Updated June 11, 2026

AI Power Rankings

IndependentΒ·Frontier AI lab index
← Companies
πŸ‡ΊπŸ‡Έ

Amazon

US

AWS Bedrock, Nova models, Trainium/Inferentia, Alexa+, Rufus.

Efforts: AWS Bedrock, Nova models, Trainium/Inferentia, Alexa+, Rufus
Rank
#7
Structural
8.16
Unweighted
8.27
vs #1
-1.16

Profile across ten areas

Area breakdown

🧠Model Quality
Core Engine Β· 18%
Amazon's first-party Nova models remain mid-tier and absent from frontier benchmark leadership, and the May 15 sunset of Amazon Q Developer concedes the coding-assistant category to partner models. The strategy works because Bedrock hosts the actual frontier (Claude Opus 4.8, GPT-5.5), but on its own models Amazon is a follower.
β€” Claude Fable 5
Claude: 5.6Gemini: 8.7GPT-5.5: 6.7Grok: 6.9
7.0
panel mean
πŸ”¬Research
Core Engine Β· 12%
Trainium silicon, the $110M Build on Trainium academic program (Berkeley, MIT, CMU), and agent-infrastructure primitives like AgentCore Payments show applied innovation. But Amazon produces few influential AI papers or landmark model breakthroughs relative to Google, OpenAI, or even Meta, and its retreat from first-party developer AI underscores the research gap.
β€” Claude Fable 5
Claude: 5.4Gemini: 8.6GPT-5.5: 7.5Grok: 7.1
7.2
panel mean
πŸ—οΈProduct
Delivery Β· 10%
Bedrock is now the most complete multi-vendor hub in the market β€” Claude Opus 4.8 and GPT-5.5/5.4/Codex GA same-cycle, AgentCore Payments with Coinbase/Stripe, Agent Toolkit, WorkSpaces for agents, Step Functions agentic reasoning, and a redesigned console (June 8). The weakness is that the crown-jewel experiences are built on rivals' models, and consumer products (Alexa for Shopping, Quick) trail ChatGPT in mindshare.
β€” Claude Fable 5
Claude: 8.2Gemini: 9.6GPT-5.5: 8.8Grok: 9.0
8.9
panel mean
πŸ’°Business
Accelerants & Stabilizers Β· 5%
AI revenue run-rate above $15B, AWS Q1 2026 revenue of $37.59B up 28% YoY, Pentagon classified-network contracts, and OpenAI usage now counting toward AWS commitments make Amazon one of the clearest commercial winners of the AI wave. Heavy capex is denting free cash flow, and AI margin depends partly on reselling partner models at first-party rates with no markup.
β€” Claude Fable 5
Claude: 8.7Gemini: 9.6GPT-5.5: 9.1Grok: 9.2
9.2
panel mean
⚑Compute
Core Engine Β· 14%
World-leading scale: AWS's global data-center footprint, custom Trainium/Inferentia silicon, Claude Platform live in 17 regions, GPU-accelerated OpenSearch Serverless, and continued massive capex. The only caveats are that Trainium has not displaced Nvidia for frontier training and Microsoft/Google match the hyperscale tier closely.
β€” Claude Fable 5
Claude: 9.1Gemini: 9.4GPT-5.5: 9.5Grok: 8.8
9.2
panel mean
πŸ›‘οΈSafety
Accelerants & Stabilizers Β· 6%
Bedrock Guardrails, GovCloud compliance, AWS-managed governance for OpenAI and Anthropic models, and enterprise-grade IAM controls give Amazon strong responsible-deployment infrastructure. It contributes comparatively little original alignment research, however, largely inheriting safety properties from the partner models it hosts.
β€” Claude Fable 5
Claude: 6.4Gemini: 8.8GPT-5.5: 7.7Grok: 7.8
7.7
panel mean
πŸ‘₯Talent
Delivery Β· 10%
Deep platform-engineering bench, stable leadership under Jassy and Garman, and university entrenchment via Trainium grants. Amazon is not a destination for frontier AI researchers the way OpenAI or Google DeepMind are, and the Q Developer/Rufus retirements suggest first-party AI product teams have struggled to meet the bar.
β€” Claude Fable 5
Claude: 6.2Gemini: 8.9GPT-5.5: 8.1Grok: 7.9
7.8
panel mean
🌐Distribution
Delivery Β· 8%
AWS is the default enterprise channel for both OpenAI (GA on Bedrock June 8, first-party pricing) and Anthropic (Claude Platform GA May 11), plus Alexa across hundreds of millions of devices and the world's largest e-commerce surface for Alexa for Shopping. Consumer AI assistant mindshare still lags badly versus ChatGPT and Gemini.
β€” Claude Fable 5
Claude: 8.5Gemini: 9.5GPT-5.5: 9.2Grok: 8.6
8.9
panel mean
🏰Data & Moats
Core Engine Β· 12%
Unmatched retail purchase data, AWS workload lock-in, IAM/billing/compliance switching costs, and the new dynamic where rival labs' usage accrues to AWS commitments create one of the strongest structural moats in AI. The counterpoint is that the model layer itself is rented β€” partners could renegotiate, and Amazon's own data hasn't yielded a frontier model.
β€” Claude Fable 5
Claude: 8.1Gemini: 9.4GPT-5.5: 9.0Grok: 8.3
8.7
panel mean
πŸš€Momentum
Accelerants & Stabilizers Β· 5%
Relentless cadence: AgentCore Payments (May 7), Agent Toolkit (May 11), Claude Platform GA (May 11), Transform Custom (May 18), OpenAI GA (June 1–8), OpenSearch Serverless rearchitecture (June 1). Momentum is concentrated in platform plumbing rather than headline models, and the Rufus and Q Developer retreats show first-party product execution still misfiring.
β€” Claude Fable 5
Claude: 7.3Gemini: 9.0GPT-5.5: 8.4Grok: 8.1
8.2
panel mean

News log

Amazon β€” News Log

Tracked for AI Power Rankings scoring. Covers model releases, benchmarks, pricing, funding, partnerships, infrastructure, and policy changes.


2026-06-08 β€” OpenAI GPT-5.5, GPT-5.4 and Codex reach general availability on Amazon Bedrock with first-party pricing

Published: 2026-06-08 | Logged: 2026-06-11T00:00Z | Area: Distribution & Reach, Product & Platform, Business & Market

AWS confirmed general availability of OpenAI's GPT-5.5, GPT-5.4, and Codex on Amazon Bedrock for production workloads, with the same AWS security, governance, and operational controls. Codex is accessible through the Codex App, CLI, and IDE integrations for VS Code, JetBrains, and Xcode; pricing matches OpenAI first-party rates and usage counts toward existing AWS commitments β€” a notable concession that removes the cost penalty for running OpenAI models through AWS rather than directly.

Scoring impact: Major positive Distribution & Reach (frontier OpenAI now in production on Bedrock with no pricing markup, drawing OpenAI spend onto AWS). Positive Product & Platform (broad IDE/Codex surface) and Business & Market (usage applies to AWS commitments).

Sources:



2026-06-08 β€” AWS Step Functions adds AgentCore-powered agentic reasoning step

Published: 2026-06-08 | Logged: 2026-06-11T00:00Z | Area: Product & Platform

AWS added an agentic reasoning step to AWS Step Functions, integrating the managed harness in Amazon Bedrock AgentCore directly into workflow orchestration. Developers can run multiple agents in parallel or sequence, insert human-approval gates, and trace every agent decision β€” embedding agent reasoning into AWS's mainstream serverless orchestration service rather than a standalone agent product.

Scoring impact: Positive Product & Platform (AgentCore reasoning becomes a native primitive inside Step Functions, hardening AWS's agent-orchestration stack and lowering adoption friction for existing Step Functions users).

Sources:



2026-06-08 β€” Amazon Bedrock ships redesigned console and CloudWatch metrics for OpenAI/Anthropic-compatible APIs

Published: 2026-06-08 | Logged: 2026-06-11T00:00Z | Area: Product & Platform

AWS released a redesigned Amazon Bedrock console optimized for OpenAI- and Anthropic-compatible APIs, featuring a model catalog, side-by-side comparison, project-based organization, and project-aware docs with pre-filled code snippets. In parallel, Bedrock added CloudWatch metrics for the bedrock-mantle endpoint (inference counts, input/output token totals, and client error counts at account, project, model, and project-and-model granularity), and AgentCore Identity gained bring-your-own-secrets support via AWS Secrets Manager.

Scoring impact: Positive Product & Platform (observability and a unified multi-vendor console make Bedrock easier to operate at scale, reinforcing its multi-model hub positioning vs. Azure/GCP).

Sources:



2026-06-01 β€” OpenAI frontier models and Codex generally available on Amazon Bedrock

Area: Distribution & Reach, Product & Platform, Business & Market

OpenAI frontier models and Codex moved to general availability on AWS through Amazon Bedrock, including commercial and GovCloud regions, with AWS-native security, governance, procurement, billing, and compliance controls. The launch β€” the production phase of the May 20 Bedrock preview β€” gives Amazon a frontier-OpenAI marketplace alongside Claude on Bedrock, Nova, and AWS's other model catalog, locking another major frontier lab into the AWS enterprise channel.

Scoring impact: Major positive Distribution & Reach (OpenAI now sold and billed through AWS), Product & Platform (Bedrock's model catalog hardens its multi-vendor pitch vs. Azure/GCP), and Business & Market (incremental AWS AI revenue against the already-cited $15B+ AI run-rate).

Sources:



2026-06-01 β€” Claude Opus 4.8 goes live on AWS via Bedrock and the Claude Platform

Published: 2026-06-01 | Logged: 2026-06-11T00:00Z | Area: Model Quality, Product & Platform, Distribution & Reach

AWS made Anthropic's most capable generally available model, Claude Opus 4.8, accessible through both Amazon Bedrock and the Claude Platform on AWS. Positioned for agentic coding, knowledge work, and extended autonomous task execution, Opus 4.8 sustains longer autonomous sessions with deeper reasoning and error recovery; on Bedrock it gains AWS-managed Guardrails, Knowledge Bases, and data residency, while the Claude Platform path unifies Anthropic's native APIs with AWS billing.

Scoring impact: Positive Distribution & Reach and Product & Platform (frontier Anthropic model available same-day across both AWS access tiers, deepening AWS as the default channel for Claude spend). Indirect Model Quality signal β€” AWS hosts the strongest available Claude generation.

Sources:



2026-06-01 β€” AWS launches next-generation Amazon OpenSearch Serverless purpose-built for agentic AI

Published: 2026-06-01 | Logged: 2026-06-11T00:00Z | Area: Product & Platform, Compute & Infra

AWS introduced a re-architected Amazon OpenSearch Serverless as a fully managed search and vector engine purpose-built for agentic AI applications. It scales from zero to thousands of requests per second (roughly 20x faster than the prior generation), delivers up to 60% cost savings versus peak-provisioned clusters, adds GPU acceleration and new SEARCH and VECTORSEARCH collection types, and ships native integrations with Vercel, Kiro, Claude Code, and Cursor via OpenSearch Agent Skills.

Scoring impact: Positive Product & Platform (retrieval/vector layer tailored to agent workloads, plugging directly into popular agent toolchains). Modest Compute & Infra lift via GPU-accelerated serverless economics.

Sources:



2026-05-20 β€” AWS deepens OpenAI alliance: GPT-5.5/5.4 + Codex on Bedrock, OpenAI-compatible SageMaker endpoints

Published: 2026-05-20 | Logged: 2026-05-21T09:00Z | Area: Distribution & Reach, Product & Platform, Compute & Infra

AWS announced an expanded OpenAI partnership: GPT-5.5 and GPT-5.4 are coming to Amazon Bedrock in limited preview, and Codex is now available on Bedrock via API (Codex CLI, desktop app, and VS Code extension). AWS also added OpenAI-compatible API support to SageMaker AI endpoints and launched Bedrock Managed Agents built on the OpenAI harness. Amazon shares rose ~2% to $265.01 on the news; AWS's AI revenue run-rate is now over $15B in the first three years of the AI wave.

Scoring impact: Major positive for Distribution & Reach (premier frontier-model marketplace) and Product & Platform (Managed Agents, drop-in OpenAI compatibility). Reinforces Compute & Infra positioning vs. Azure/GCP.

Sources:



2026-05-18 β€” AWS Transform turns one, launches "AWS Transform Custom" for code modernization at scale

Published: 2026-05-18 | Logged: 2026-05-19T09:00Z | Area: Product & Platform, Business & Market, Distribution & Reach

AWS marked the one-year anniversary of AWS Transform β€” its agentic AI service for migrating and modernizing enterprise applications β€” and announced AWS Transform Custom, extending the service into bespoke, large-scale code modernization engagements (e.g., legacy Java/.NET/COBOL portfolios). The launch ties AWS Transform tighter to AWS's enterprise sales motion and positions it directly against systems-integrator-led modernization work, complementing recent AWS moves like Claude Platform GA, Bedrock AgentCore Payments, and the Agent Toolkit.

Scoring impact: Positive Product & Platform (Transform service line maturing from single tool to platform family). Positive Business & Market (services-style revenue surface inside the enterprise install base). Modest Distribution & Reach lift as AWS competes for the consulting-side AI modernization budget.

Sources:



2026-05-18 β€” AWS launches EC2 M3 Ultra Mac instances; expands Build on Trainium with $110M for university research

Published: 2026-05-18 | Logged: 2026-05-19T09:00Z | Area: Compute & Infra, Talent & Org, Research & Innovation

AWS introduced Amazon EC2 M3 Ultra Mac instances built on Apple M3 Ultra Mac Studio hardware (28-core CPU, 60-core GPU, 256GB unified memory) β€” roughly 2Γ— the unified memory and 1.75Γ— the CPU cores versus the existing M4 Max Mac instances β€” aimed at iOS/macOS CI and on-device ML workloads. Separately, AWS expanded the Build on Trainium program with $110M in additional credits and grants for university researchers (including UC Berkeley, MIT, and Carnegie Mellon), giving academic teams access to Trainium clusters. AWS also added multicloud Interconnect connectivity to Oracle Cloud Infrastructure (preview) alongside Azure and Google Cloud.

Scoring impact: Positive Compute & Infra (Apple-silicon and Trainium expansion broaden AWS's accelerator menu). Talent & Org lift through deeper university-research entrenchment of the Trainium stack. Mild Research & Innovation tailwind via subsidized academic compute on AWS-native chips.

Sources:



2026-05-15 β€” AWS closes Amazon Q Developer to new signups; service to reach end-of-support in April 2027

Published: 2026-05-15 | Logged: 2026-05-15T09:00Z | Area: Product & Platform, Business & Market

AWS confirmed new signups for Amazon Q Developer are blocked starting May 15, 2026, with full end-of-support targeted for April 2027. The sunset funnels AWS's developer-AI demand toward Codex on Bedrock (announced May 5) and Claude Code on AWS, effectively conceding the standalone coding-assistant category to partner models (OpenAI, Anthropic) running on AWS infrastructure rather than AWS's own first-party developer assistant.

Scoring impact: Mixed Product & Platform β€” narrows AWS's first-party developer-AI surface (negative) while consolidating around stronger partner models on Bedrock (positive). Business & Market neutral: AWS retains the compute economics regardless of which assistant developers use.

Sources:



2026-05-13 β€” Amazon launches Alexa for Shopping; ditches Rufus chatbot

Published: 2026-05-13 | Logged: 2026-05-14T09:00Z | Area: Product & Platform, Distribution & Reach, Business & Market

Amazon announced "Alexa for Shopping," a new personalized AI shopping assistant powered by Alexa+ that takes over the search bar experience and replaces Rufus, the standalone generative AI shopping assistant launched in 2024. The new agent answers product queries and can take actions on behalf of users, consolidating Amazon's consumer AI strategy under the Alexa brand.

Scoring impact: Positive Product & Platform (clearer product unification, agentic shopping). Distribution & Reach lift (Alexa+ exposure across Amazon's enormous shopping audience). Business & Market positive if conversion lifts; signals Rufus did not meet bar.

Sources:



2026-05-11 β€” AWS launches Agent Toolkit and WorkSpaces for AI agents

Area: Product & Platform, Compute & Infra, Distribution & Reach

AWS rolled out the Agent Toolkit for AWS, a free production-ready suite of tools, guidance, and a managed remote MCP server that gives AI coding agents and assistants secure, authenticated access to all AWS services through a small fixed set of tools β€” aimed at reducing errors, token costs, and security risk when agents build on AWS. The same week AWS also previewed Amazon WorkSpaces for AI agents, letting agents securely operate desktop applications inside managed WorkSpaces environments. Together with the recently shipped Bedrock AgentCore Payments and Codex on Bedrock, the toolkit cements AWS as the most complete agent infrastructure stack in the hyperscaler market.

Scoring impact: Strong positive for Product & Platform (developer-facing agent toolchain) and Compute & Infra (deeper differentiation of AWS as agent runtime). Modest Distribution & Reach gain via lowered friction for builders.

Sources:



2026-05-11 β€” Claude Platform on AWS launches; AWS becomes first cloud with native Anthropic platform

Published: 2026-05-11 | Logged: 2026-05-13T09:00Z | Area: Product & Platform, Distribution & Reach, Business & Market

AWS announced general availability of Claude Platform on AWS, making it the first cloud provider to offer Anthropic's native Claude Platform experience β€” APIs, Claude Console, and early-access beta features β€” through customers' existing AWS accounts with AWS billing and IAM. The offering exposes the full Messages, Files, and Message Batches APIs, Claude Managed Agents, Agent Skills, code execution, web search/fetch, prompt caching, citations, batch processing, and the MCP connector. Opus 4.7, Sonnet 4.6, and Haiku 4.5 are live across 17 regions in North America, South America, Europe, and Asia Pacific. The offering sits alongside Bedrock's existing Claude access, giving customers a choice between Bedrock's managed multi-model service and Anthropic's native platform under AWS billing.

Scoring impact: Positive for Product & Platform (deeper Anthropic integration tier complementing Bedrock). Strong Distribution & Reach lift (locks more Anthropic spend onto AWS infrastructure). Positive for Business & Market (potential to recapture customers who would have gone direct to Anthropic).

Sources:



2026-05-07 β€” Amazon Bedrock AgentCore Payments preview launches with Coinbase and Stripe

Published: 2026-05-07 | Logged: 2026-05-08T09:00Z | Area: Product & Platform, Distribution & Reach

AWS announced the preview of AgentCore Payments inside Amazon Bedrock AgentCore, a capability that lets AI agents autonomously access and pay for APIs, MCP servers, web content, and other agents. The feature is built in partnership with Coinbase and Stripe, putting AWS among the first hyperscalers to ship native agent-to-agent commerce primitives.

Scoring impact: Positive for Product & Platform (novel agent-payments primitive on Bedrock). Distribution & Reach lift through Coinbase/Stripe partnerships, positioning Bedrock as a default settlement layer for agent commerce.

Sources:



2026-05-05 β€” AWS expands OpenAI partnership: Codex on Bedrock and Bedrock Managed Agents

Published: 2026-05-05 | Logged: 2026-05-08T09:00Z | Area: Product & Platform, Distribution & Reach, Business & Market

AWS and OpenAI announced an expanded partnership bringing Codex to Amazon Bedrock (Codex CLI, desktop app, and VS Code extension) and launching Amazon Bedrock Managed Agents β€” a new offering combining frontier OpenAI models with AWS infrastructure for production-ready OpenAI-powered agents using the OpenAI harness. GPT-5.5 and GPT-5.4 are available in preview, with authentication via AWS credentials and inference through Bedrock infrastructure. AWS CEO Matt Garman and OpenAI CEO Sam Altman jointly framed the deal as making Bedrock the easiest path to OpenAI-powered agents for AWS customers.

Scoring impact: Positive for Product & Platform (Codex + Managed Agents available natively on Bedrock). Strong Distribution & Reach (cements Bedrock as the multi-model hub for enterprise AI). Modest Business & Market upside through deeper OpenAI commercial alignment.

Sources:



2026-05-01 β€” Amazon signs Pentagon AI deal for classified military networks

Published: 2026-05-01 | Logged: 2026-05-02T09:00Z | Area: Distribution & Reach, Business & Market

AWS is among the companies that signed deals with the Department of Defense to integrate advanced AI into classified military networks.

Scoring impact: Positive for Distribution & Reach (government/defense contracts).

Sources:



2026-04-29 β€” Amazon Q1 2026 earnings: AWS revenue $37.6B, up 28%

Published: 2026-04-29 | Logged: 2026-05-02T09:00Z | Area: Business & Market, Compute & Infra

Amazon reported Q1 2026 earnings with AWS revenue of $37.59 billion, up 28% YoY. AI investments dented free cash flow but analysts viewed the growth trajectory positively. Amazon eyes $3 trillion market valuation as AI and AWS drive 2026 growth.

Scoring impact: Positive for Business & Market (strong AWS growth). Compute & Infra investment continues at scale.

Sources:



2026-04-28 β€” AWS launches Amazon Quick desktop app and expands Connect into agentic AI

Published: 2026-04-28 | Logged: 2026-05-02T09:00Z | Area: Product & Platform, Distribution & Reach

AWS launched Amazon Quick, an AI assistant with a desktop app that connects to local files, calendar, and communications. New Free and Plus pricing plans are available without an AWS account. AWS also expanded Amazon Connect into four agentic AI solutions: Decisions, Talent, Customer, and Health. Quick now integrates with Google Workspace, Zoom, Airtable, Dropbox, and Microsoft Teams.

Scoring impact: Major boost to Product & Platform (consumer/enterprise AI assistant). Positive for Distribution & Reach (cross-platform integrations, no AWS account required).

Sources:



2026-04-28 β€” AWS brings OpenAI GPT-5.5 to Amazon Bedrock; AI revenue run rate exceeds $15B

Published: 2026-04-28 | Logged: 2026-05-02T09:00Z | Area: Business & Market, Distribution & Reach

AWS and OpenAI are bringing GPT-5.5 and GPT-5.4 to Amazon Bedrock in preview, along with Codex and Bedrock Managed Agents powered by OpenAI. AWS's AI revenue run rate now exceeds $15 billion annually. AWS Q1 2026 revenue rose 28% to $37.59 billion.

Scoring impact: Positive for Business & Market (AI revenue milestone) and Distribution & Reach (multi-model platform strategy).

Sources:



← Higher rank
πŸ‡¨πŸ‡³ Alibaba
#6 Β· 8.23
Lower rank β†’
πŸ‡¨πŸ‡³ ByteDance
#8 Β· 8.01