---
title: "Amazon CEO Andy Jassy Confirms Continued Reliance on Nvidia Chips"
description: "Amazon CEO Andy Jassy has reaffirmed the company's commitment to Nvidia hardware while simultaneously scaling its own custom AI silicon business to an annual run rate exceeding $20 billion."
author: "CryptoResearch AI"
published: "2026-09-19T10:03:29.130Z"
updated: "2026-09-19T10:03:29.131Z"
category: "ai"
reading_time_minutes: 2
content_type: "editorial"
canonical: "https://cryptoresearch.news/news/amazon-ceo-andy-jassy-confirms-continued-reliance-on-nvidia-chips"
tags: ["Amazon", "Nvidia", "AI", "AWS", "Hardware", "Cloud"]
---

# Amazon CEO Andy Jassy Confirms Continued Reliance on Nvidia Chips

> Editorial content, written by CryptoResearch.

Amazon CEO Andy Jassy has reaffirmed the company's commitment to Nvidia hardware while simultaneously scaling its own custom AI silicon business to an annual run rate exceeding $20 billion.

Amazon CEO Andy Jassy has clarified the company's stance on its hardware strategy, emphasizing that Nvidia will remain a core part of its infrastructure. In an April 2026 letter to shareholders, Jassy noted that nearly all AI development to date has relied on Nvidia chips and confirmed that AWS intends to remain a primary destination for running them.

The company is currently executing a dual-track strategy. In August 2026, Amazon committed to purchasing 2 million additional Nvidia GPUs for deployment between 2027 and 2028. This order adds to a previous commitment of 1 million chips, creating a total pipeline of approximately 3 million units.

Simultaneously, Amazon is scaling its internal AI chip business, which includes the Graviton, Trainium, and Nitro product lines. This division has reached an annual revenue run rate of over $20 billion. Jassy suggested this figure could reach $50 billion if Amazon decided to commercialize these chips for external use.

Amazon's current AI training chip, Trainium2, is reportedly nearly sold out and offers roughly 30% better price-performance than comparable GPUs. The upcoming Trainium3, scheduled for launch in early 2026, is expected to provide a 30-40% performance improvement over its predecessor and is already nearly fully subscribed.

The company expects its Trainium line to eventually generate annual capital expenditure savings in the tens of billions of dollars. Jassy acknowledged that many customers will continue to prefer Nvidia for specific workloads, particularly those requiring the mature CUDA software ecosystem and high-end performance.

Amazon's approach mirrors that of other major hyperscalers, including Google, Microsoft, and Meta, all of which are developing custom silicon to improve internal economics while maintaining relationships with Nvidia. For Amazon, the goal for its custom chips is to provide a cost-effective alternative for routine inference and fine-tuning tasks.

Market participants are monitoring the $20 billion revenue run rate of Amazon's custom chip business. A shift toward the $50 billion external commercialization scenario could signal a change in how AI compute is priced and provisioned across the industry.

## Sources

- [Crypto Briefing](https://cryptobriefing.com/amazon-jassy-nvidia-chips-long-term/)

---

Published by CryptoResearch. Canonical version: https://cryptoresearch.news/news/amazon-ceo-andy-jassy-confirms-continued-reliance-on-nvidia-chips