What is the latest version of AWS Trainium2?

The current shipped version of AWS Trainium2 is Trainium2, as of September 15, 2026.

How much does AWS Trainium2 cost?

AWS Trainium2 has no published flat entry price. Pricing not published; available via AWS EC2 instance pricing.

What AI model does AWS Trainium2 use?

AWS Trainium2 runs primarily on Neuron compiler.

← Chips & Compute

AWS Trainium2

Name: AWS Trainium2
Author: AWS

AWS · #5 most active of 18 in Chips & Compute

Compare →

Trn2 UltraServers and Project Rainier capacity, programmed through the Neuron SDK.

Trust score · Mostly verified

2 data points checked · 2026-09-11

Model stack: DisputedVersion: VerifiedHow this works →

Current version

Trainium2

Entry cost

Pricing not published; available via AWS EC2 instance pricing.

Changes / 30d

The short answers · verified September 15, 2026

What is the latest version of AWS Trainium2?: The current shipped version of AWS Trainium2 is Trainium2, as of September 15, 2026.
How much does AWS Trainium2 cost?: AWS Trainium2 has no published flat entry price. Pricing not published; available via AWS EC2 instance pricing.
What AI model does AWS Trainium2 use?: AWS Trainium2 runs primarily on Neuron compiler.

Source: AWS — aws.amazon.com · Reusable under CC BY 4.0 — cite Tomorrow

Capabilities

Trn2 UltraServer
64-chip NeuronLink domain
Neuron SDK
Rainier cluster
Generative AI training
Generative AI inference
Up to 20.8 FP8 petaflops
1.5 TB HBM3 memory
3.2 Tbps EFAv3 networking
NeuronLink chip interconnect
EC2 UltraClusters scale
AWS Neuron SDK integration
Native PyTorch & JAX support
Trn2 UltraServers support
NeuronLink interconnect
1.5 TB HBM3 accelerator memory
Up to 20.8 FP8 Petaflops of compute
EFAv3 networking (3.2 Tbps)
EC2 UltraClusters scalability
Up to 8 TB local NVMe storage
Neuron SDK integration
Neuron Kernel Interface (NKI)
16 AWS Trainium2 chips
Up to 20.8 FP8 PFLOPS compute
46 TBps memory bandwidth
NeuronLink chip-to-chip interconnect
AWS Neuron SDK support
EC2 UltraClusters support
3x more energy efficient than Trn1
Configurable FP8 (cFP8) support
Stochastic rounding & 4x sparsity (16:4)
Local NVMe storage up to 8 TB
EC2 UltraClusters scaling
EFAv3 networking
Neuron SDK support
PyTorch support
JAX support
vLLM support
Generative AI Training
Inference
NeuronLink Interconnect
EFAv3 Networking
PyTorch Support
JAX Support
vLLM Support
Elastic Fabric Adapter (EFAv3) networking
NeuronLink
EC2 UltraClusters
Generative AI Inference
AWS Neuron SDK
PyTorch integration
JAX integration
generative AI training
generative AI inference
AWS Nitro System encryption
FP8 compute
HBM3 memory
ML Training
ML Inference
High Bandwidth Memory
vLLM integration
AI Training
AI Inference
Gen AI Training
AWS Nitro System
Elastic Fabric Adapter (EFAv3)
FP8 compute support
Up to 20.8 FP8 Petaflops (Trn2)
Up to 83.2 FP8 Petaflops (UltraServer)
Up to 6 TB HBM (UltraServer)
Gen AI training
Gen AI inference
3.2 Tbps EFAv3
1.5 TB HBM3
Hugging Face integration
Elastic Fabric Adapter

Pricing

Pricing not published; available via AWS EC2 instance pricing.

Entry price over time

Not enough pricing history yet — we start charting from the second observation.

Geek mode

Models under the hood

Neuron compiler
AWS
disclosed
Trainium2
AWS
disclosed

Context window

—

Public API

yes

Multi-model routing

Stack signals

PyTorch
JAX
Hugging Face
Ray
Amazon SageMaker
Amazon EKS
Amazon ECS
AWS ParallelCluster
AWS Batch

Shared model stack

Other tracked products running on the same foundation models — a quick read on how much of the catalog moves when one of these models changes.

Neuron compilerAWS · 1 product
- AWS Trainium2
Trainium2AWS · 1 product
- AWS Trainium2

green disclosed · amber inferred · grey unattributed. Aliases are folded into one model; provider concentration counts only models with an established vendor.

Reported scale

Reference data. Each figure is whatever the source actually said — weekly users, downloads, revenue run-rate — with its own definition and date. These are not comparable between tools and are never used to rank anything.

No public usage figure on record for this product.

Also in Chips & Compute

All Chips & Compute

Tenstorrent Blackhole

Tenstorrent

Open RISC-V AI hardware

Version: Blackhole
Cost: from $999/unit
Model: Tensix cores

Google TPU Ironwood

Google

Inference-first TPU pods

Version: Ironwood (7th generation)
Cost: from $5.4/unit
Model: Gemini

Cerebras WSE-3

Cerebras

Wafer-scale inference and training

Version: WSE-3
Cost: Free tier / developer tier pricing available (contact/partner integrations)
Model: Codex-Spark

NVIDIA Blackwell

NVIDIA

GB200/B200 rack-scale AI systems

Version: Blackwell
Cost: Pricing is not publicly listed (enterprise hardware architecture/system)
Model: Blackwell

Intel Gaudi 3

Intel

Cost-focused training and inference accelerator.

Version: 1.24.1
Cost: Pricing not published. Available via Intel Tiber AI Cloud or OEM platforms.
Model: Accelerator

SambaNova SN40L

SambaNova

Dataflow chips for model serving

Version: SN40L
Cost: Free tier available. Paid plans include Developer (Pay-as-you-go) and Enterprise (Subscription-based).
Model: DeepSeek-V3.1

Change history

Sep 15, 2026first seenversion
AWS Trainium2 moved to Trainium2
AWS Trainium2Trainium2
source
Sep 14, 2026first seenversion
AWS Trainium2 moved to AWS Trainium2
Trainium2AWS Trainium2
source
Sep 13, 2026first seencapability
New capabilities: Elastic Fabric Adapter
75 tracked76 tracked · +Elastic Fabric Adapter
source
Sep 12, 2026first seencapability
New capabilities: Hugging Face integration
74 tracked75 tracked · +Hugging Face integration
source
Sep 10, 2026first seencapability
New capabilities: Gen AI training, Gen AI inference, 3.2 Tbps EFAv3
70 tracked74 tracked · +Gen AI training, Gen AI inference, 3.2 Tbps EFAv3, 1.5 TB HBM3
source
Sep 5, 2026first seencapability
New capabilities: Up to 20.8 FP8 Petaflops (Trn2), Up to 83.2 FP8 Petaflops (UltraServer), Up to 6 TB HBM (UltraServer)
67 tracked70 tracked · +Up to 20.8 FP8 Petaflops (Trn2), Up to 83.2 FP8 Petaflops (UltraServer), Up to 6 TB HBM (UltraServer)
source
Sep 3, 2026first seencapability
New capabilities: Elastic Fabric Adapter (EFAv3), FP8 compute support
65 tracked67 tracked · +Elastic Fabric Adapter (EFAv3), FP8 compute support
source
Aug 28, 2026first seencapability
New capabilities: Gen AI Training, AWS Nitro System
63 tracked65 tracked · +Gen AI Training, AWS Nitro System
source
Aug 26, 2026first seenversion
AWS Trainium2 moved to Trainium2
AWS Trainium2Trainium2
source
Aug 25, 2026first seenversion
AWS Trainium2 moved to AWS Trainium2
Trainium2AWS Trainium2
source
Aug 22, 2026first seencapability
New capabilities: AI Training, AI Inference
61 tracked63 tracked · +AI Training, AI Inference
source
Aug 20, 2026first seencapability
New capabilities: vLLM integration
60 tracked61 tracked · +vLLM integration
source
Aug 19, 2026first seencapability
New capabilities: ML Training, ML Inference, High Bandwidth Memory
57 tracked60 tracked · +ML Training, ML Inference, High Bandwidth Memory
source
Aug 18, 2026first seencapability
New capabilities: FP8 compute, HBM3 memory
55 tracked57 tracked · +FP8 compute, HBM3 memory
source
Aug 16, 2026first seencapability
New capabilities: AWS Nitro System encryption
54 tracked55 tracked · +AWS Nitro System encryption
source
Aug 14, 2026first seencapability
New capabilities: generative AI training, generative AI inference
52 tracked54 tracked · +generative AI training, generative AI inference
source
Aug 12, 2026first seencapability
New capabilities: PyTorch integration, JAX integration
50 tracked52 tracked · +PyTorch integration, JAX integration
source
Aug 11, 2026first seenversion
AWS Trainium2 moved to Trainium2
AWS Trainium2Trainium2
source
Aug 11, 2026first seencapability
New capabilities: Generative AI Inference, AWS Neuron SDK
48 tracked50 tracked · +Generative AI Inference, AWS Neuron SDK
source
Aug 10, 2026first seenversion
AWS Trainium2 moved to AWS Trainium2
Trainium2AWS Trainium2
source
Aug 10, 2026first seencapability
New capabilities: NeuronLink, EC2 UltraClusters
46 tracked48 tracked · +NeuronLink, EC2 UltraClusters
source
Aug 9, 2026first seencapability
New capabilities: Elastic Fabric Adapter (EFAv3) networking
45 tracked46 tracked · +Elastic Fabric Adapter (EFAv3) networking
source
Aug 8, 2026first seencapability
New capabilities: Generative AI Training, Inference, NeuronLink Interconnect
38 tracked45 tracked · +Generative AI Training, Inference, NeuronLink Interconnect, EFAv3 Networking, PyTorch Support, JAX Support, vLLM Support
source
Aug 7, 2026first seencapability
New capabilities: EC2 UltraClusters scaling, EFAv3 networking, Neuron SDK support
3210
source
Aug 6, 2026first seenversion
AWS Trainium2 moved to Trainium2
Trn2Trainium2
source
Aug 6, 2026first seencapability
New capabilities: Configurable FP8 (cFP8) support, Stochastic rounding & 4x sparsity (16:4), Local NVMe storage up to 8 TB
2910
source
Aug 5, 2026first seenversion
AWS Trainium2 moved to Trn2
Trainium2Trn2
source
Aug 5, 2026first seencapability
New capabilities: 16 AWS Trainium2 chips, Up to 20.8 FP8 PFLOPS compute, 46 TBps memory bandwidth
2210
source
Aug 4, 2026first seencapability
New capabilities: NeuronLink interconnect, 1.5 TB HBM3 accelerator memory, Up to 20.8 FP8 Petaflops of compute
1410
source
Aug 3, 2026first seencapability
New capabilities: Generative AI training, Generative AI inference, Up to 20.8 FP8 petaflops
410
source
Jul 2, 2026first seenversion
AWS Trainium2 moved to Trn2 / Trainium2
Trn2Trn2 / Trainium2
source
Jul 2, 2026first seencapability
New capabilities: Gen AI Compute, Trainium2 chips, 20.8 FP8 Petaflops
10 tracked17 tracked · +Gen AI Compute, Trainium2 chips, 20.8 FP8 Petaflops, 1.5 TB HBM3 memory, EFAv3 networking, Neuron SDK support, PyTorch & JAX support
source
Jun 19, 2026first seencapability
Trn2 UltraServers open to general customers
—64-chip NeuronLink domain
source
Jun 6, 2026first seencapability
New capabilities: Trn2 instances, Trn2 UltraServers, 16 Trainium2 chips
10 tracked20 tracked · +Trn2 instances, Trn2 UltraServers, 16 Trainium2 chips, 64 Trainium2 chips, 20.8 FP8 petaflops, 83.2 FP8 petaflops, 1.5 TB HBM3, 6 TB HBM3, 3.2 Tbps EFAv3, 12.8 Tbps EFAv3
source
May 24, 2026first seencapability
New capabilities: 1.5 TB HBM3 memory, 3.2 Tbps EFAv3 networking, NeuronLink interconnect
10 tracked19 tracked · +1.5 TB HBM3 memory, 3.2 Tbps EFAv3 networking, NeuronLink interconnect, AWS Nitro System encryption, EC2 UltraClusters scaling, FP32/TF32/BF16/FP16/cFP8 support, Stochastic rounding, 4x sparsity (16:4), Neuron Kernel Interface (NKI)
source
Apr 20, 2026first seencapability
New capabilities: 20.8 FP8 Petaflops, 1.5 TB HBM3, 3.2 Tbps EFAv3
9 tracked18 tracked · +20.8 FP8 Petaflops, 1.5 TB HBM3, 3.2 Tbps EFAv3, 8 TB local NVMe, FP32 support, TF32 support, BF16 support, FP16 support, cFP8 support
source
Mar 10, 2026first seencapability
New capabilities: Trainium2, NeuronLink, Up to 20.8 FP8 Petaflops
10 tracked18 tracked · +Trainium2, NeuronLink, Up to 20.8 FP8 Petaflops, 1.5 TB HBM3 memory, 46 TBps memory bandwidth, 3.2 Tbps EFAv3 networking, FP8 cFP8 support, Neuron SDK
source
Mar 10, 2026first seenversion
AWS Trainium2 moved to Trn2
Trainium2Trn2
source
Feb 17, 2026first seencapability
New capabilities: Generative AI training, Generative AI inference, NeuronLink chip interconnect
8 tracked18 tracked · +Generative AI training, Generative AI inference, NeuronLink chip interconnect, EFAv3 networking, AWS Nitro System encryption, FP8 FP16 FP32 TF32 BF16 support, Up to 83.2 FP8 petaflops, Up to 6 TB HBM3 memory, Local NVMe storage, PyTorch and JAX support
source
Jan 18, 2026first seenversion
AWS Trainium2 moved to Trainium2
Trn2Trainium2
source

Subscribe to AWS Trainium2 changes

AWS Trainium2 compared

Straight head-to-head pages against the busiest products in Chips & Compute.

AWS Trainium2 vs Tenstorrent Blackhole AWS Trainium2 vs Google TPU Ironwood AWS Trainium2 vs Cerebras WSE-3 AWS Trainium2 vs NVIDIA Blackwell AWS Trainium2 vs Intel Gaudi 3 AWS Trainium2 vs SambaNova SN40L

How to cite this page

Free to cite and reuse under CC BY 4.0. Permalink: https://tomorrow.aliensquad.ai/tools/aws-trainium

APA

Tomorrow. (2026). AWS Trainium2 — version, pricing and model stack [Data set entry]. AlienSquad. Retrieved 2026-09-17, from https://tomorrow.aliensquad.ai/tools/aws-trainium

BibTeX

@misc{tomorrow-tools-aws-trainium,
  author       = {{Tomorrow}},
  title        = {AWS Trainium2 — version, pricing and model stack},
  year         = {2026},
  publisher    = {AlienSquad},
  howpublished = {\url{https://tomorrow.aliensquad.ai/tools/aws-trainium}},
  note         = {Accessed: 2026-09-17}
}