Amazon's Nvidia GPU spending nearly doubles as AI chip demand reshapes cloud infrastructure

Aug 271 views

Key takeaways

Amazon ordering 2 million Nvidia GPUs for AWS data centers in 2027–2028, valued in tens of billions
Deal includes Blackwell Ultra and Rubin chips plus integrated networking, software, and robotics platforms
Expansion happens despite Amazon's competing Trainium and Graviton custom chip programs

Amazon and Nvidia deepened their partnership Wednesday, announcing a 2 million GPU order for AWS data centers in 2027 and 2028—nearly tripling Amazon's commitment from five months prior. The new order includes Blackwell Ultra, Rubin, and Rubin Ultra chips, worth tens of billions of dollars by industry estimates.

The expansion extends beyond raw compute capacity. Nvidia will integrate its networking hardware, open-source models, CPUs, data processing software, and robotics platforms across AWS. Nvidia cited "surging demand" from startups, enterprises, AI labs, and governments as the driver. The deal arrives during Nvidia's record earnings quarter, with data center revenue hitting $89 billion, up 117 percent year-over-year.

Amazon continues investing in custom chips through its Trainium and Graviton lines, but the GPU expansion reflects market realities: Nvidia's ecosystem remains the dominant platform for large-scale AI workloads. AWS will also deploy Nvidia's Vera CPUs and robotics stack, including Omniverse simulation and Isaac development tools.

The bigger picture

This partnership crystallizes a paradox in enterprise AI infrastructure. Amazon competes with Nvidia through custom chips yet remains its largest customer, unable to fully transition AWS workloads away from Nvidia's architecture. Google and Meta face similar dynamics—all three hyperscalers are building alternatives while deepening Nvidia dependency. The deal signals that custom chip strategies alone cannot yet displace Nvidia's ecosystem, though the game remains in early innings.

LagPing's take

We're covering this because it reveals how even tech giants with massive engineering resources remain tethered to Nvidia's dominance. The speed of this escalation—doubling orders in five months—matters more than the headline figure. It shows the gap between building competitive chips and actually replacing entrenched infrastructure is still measured in years, not quarters.

X / Twitter WhatsApp

View at TechCrunch AI

Find "NVIDIA" on Amazon

As an Amazon Associate, LagPing earns from qualifying purchases. Product links are affiliate links.