AWS Trainium
Amazon's own training accelerator, rented by the hour inside EC2 instances rather than sold as a part.
Find alternatives to AWS Trainium
Consumed as EC2 capacity and billed by the hour; the silicon is not sold separately.
What it does
- Accelerator silicon
- Model training
How it compares
-
NVIDIA Blackwell
Named in NVIDIA's own 10-K among cloud companies designing competing AI silicon; the rivalry is in the accelerator, not in the cloud platform Trainium runs on.
large cloud services companies with internal teams designing hardware and software that incorporate accelerated or AI computing functionality as part of their internal solutions or platforms, such as Alibaba Group, Alphabet Inc., Amazon, Inc., or Amazon, Baidu, Inc., Huawei, and Microsoft Corporation, or Microsoft
Regulatory filing · 25 Aug 2026 -
Intel Gaudi AI Accelerator
Both are data-centre AI accelerators aimed at the same workloads; Intel sells its silicon to cloud operators while Amazon builds Trainium for its own, so they meet as vendor against in-house design.
DCAI competitors include AMD, which utilizes the x86 architecture and competes with us across the full spectrum of CPUs, GPUs, accelerators and other products; providers of GPU systems such as NVIDIA, whose GPU systems have experienced the highest demand in the market as AI workloads have become the dominant source of compute demand; companies developing their own custom silicon, including many hyperscalers such as Amazon, Google, Meta and Microsoft
Regulatory filing · 25 Aug 2026 -
Nebius AI Cloud
We also compete with general purpose cloud computing providers including Amazon (AWS), Google (Google Cloud Platform), Microsoft (Azure), and Oracle.
Regulatory filing · 18 Sep 2026 -
GroqCloud
Beyond Nvidia, Groq competes with Amazon, Google and Microsoft, all of which offer -- or will soon offer -- custom chips for AI workloads in the cloud.
News reporting · 19 Sep 2026
Sources
- Pricing
-
Official documentation · 22 Aug 2026
Consumed as EC2 capacity and billed by the hour; the silicon is not sold separately.
- Sold within
-
Trainium is not sold as a part: it is reached by renting the EC2 instance types built on it (Inf2, Trn1/Trn2), which are lines on EC2's own rate card. The chip has no price of its own anywhere on aws.amazon.com.
Also from Amazon
-
Amazon Kendra Data service
A managed AWS enterprise search service that answers natural-language queries and powers retrieval-augmented generation over indexed content.
-
Amazon Lex API service
A managed AWS service for building voice and text conversational bots, built on the same technology as Alexa.
-
Amazon Nova Act Agent platform
An AWS platform for building and running AI agents that automate browser-based UI workflows, billed per agent-hour of active work.
-
Amazon Polly API service
A managed AWS API that converts text into synthesized speech across more than 40 languages and voice styles.
-
Amazon Q Developer Developer tool
An AWS coding assistant that generates, reviews, tests and autonomously modifies code inside IDEs and the CLI.
-
Amazon Quick Agent platform
AWS's subscription AI work assistant combining chat-based research, business-intelligence dashboards and workflow automation, positioned as the successor to Amazon Q Business.
-
Amazon Rekognition API service
A managed AWS API for image and video analysis, including object and face detection, text extraction from images and content moderation.
-
Amazon SageMaker AI Platform
AWS's unified platform for building, training and deploying machine learning and generative AI models.
-
Amazon Textract API service
A managed AWS API that extracts text, handwriting, forms and tables from scanned documents.
-
Amazon Transcribe API service
A managed AWS API that converts speech to text using an automatic speech recognition model.
-
Amazon Bedrock Model API
AWS service that serves models from several providers behind one API, with tooling for agents and evaluation.
-
AWS Inferentia Hardware
AWS's custom silicon for machine learning inference, purchased via Amazon EC2 Inf1 and Inf2 instances rather than as a standalone chip.
-
Alexa+ Application
Amazon's generative-AI Alexa: a conversational assistant across Echo devices and the Alexa app that handles multi-step requests, remembers context and acts on behalf of the user.
-
Amazon Comprehend API service
Natural-language-processing service that extracts key phrases, sentiment, entities, document classification and PII redaction from text.
-
Amazon Connect Customer Platform
Cloud contact-centre service, formerly simply 'Amazon Connect', now positioned as an agentic-AI customer-experience solution: AI agents and intelligent routing across voice, chat, email, SMS, web and WhatsApp, designed on a no-code canvas, with real-time performance tracking for agents and supervisors.
-
Amazon Connect Health Agent platform
Agentic AI for healthcare providers and health-tech developers: patient verification, appointment management, ambient clinical documentation, patient insights and medical coding, delivered inside Amazon Connect Customer for patient engagement and via an SDK at the point of care.
-
Kiro Developer tool
Agentic AI development platform (IDE, CLI, web, mobile) that turns prompts into executable specs, validates code with property-based testing and runs parallel agents across large codebases.
-
Amazon Personalize API service
Recommendation engine trained on user-interaction data to deliver real-time personalized recommendations across web, app and marketing channels.
-
Amazon Translate API service
Neural machine translation service for batch and real-time cross-lingual text translation with brand-terminology customization.
-
Amazon Bedrock AgentCore Platform
Production AI-agent operational platform (framework- and model-agnostic) providing security controls, tool integration, debugging and evaluation for deployed agents.
-
AWS HealthScribe API service
HIPAA-eligible service that transcribes patient-clinician conversations and generates preliminary clinical documentation with evidence mapping back to the transcript.
-
AWS Transform Platform
Agentic-AI enterprise IT transformation workbench that accelerates cloud migration, application modernization (mainframe, Windows, custom code) and continuous tech-debt reduction.
Something wrong here? Send a correction — quote this product id: aws-trainium.