The AI chip race heats up as AWS unveils Trainium2 (T2) chips, designed to train and deploy large language models faster than Nvidia GPUs though Blackwell is expected to surpass them.
In a surprising move Apple announced it’s using AWS chips for search and exploring their use for Apple Intelligence
In a surprising move Apple announced it’s using AWS chips for search and exploring their use for Apple Intelligence
AWS' Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025 | TechCrunch
At its re:Invent conference, AWS today announced the general availably of its Trainium2 (T2) chips for training and deploying large language models
techcrunch.com
December 3, 2024 at 9:41 PM
The AI chip race heats up as AWS unveils Trainium2 (T2) chips, designed to train and deploy large language models faster than Nvidia GPUs though Blackwell is expected to surpass them.
In a surprising move Apple announced it’s using AWS chips for search and exploring their use for Apple Intelligence
In a surprising move Apple announced it’s using AWS chips for search and exploring their use for Apple Intelligence
AWS’ Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025
AWS’ Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025
At its re:Invent conference, AWS today announced the general availably of its Trainium2 (T2) chips for training and deploying large language models (LLMs). These chips, which AWS first announced a year ago, will be four times as fast as their…
tcrn.ch
December 3, 2024 at 4:38 PM
AWS’ Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025
This represents about a fifth of all Trainium2 chips, according to our AI Chip Sales data (epoch.ai/data/ai-chi...).
January 9, 2026 at 10:56 PM
This represents about a fifth of all Trainium2 chips, according to our AI Chip Sales data (epoch.ai/data/ai-chi...).
I mean, the real reason is "a founder testimonial onstage of Trainium2 was doubtless a contractual requirement for the investment round."
December 3, 2024 at 4:52 AM
I mean, the real reason is "a founder testimonial onstage of Trainium2 was doubtless a contractual requirement for the investment round."
Just a reminder. The hyperscalers want their own AI chips.
Amazon AWS just launched its Trainium3 AI chip
– Up to 4.4× higher performance
– ~4× more memory bandwidth
– 4× better performance-per-watt vs Trainium2
– Scale to 144 chips (362 FP8 PFLOPs) per server and 1M+ chips in UltraClusters
Amazon AWS just launched its Trainium3 AI chip
– Up to 4.4× higher performance
– ~4× more memory bandwidth
– 4× better performance-per-watt vs Trainium2
– Scale to 144 chips (362 FP8 PFLOPs) per server and 1M+ chips in UltraClusters
December 2, 2025 at 8:01 PM
Just a reminder. The hyperscalers want their own AI chips.
Amazon AWS just launched its Trainium3 AI chip
– Up to 4.4× higher performance
– ~4× more memory bandwidth
– 4× better performance-per-watt vs Trainium2
– Scale to 144 chips (362 FP8 PFLOPs) per server and 1M+ chips in UltraClusters
Amazon AWS just launched its Trainium3 AI chip
– Up to 4.4× higher performance
– ~4× more memory bandwidth
– 4× better performance-per-watt vs Trainium2
– Scale to 144 chips (362 FP8 PFLOPs) per server and 1M+ chips in UltraClusters
🤖 Claude 3.5 Haiku llega a AWS Trainium2 + destilación en Bedrock
https://claude.com/blog/trainium2-and-distillation
#Claude #AWS #IA #Bedrock
https://claude.com/blog/trainium2-and-distillation
#Claude #AWS #IA #Bedrock
July 19, 2026 at 11:25 AM
🤖 Claude 3.5 Haiku llega a AWS Trainium2 + destilación en Bedrock
https://claude.com/blog/trainium2-and-distillation
#Claude #AWS #IA #Bedrock
https://claude.com/blog/trainium2-and-distillation
#Claude #AWS #IA #Bedrock
Trainium2 is now GA a year after being announced and they're terrific--just kidding, they're comparative crap compared to Trainium3, which you should totally hold out for at next year's re:Invent.
AWS has apparently never heard of the Osborne Effect.
AWS has apparently never heard of the Osborne Effect.
AWS’ Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025
AWS’ Trainium2 chips for building LLMs are now generally available, with Trainium3 coming in late 2025
At its re:Invent conference, AWS today announced the general availably of its Trainium2 (T2) chips for training and deploying large language models (LLMs). These chips, which AWS first announced a year ago, will be four times as fast as their…
tcrn.ch
December 4, 2024 at 12:03 AM
Trainium2 is now GA a year after being announced and they're terrific--just kidding, they're comparative crap compared to Trainium3, which you should totally hold out for at next year's re:Invent.
AWS has apparently never heard of the Osborne Effect.
AWS has apparently never heard of the Osborne Effect.
Poolside is mentioned as a Trainium2 customer. That's Jason Warner's thing...
December 3, 2024 at 4:27 PM
Poolside is mentioned as a Trainium2 customer. That's Jason Warner's thing...
AWS says Anthropic's Claude Opus 4 launched on its Trainium2 GPUs, and Project Rainier, AWS' supercomputer built for Anthropic, uses 500K+ Trainium2 chips (Kristina Partsinevelos/CNBC)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
June 18, 2025 at 9:45 AM
AWS says Anthropic's Claude Opus 4 launched on its Trainium2 GPUs, and Project Rainier, AWS' supercomputer built for Anthropic, uses 500K+ Trainium2 chips (Kristina Partsinevelos/CNBC)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
An example ...
"Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia."
www.cnbc.com/2025/06/17/a...
"Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia."
www.cnbc.com/2025/06/17/a...
AWS' custom chip strategy is showing results, and cutting into Nvidia's AI dominance
Anthropic's Claude Opus 4 launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of its own.
www.cnbc.com
June 18, 2025 at 3:03 AM
An example ...
"Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia."
www.cnbc.com/2025/06/17/a...
"Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia."
www.cnbc.com/2025/06/17/a...
Talking extensively about the Trainium2 / Graviton4 packaging.
There's zero chance this wins AWS any business (directly); it's clear that Peter DeSantis is reveling in the geekery here, and I'm totally here for it. Not everything has to be a sales pitch.
There's zero chance this wins AWS any business (directly); it's clear that Peter DeSantis is reveling in the geekery here, and I'm totally here for it. Not everything has to be a sales pitch.
December 3, 2024 at 4:20 AM
Talking extensively about the Trainium2 / Graviton4 packaging.
There's zero chance this wins AWS any business (directly); it's clear that Peter DeSantis is reveling in the geekery here, and I'm totally here for it. Not everything has to be a sales pitch.
There's zero chance this wins AWS any business (directly); it's clear that Peter DeSantis is reveling in the geekery here, and I'm totally here for it. Not everything has to be a sales pitch.
AWS announces general availability of Trainium2 chips, and EC2 Trn2 UltraServers, which feature 64 Trainium2 chips; Trainium3 is slated for 2025 (Frederic Lardinois/TechCrunch)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
December 3, 2024 at 5:35 PM
AWS announces general availability of Trainium2 chips, and EC2 Trn2 UltraServers, which feature 64 Trainium2 chips; Trainium3 is slated for 2025 (Frederic Lardinois/TechCrunch)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
$AMZN AWS is building over 1.3GW of datacenter capacity for Anthropic -- hosting nearly 1M Trainium2 chips.
Anthropic is now at a $5B run-rate (5x YoY) -- with AWS as its main training partner and inference largely on $GOOGL Cloud TPUs.
Anthropic is now at a $5B run-rate (5x YoY) -- with AWS as its main training partner and inference largely on $GOOGL Cloud TPUs.
September 4, 2025 at 4:20 AM
$AMZN AWS is building over 1.3GW of datacenter capacity for Anthropic -- hosting nearly 1M Trainium2 chips.
Anthropic is now at a $5B run-rate (5x YoY) -- with AWS as its main training partner and inference largely on $GOOGL Cloud TPUs.
Anthropic is now at a $5B run-rate (5x YoY) -- with AWS as its main training partner and inference largely on $GOOGL Cloud TPUs.
WSJ published an article about the new Trainium2 cluster which has a super-cool photo of an UltraServer, as well as details about AWS's next flagship training cluster, Rainier.
https://www.wsj.com/articles/amazon-announces-supercomputer-new-server-powered-by-homegrown-ai-chips-18c196fc
#HPC #AI
https://www.wsj.com/articles/amazon-announces-supercomputer-new-server-powered-by-homegrown-ai-chips-18c196fc
#HPC #AI
December 3, 2024 at 6:45 PM
WSJ published an article about the new Trainium2 cluster which has a super-cool photo of an UltraServer, as well as details about AWS's next flagship training cluster, Rainier.
https://www.wsj.com/articles/amazon-announces-supercomputer-new-server-powered-by-homegrown-ai-chips-18c196fc
#HPC #AI
https://www.wsj.com/articles/amazon-announces-supercomputer-new-server-powered-by-homegrown-ai-chips-18c196fc
#HPC #AI
Amazon promises 4x faster AI silicon in 2025, turns Trainium2 loose on the net
Amazon promises 4x faster AI silicon in 2025, turns Trainium2 loose on the net
Tens of thousands of AWS’ Trn2 instances to fuel Anthropic's next-gen models
Re:Invent Amazon Web Services teased its next gen AI accelerator dubbed Trainium3 at re:Invent on Tuesday, which it says will deliver 4x higher performance than its…
dlvr.it
December 3, 2024 at 6:38 PM
Amazon promises 4x faster AI silicon in 2025, turns Trainium2 loose on the net
AppleはAmazon製のTrainiumとGravitonチップで検索サービスを提供中、Trainium2チップをAIモデルの事前トレーニングに使用できるかどうかも評価中であることを明かす
https://gigazine.net/news/20241204-apple-amazon-chip/
https://gigazine.net/news/20241204-apple-amazon-chip/
AppleはAmazon製のTrainiumとGravitonチップで検索サービスを提供中、Trainium2チップをAIモデルの事前トレーニングに使用できるかどうかも評価中であることを明かす
Amazonが2024年12月3日に開催されたAmazon Web Services(AWS)のイベント「AWS re:Invent 2024」にて、AIトレーニング用チップ「Trainium2」の一般提供を開始したことを発表しました。イベントにはAppleの幹部も登壇し、AppleでのAmazon製チップの使用例について報告しています。
gigazine.net
December 4, 2024 at 3:31 AM
AppleはAmazon製のTrainiumとGravitonチップで検索サービスを提供中、Trainium2チップをAIモデルの事前トレーニングに使用できるかどうかも評価中であることを明かす
https://gigazine.net/news/20241204-apple-amazon-chip/
https://gigazine.net/news/20241204-apple-amazon-chip/
Based on our discussions with experts, we don't expect this type of cooling to be feasible for the most powerful AI servers, like NVIDIA Blackwell and beyond. But for this data center, AWS uses their custom Trainium2 servers, which are less power-dense.
January 9, 2026 at 10:56 PM
Based on our discussions with experts, we don't expect this type of cooling to be feasible for the most powerful AI servers, like NVIDIA Blackwell and beyond. But for this data center, AWS uses their custom Trainium2 servers, which are less power-dense.
AWS has lagged behind Azure and Google Cloud in cloud AI adoption. But Anthropic may be its ace as Amazon’s building custom Trainium2 chips tailored for Claude, which is rapidly gaining traction.
If AWS becomes the best Claude host, it could steal Claude users from Google Cloud and regain ground.
If AWS becomes the best Claude host, it could steal Claude users from Google Cloud and regain ground.
Amazon’s AI Resurgence: AWS & Anthropic’s Multi-Gigawatt Trainium Expansion
Two-and-a-half years ago, we flagged a looming “cloud crisis” at AWS. Today, the evidence has mounted. AWS is the crown jewel of the Amazon empire, generating ~60% of group profits, and dominating …
semianalysis.com
September 7, 2025 at 4:17 PM
AWS has lagged behind Azure and Google Cloud in cloud AI adoption. But Anthropic may be its ace as Amazon’s building custom Trainium2 chips tailored for Claude, which is rapidly gaining traction.
If AWS becomes the best Claude host, it could steal Claude users from Google Cloud and regain ground.
If AWS becomes the best Claude host, it could steal Claude users from Google Cloud and regain ground.
"Here's how we're using Trainium2 to build the most powerful AI server ever."
Were that true, why would the recent $4 billion investment in Anthropic require them to use these "most powerful AI servers ever?" Wouldn't they naturally be beating AWS's door down to use them?
Instead, NVIDIA.
Were that true, why would the recent $4 billion investment in Anthropic require them to use these "most powerful AI servers ever?" Wouldn't they naturally be beating AWS's door down to use them?
Instead, NVIDIA.
December 3, 2024 at 4:15 AM
"Here's how we're using Trainium2 to build the most powerful AI server ever."
Were that true, why would the recent $4 billion investment in Anthropic require them to use these "most powerful AI servers ever?" Wouldn't they naturally be beating AWS's door down to use them?
Instead, NVIDIA.
Were that true, why would the recent $4 billion investment in Anthropic require them to use these "most powerful AI servers ever?" Wouldn't they naturally be beating AWS's door down to use them?
Instead, NVIDIA.
Amazonが3nmの新AIチップ「Trainium3」を発表、「Trainium2」より4倍高速＆コスト最大50％削減＆さらに「Trainium4」も予告
https://gigazine.net/news/20251203-amazon-trainium-3/
https://gigazine.net/news/20251203-amazon-trainium-3/
Amazonが3nmの新AIチップ「Trainium3」を発表、「Trainium2」より4倍高速＆コスト最大50％削減＆さらに「Trainium4」も予告
Amazon Web Services(AWS)が、最新世代のTrainiumチップである「Trainium3」をUltraServerサービスを通じて提供開始しました。Trainium3は同社初の3nmというプロセスノードで製造されたAIチップであり、前世代と比較してさまざまな性能が向上しています。
gigazine.net
December 3, 2025 at 2:40 AM
Amazonが3nmの新AIチップ「Trainium3」を発表、「Trainium2」より4倍高速＆コスト最大50％削減＆さらに「Trainium4」も予告
https://gigazine.net/news/20251203-amazon-trainium-3/
https://gigazine.net/news/20251203-amazon-trainium-3/
#reinvent2024 keynote, announcing Trainium2 and then Trainium3 next year
December 3, 2024 at 4:38 PM
#reinvent2024 keynote, announcing Trainium2 and then Trainium3 next year
🚀 AWS Project Rainier goes live with 500k Trainium2 chips - one of world's largest AI compute clusters
🤖 Powering @Anthropic's Claude
#AWS #AI #CloudComputing
🤖 Powering @Anthropic's Claude
#AWS #AI #CloudComputing
AWS activates Project Rainier: One of the world’s largest AI compute clusters comes online
Our new AI “compute cluster’ will connect hundreds of thousands of Trainium2 chips across the U.S.
ift.tt
November 7, 2025 at 9:43 AM
🚀 AWS Project Rainier goes live with 500k Trainium2 chips - one of world's largest AI compute clusters
🤖 Powering @Anthropic's Claude
#AWS #AI #CloudComputing
🤖 Powering @Anthropic's Claude
#AWS #AI #CloudComputing
Look ma, no #GPUs
AWS is building a machine with hundreds of thousands of #Tranium2 processors to achieve performance of ~65 ExaFLOPS for #AI
AWS also unveiled its Trainium3 processor which will quadruple performance compared to the Trainium2
www.tomshardware.com/tech-industr...
#HPC
AWS is building a machine with hundreds of thousands of #Tranium2 processors to achieve performance of ~65 ExaFLOPS for #AI
AWS also unveiled its Trainium3 processor which will quadruple performance compared to the Trainium2
www.tomshardware.com/tech-industr...
#HPC
AWS building ExaFLOPS-class supercomputer for AI with hundreds of thousands homegrown Trainium2 processors — AWS forges a path without Nvidia GPUs
130 FP8 ExaFLOPS?
www.tomshardware.com
December 5, 2024 at 2:05 PM
Look ma, no #GPUs
AWS is building a machine with hundreds of thousands of #Tranium2 processors to achieve performance of ~65 ExaFLOPS for #AI
AWS also unveiled its Trainium3 processor which will quadruple performance compared to the Trainium2
www.tomshardware.com/tech-industr...
#HPC
AWS is building a machine with hundreds of thousands of #Tranium2 processors to achieve performance of ~65 ExaFLOPS for #AI
AWS also unveiled its Trainium3 processor which will quadruple performance compared to the Trainium2
www.tomshardware.com/tech-industr...
#HPC
There will be lots of superficial pieces written about AWS Trainium2 and the various skus. And then there is this piece from semianalysis. Long read, worth every moment you spend on it. semianalysis.com/2024/12/03/a... #reinvent
Amazon’s AI Self Sufficiency | Trainium2 Architecture & Networking
Amazon is currently conducting one of the largest build-out of AI clusters globally, deploying a considerable number of Hopper and Blackwell GPUs. In addition to a massive Capex invested into Nvidi…
semianalysis.com
December 4, 2024 at 11:22 AM
There will be lots of superficial pieces written about AWS Trainium2 and the various skus. And then there is this piece from semianalysis. Long read, worth every moment you spend on it. semianalysis.com/2024/12/03/a... #reinvent
@awscloud.bsky.social just unveiled Project Rainier, a distributed AI supercluster powered by Trainium2 chips that delivers up to 5x more compute power than @anthropic.com’s largest cluster.
See how it is reshaping the future of AI training: awsinsider.net/articles/202...
#AWS #AI #Trainium2
See how it is reshaping the future of AI training: awsinsider.net/articles/202...
#AWS #AI #Trainium2
AWS Says Project Rainier Ushers New Era of AI Training Superclusters -- AWSInsider
AWS's new Project Rainier supercluster uses Trainium2 chips and distributed multi-site design to power next-generation AI model training, offering five times more compute than existing clusters.
awsinsider.net
November 5, 2025 at 3:21 PM
@awscloud.bsky.social just unveiled Project Rainier, a distributed AI supercluster powered by Trainium2 chips that delivers up to 5x more compute power than @anthropic.com’s largest cluster.
See how it is reshaping the future of AI training: awsinsider.net/articles/202...
#AWS #AI #Trainium2
See how it is reshaping the future of AI training: awsinsider.net/articles/202...
#AWS #AI #Trainium2