Tuesday 15 September 2026
Evening EditionRSS

ZOTPAPER

News without the noise

ZOTPAPER
AI

Amazon triples Nvidia GPU order to 2 million chips over two years as AI demand surges

Cloud giant deepens partnership with chipmaker beyond pure hardware procurement

edit
By LineZotpaper
Published
Read Time2 min
Sources2 outlets
Amazon has placed an additional order for 2 million Nvidia GPU chips to be deployed across its data centers over the next two years, tripling its previous commitment. The expansion, driven by surging demand for AI workloads, is part of a broader partnership that extends beyond chip purchases, according to sources close to the deal.

Amazon Web Services (AWS), the world’s largest cloud provider, is dramatically scaling its Nvidia GPU footprint to keep pace with the explosive growth of artificial intelligence computing. The 2 million additional chips, likely a mix of H100, B200, or next-generation Blackwell GPUs, will be integrated into AWS’s global infrastructure over 2027 and 2028.

The move underscores the fierce competition among cloud giants — Microsoft Azure and Google Cloud have also announced large GPU procurement rounds — as enterprises race to train and deploy large language models and AI applications. “This tripling reflects the real demand we’re seeing from customers who want to build, fine-tune, and run AI at scale,” said an AWS spokesperson in a statement.

Beyond the chip order, Amazon and Nvidia are said to be expanding their collaboration on software orchestration, energy efficiency, and co-developed AI services. Details remain sparse, but the partnership is expected to include deeper integration of Nvidia’s CUDA ecosystem with AWS’s custom Trainium and Inferentia chips, as well as collaborative work on Nvidia’s DGX Cloud hosted on AWS.

Financial terms of the expanded deal were not disclosed, though analysts estimate the chip order alone could be worth billions of dollars at current Nvidia pricing. The announcement comes amid growing scrutiny of the energy consumption and water usage of massive AI data centers, with environmental groups urging greater transparency from both Amazon and Nvidia.

Critics also warn that such massive GPU deployments could entrench vendor lock-in, making it harder for AWS customers to migrate to competing AI accelerators. Meanwhile, Nvidia continues to face export control restrictions on selling advanced chips to certain countries, though the Amazon order is not expected to be affected.

§

Analysis

Why This Matters

  • Infrastructure race: The tripling confirms that cloud providers see AI compute as a non-negotiable strategic asset. AWS aims to maintain its lead over Azure and Google Cloud.
  • Cost implications: Massive GPU procurement will pressure AWS margins initially, but could lower AI inference prices for customers over time as scale drives efficiency.
  • Environmental impact: Data centers with millions of high-power GPUs will strain local power grids and water supplies, likely triggering more regulatory and community pushback.

Background

Amazon and Nvidia have been partners for over a decade, with AWS offering Nvidia GPUs in its EC2 instances since 2010. The relationship deepened in 2023 when AWS became the first cloud provider to offer Nvidia’s H100 GPUs. In early 2026, Amazon announced a $5 billion commitment to Nvidia chips over three years. This new order triples that commitment, reflecting the acceleration of AI adoption across industries. Nvidia now supplies GPUs to all major cloud providers, but Amazon’s AWS is the largest single customer by volume. The extended partnership also comes as Nvidia faces increasing competition from AMD, Intel, and custom chips from cloud providers themselves.

Key Perspectives

[AWS]: Amazon frames the deal as a response to customer demand. It strengthens AWS’s position as the most GPU-rich cloud, attracting AI startups and enterprises. The partnership also allows AWS to influence Nvidia’s roadmap for cloud-optimized chips. [Nvidia]: Securing a multi-million-chip order from the world’s largest cloud provider provides revenue certainty and validates Nvidia’s data center strategy. It also gives Nvidia a showcase for its latest architectures. [Competing cloud providers]:: Microsoft and Google will likely accelerate their own GPU purchases or in-house chip development (e.g., Azure’s Cobalt, Google’s TPU) to avoid falling behind on AI capacity. [Critics/Skeptics]: Environmental activists call for a moratorium on new data center construction until renewable energy and water recycling requirements are met. Some analysts question whether the demand is sustainable, warning of a potential GPU glut if AI adoption slows.

What to Watch

  • Delivery timelines: Whether Nvidia can deliver 2 million chips by 2028 given its own supply chain constraints (fabrication capacity at TSMC, advanced packaging).
  • AWS margin impact: How the massive hardware investment affects AWS’s operating margins in upcoming quarterly reports.
  • Regulatory developments: Potential new U.S. export controls that could affect Nvidia’s ability to serve Amazon’s global data centers in certain regions.
  • Custom chip progress: Amazon’s own Trainium and Inferentia chips — if they mature, they could reduce dependence on Nvidia long-term.

Sources

newspaper

Zotpaper

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.

Editorial ProcessSubscribe