Amazon’s Newest Gambit: Selling AI Chips

2026-06-19 · Source: Big Data & AI News - EE Times · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Amazon, a major hyperscaler, is expanding its in-house semiconductor business by planning to sell its custom AI chips, including Trainium accelerators, directly to other data center companies. This move follows Google's similar initiative with its TPUs. Amazon CEO Andy Jassy indicated this direction, noting the company's chip business already surpassed a \$20 billion annual revenue run rate. Amazon's portfolio includes Graviton CPUs, Trainium and Inferentia AI accelerators, and Nitro networking cards. Inferentia chips are claimed to be 40% cheaper for AI model inference. The company's chip trajectory began with the 2015 acquisition of Annapurna Labs, leading to the 2020 unveiling of Trainium1, followed by Trainium2 in 2024, offering 4x performance and 3x memory, and Trainium3 in late 2025, a 3-nm chip delivering up to 4x performance of its predecessor. Uber and Anthropic, which committed to deploying over one million Trainium chips, are early partners. While Amazon's efforts challenge Nvidia in specific workloads, Nvidia's full stack, including CUDA and developer tools, maintains its general-purpose AI infrastructure dominance.

Key takeaway

For AI Architects evaluating infrastructure, Amazon's move to sell its Trainium and Inferentia AI chips directly offers new options beyond Nvidia. You should assess these custom accelerators for specific generative AI training and inference workloads, especially given Inferentia's claimed 40% cost efficiency. Consider integrating AWS Neuron for your Trainium-based AI model development to leverage Amazon's growing ecosystem, potentially diversifying your hardware dependencies and optimizing costs for specialized tasks.

Key insights

Amazon and Google are selling custom AI chips, challenging Nvidia's market dominance in specific workloads.

Principles

• In-house silicon development reduces cloud costs.
• Full-stack ecosystems drive AI chip adoption.
• Custom chips excel in specific, optimized tasks.

Method

Amazon's chip development began with the 2015 Annapurna Labs acquisition, leading to custom AI silicon development around 2019 and subsequent Trainium accelerator generations (Trainium1, Trainium2, Trainium3).

In practice

• Consider Trainium3 for large-scale generative AI training.
• Evaluate Inferentia for 40% cheaper AI inference.
• Explore AWS Neuron for Trainium-based AI model development.

Topics

• AI Accelerators
• Amazon Web Services
• Trainium Chips
• Inferentia Chips
• Custom Silicon
• Data Center Infrastructure
• Generative AI

Best for: Investor, CTO, VP of Engineering/Data, Director of AI/ML, AI Architect, Tech Journalist

Related on AIssential

• Amazon Wants to Sell Its Own AI Chips — Amazon's direct AI chip sales could disrupt Nvidia's dominance but face complex revenue and supply chain trade-offs.
• An exclusive tour of Amazon’s Trainium lab, the chip that’s won over Anthropic, OpenAI, even Apple — Amazon Web Services (AWS) is significantly advancing its custom AI chip development, underscored by a \$50 billion investment deal with OpenAI that includes supplying 2 gigawatts of Trainium computing capacity.
• In another wild turn for AI chips, Meta signs deal for millions of Amazon AI CPUs — AWS secured a significant Meta deal for Graviton CPUs, intensifying cloud and custom AI chip competition.
• Amazon in Talks to Sell Custom AI Chips in Bid to Undercut Nvidia — Amazon aims to sell custom AI chips, directly challenging Nvidia's market dominance.
• Meta Is Developing 4 New Chips to Power Its AI and Recommendation Systems — Meta is developing custom MTIA chips to power its internal generative AI and content ranking systems.
• Can the Nvidia Monopoly on AI Chips Be Broken? — AI chip market diversity demands power-optimized, application-specific hardware supported by robust, user-friendly software stacks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Big Data & AI News - EE Times.