also in on chip/memory design.
Amazon (AWS) 
Anthropic 
OpenAI 
SpaceX 
SK hynix 
Samsung Foundry