AWS Trainium2 · Trainium

667 teraflops of dense BF16 compute a chip; AWS publishes no power figure

Shipped 3 Dec 2024; shipped, no dated promise on record. Epoch AI names it at 5 AI data centers and no GPU cluster. AWS across the site.

Embed

Trainium2

<iframe src="https://jesseahawkins.com/chips/embed/accelerators/aws-trainium2/" title="AWS Trainium2: 667 teraflops of dense BF16 compute a chip; AWS publishes no power figure" width="100%" height="520" style="border:0;max-width:100%" loading="lazy"></iframe>

The frame credits AWS, Epoch AI, AI data centers and Epoch AI, Data on GPU clusters and links back to its page. Preview it.

Cite

This page

Jesse Hawkins, “AWS Trainium2: 667 teraflops of dense BF16 compute a chip; AWS publishes no power figure”, jesseahawkins.com, data retrieved 23 Oct 2026. https://jesseahawkins.com/chips/accelerators/aws-trainium2/. Data: Each vendor’s own datasheets and releases, quoted; Epoch AI, “AI data centers” and “Data on GPU clusters”, CC BY 4.0.

BibTeX

@misc{hawkins2026chipsacceleratorsawstrainium2,
  author = {Hawkins, Jesse},
  title = {{AWS Trainium2: 667 teraflops of dense BF16 compute a chip; AWS publishes no power figure}},
  year = {2026},
  howpublished = {\url{https://jesseahawkins.com/chips/accelerators/aws-trainium2/}},
  note = {Data: Each vendor’s own datasheets and releases, quoted; Epoch AI, ``AI data centers'' and ``Data on GPU clusters'', CC BY 4.0, retrieved 2026-10-23}
}

The data · Epoch AI

Epoch AI, 'AI data centers'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/ai-data-centers' [online resource]. CC BY 4.0.

The clusters · Epoch AI

Epoch AI, 'Data on GPU clusters'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/gpu-clusters' [online resource]. CC BY 4.0.

CSV Post on X

Each figure in AWS’s own words

one chip, dense compute; memory is high-bandwidth memory

BF16, dense: 667 TFLOPS 667 BF16/FP16/TF32 TFLOPS Trainium2 architecture (AWS Neuron documentation)
FP8, dense: 1,299 TFLOPS 1,299 FP8 TFLOPS Trainium2 architecture (AWS Neuron documentation)
Memory: 96 GiB 96 GiB of device memory with 2.9 TB/sec of bandwidth Trainium2 architecture (AWS Neuron documentation)
Memory bandwidth: 2.9 TB/s 96 GiB of device memory with 2.9 TB/sec of bandwidth Trainium2 architecture (AWS Neuron documentation)

The AWS Neuron docs give per-chip figures; the Trn2 page gives only 16-chip instance figures, so the docs are the source. AWS states no chip power. The launch release of 2023-11-28 and the Amazon filings of early 2024 give no availability window, so there is no promise. The Trn2 page states no on-demand price.

The shipment

each date links to the page it is quoted from

3 Dec 2024 Shipped: Trn2 instances generally available. Today, AWS announces the general availability of Amazon Elastic Compute Cloud (Amazon EC2) Trn2 instances Amazon EC2 Trn2 instances are generally available

Named at 5 AI data centers

H100-equivalents and power are the whole site’s today, with every chip it holds

AI data centers that name this chip
Site	Owner	Country	H100e	MW	Chips
Anthropic-Amazon New Carlisle	Amazon	United States	818,428	1,244	Trainium2
Amazon Madison Mega Site	Amazon	United States	214,348	341	Trainium2
AWS New Albany	Amazon	United States	201,112	255	Trainium2, H100, H200, B200
Amazon Ridgeland	Amazon	United States	171,299	273	Trainium2, Trainium3
AWS Berwick	Amazon	United States	81,354	139	Trainium2, Trainium3

In no GPU cluster of Epoch AI’s list

Epoch AI’s list of AI supercomputers, which stops in July 2025

Epoch AI’s names for this chip: Trainium2.

Other AWS chips

Trainium3 671 teraflops of dense BF16 compute a chip; AWS publishes no power figure

Sources

accelerators.csv Every dataset → Status → Glossary →

AWS’s own pages, each linked beside the figure or date it gives. Each page was read and each quote checked on 23 Oct 2026.
Where it runs: Epoch AI, AI data centers and Epoch AI, GPU clusters, CC BY 4.0.