Hyperscalers Expand AI Chip Dual Sourcing as Nvidia Data Center Hits $75 Billion, AMD Grows 107%
Updated
Updated · Yahoo Finance · Aug 25
Hyperscalers Expand AI Chip Dual Sourcing as Nvidia Data Center Hits $75 Billion, AMD Grows 107%
3 articles · Updated · Yahoo Finance · Aug 25
Summary
$75.25 billion in Nvidia data center revenue and $6.72 billion at AMD are accelerating hyperscaler moves to buy AI chips from both vendors rather than rely on a single supplier.
Nvidia still widened its lead with Q1 FY2027 revenue of $81.61 billion, data center growth of 92%, and networking up 199% as Blackwell and Rubin systems pulled in rack-scale deployments.
AMD's challenge rests on price-performance: Q2 2026 revenue reached $11.54 billion, Instinct sales more than doubled on the MI350 ramp, and data center now makes up 58% of total revenue.
New product roadmaps sharpen that split, with Nvidia's Vera Rubin shipping in Q3 and promising 35x higher inference throughput, while AMD says Helios can deliver 30% more tokens per dollar.
Customer commitments show why dual sourcing is gaining traction: AWS plans to add more than 1 million Blackwell and Rubin GPUs, while Anthropic backed AMD's MI450-based Helios with up to 2 gigawatts and Microsoft plans Azure deployment.
Why is AMD betting billions on an AI lab to break Nvidia's grip on the gigawatt-scale data center market?
Will a new open networking standard finally shatter Nvidia's proprietary AI factory model before the next generation of chips arrives?
Could the sudden rise of agentic AI make traditional server CPUs the unexpected bottleneck in a trillion-dollar infrastructure empire?
Inside the $1 Trillion AI Hardware Boom: Nvidia, AMD, and the Hyperscaler Arms Race to 2028
Overview
In 2026, hyperscalers like Microsoft, Alphabet, Amazon, and Meta are driving a historic surge in AI hardware demand by massively expanding data centers and capital expenditures. This fuels a fierce competition between Nvidia and AMD, with Nvidia maintaining dominance through its CUDA software lock-in and high margins, while AMD challenges with advanced hardware and open software. To optimize costs, hyperscalers increasingly adopt custom ASICs and dual-sourcing strategies, reducing reliance on Nvidia. However, physical limits at advanced foundries and geopolitical tensions—such as export controls and the push for sovereign AI—create new bottlenecks. If software monetization lags, the industry risks a sharp correction in hardware demand.