NVIDIA and AWS Are Expanding Their GPU Deal Deep Into 2027 and 2028
As seen on the 24/7 Wall St. homepage on August 26, 2026.
AWS and NVIDIA to Deliver 2 Million Additional GPUs and Next-Generation Infrastructure for Agentic and Physical AI
AWS demand has already blown past the 1 million GPUs pledged at GTC 2026, and the new Blackwell Ultra, Rubin and Rubin Ultra commitment covers 2027-2028, extending visibility well past the current cycle. The deal also puts Vera CPUs, NVLink Fusion with custom memory inside Trainium racks, and 100,000 GPUs for US government AI factories in play.
Continue ReadingShow less
AWS had already consumed the GPUs NVIDIA pledged at GTC 2026, and the two companies have now committed to 2 million more, covering Blackwell Ultra, Rubin, and Rubin Ultra chips and extending AWS's confirmed supply well into 2027 and 2028. That is unusually deep forward visibility for an industry where allocation fights are fierce.
The deal goes beyond raw GPU counts. It brings NVIDIA Vera CPUs and NVLink Fusion with custom memory into Amazon's Trainium racks, tying NVIDIA's interconnect architecture directly into AWS's own custom silicon stack.
Sponsored
Are You Ready To Retire, Or Years Behind?
Most Americans suspect they're behind on retirement and never find out. Advisor.com's free matching tool pairs you in about three minutes with a vetted fiduciary advisor who can help you with investing, taxes, retirement, estate planning, and more. No minimums. No sales call. Find out where you stand.
A dedicated slice of 100,000 GPUs is earmarked for US government AI factories, which adds a public-sector dimension to what has largely been a commercial cloud buildout. That commitment signals both companies are positioning this infrastructure as strategically important.
For investors tracking AI infrastructure spend, the pace matters: AWS burned through its first tranche of GPUs before the next was even formally announced. The 2027 to 2028 window of the new commitment gives both NVIDIA and AWS rare multi-year revenue and supply-chain clarity at a moment when most hyperscaler GPU agreements remain short-cycle.