Skip to main content
AWS logo beside a rack of Trainium3 UltraServers, with a Trainium4 chip diagram and a 6-FP4 performance boost label.

Editorial illustration for AWS Unveils Trainium3 UltraServers, Teases Trainium4 with 6x Performance Boost

Business & Startups

AWS Trainium3 Servers Supercharge AI Computing Power

AWS unveils Trainium3 UltraServers, previews Trainium4 with 6× FP4 boost

By AI Daily Post Edited by
• Updated: • 3 min read

AWS announced its Trainium3 UltraServers today. More importantly, the company has now deployed over one million of its custom Trainium accelerators. That scale is the foundation for its real play: using NVLink Fusion inside its MGX racks to directly wire those chips to NVIDIA GPUs and its own Graviton processors.

This builds an integrated system. And the teased Trainium4 chip, promising at least six times the FP4 processing punch, shows where this arsenal is headed next.

AWS also revealed early details of Trainium4, expected to deliver at least 6x the processing performance in FP4, along with higher FP8 performance and memory bandwidth. The next-generation chip will support NVIDIA NVLink Fusion interconnects to operate alongside NVIDIA GPUs and AWS Graviton processors in MGX racks. AWS has already deployed more than 1 million Trainium chips to date.

The company says the latest performance improvements translate to faster training and lower inference latency. In internal tests using OpenAI's GPT-OSS open-weight model, Trn3 UltraServers delivered three times higher throughput per chip and four times faster response times compared to Trn2 UltraServers.

Related Topics: #AWS #Trainium3 #Trainium4 #AI hardware #machine learning #cloud computing #custom silicon #OpenAI #GPUs #semiconductor

Take that internal GPT-OSS benchmark. Triple the throughput. Response times four times faster.

Those numbers prove the performance leap from Trainium2 to Trainium3. But the strategic shift is the hardware integration. NVLink Fusion physically binds AWS silicon to NVIDIA’s dominant GPUs.

With a million chips already in circulation, this is beyond testing. It is AWS building a complete, scaled hardware stack—a concrete alternative to relying solely on external suppliers.

Related Reading

Nordic classroom where a teacher demonstrates Gemini for Education on a laptop while students view NotebookLM on screen Business & Startups

Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy

• 3 min read
Kling exec unveils the sleek Video O1 on stage, holding the all-in-one unit beside a glowing MVL bridge and transformer Business & Startups

Kling launches Video O1, all-in-one model with MVL bridge using transformer

• 2 min read
DeepSeek AI startup team celebrating after securing a historic USD 7 billion funding round, signaling rapid growth and indust Business & Startups

DeepSeek Seeks More Capital Weeks After USD 7B Funding Round

• 4 min read
Tech reporter in studio beside a giant Doge meme screen, a glowing Facebook dating logo, and an Amazon AI robot arm. Related Topic

WIRED Roundup: DOGE Survives, Facebook Dating Launches, Amazon Chases AI Lead

Open Source

• 2 min read
Engineers and architects engage at the AWS AI Conclave Bengaluru 2026, exploring Agentic AI and data innovations. Related Topic

AWS AI Conclave Bengaluru 2026: Deep Tech Insights for Engineers and Architects

Market Trends

• 3 min read
AWS logo beside a sleek AI robot typing code on a laptop against a cloud-filled background From Archives

AWS launches 'frontier agents' AI that can code autonomously for days

• 3 min read
Matt Garman AWS chief executive stands at a podium delivering a keynote on cloud driven AI as industry peers watch and discuss product strategy From Archives

AWS CEO Matt Garman pushes cloud lead in AI era as peers urge product rethink

• 3 min read
Amazon executives unveil Leo satellite model beside a glowing data-center rack, with a 1.3 GW power board. From Archives

Amazon adds Leo satellite internet, pledges USD 50 bn for AI HPC, 1.3 GW compute

• 4 min read
Amazon team in a sleek data center watches monitors showing AI-crafted cyber-attack tactics and suggested patches. From Archives

Amazon Deploys AI Agents to Generate Offensive Techniques and Suggest Fixes

• 4 min read

Common Questions Answered

What performance improvements can users expect from AWS's Trainium3 UltraServers?

AWS's Trainium3 UltraServers represent a significant leap forward in machine learning infrastructure with enhanced computational capabilities. The new servers are designed to improve AI workload processing, demonstrating AWS's commitment to custom silicon development for advanced AI computing.

How does AWS plan to integrate Trainium4 with existing GPU technologies?

Trainium4 will support NVIDIA NVLink Fusion interconnects, allowing seamless operation alongside NVIDIA GPUs and AWS Graviton processors in MGX racks. This integration strategy enables a more flexible and powerful multi-processor ecosystem for handling complex AI workloads.

What performance boost is AWS promising with the upcoming Trainium4 chip?

AWS is teasing Trainium4 with a promised 6x performance improvement in FP4 processing, along with enhanced FP8 performance and increased memory bandwidth. These advancements are expected to translate into faster AI training times and lower inference latency for users.

Further Reading

PRESENTED BY NO CODE MBA

Ship an AI product this weekend — no engineers required.

Structured, in-depth lessons on the exact no-code tools — not scattered tutorials.

  • The exact platforms, taught in depth
  • Build real, working projects
  • Our honest review + a reader discount
Read the review →

Latest News

1

OpenAI's Miles Wang in Talks for USD 2B AI Drug Discovery Startup

2

Mistral Vibe for Code Leads in Multi-Agent Programming Benchmark

3

OpenAI's First Hardware Device Is a Movable, Screenless Speaker

4

PrismML's Bonsai 27B Runs Qwen3.6 on Laptops With 1-bit and Ternary Builds

5

OpenAI Targets 2027 for First Major Hardware: A ChatGPT Speaker

6

Publishers sue Google over unauthorized AI book training

7

Anthropic's Claude for Teachers Vows Not to Train on Student Data

8

DeepSeek Seeks More Capital Weeks After USD 7B Funding Round

9

Anthropic's New AI Ad Campaign Draws Criticism for 'Creepy' Tactics

10

DeepMind CEO proposes independent AI regulator as White House advisor voices skepticism

Most recent articles

LIVE03:21OpenAI's Miles Wang in Talks for USD 2B AI Drug Discovery Startup