← The AI Systems Map
L1 · CloudCloud / computeUnited StatesAMZN

Amazon Web Services (AWS)

Largest hyperscaler; deepest custom-silicon stack (Trainium, Inferentia, Graviton).

Website ↗

AWS is the world's largest cloud provider and Anthropic's primary training and inference partner. It designs its own AI accelerators (Trainium/Inferentia) and Arm CPUs (Graviton) via the Annapurna Labs team, reducing dependence on Nvidia. In 2024-2025 Amazon committed over $8B to Anthropic and built Project Rainier, a massive Trainium2 cluster (hundreds of thousands of chips) for Anthropic model training. Still buys huge volumes of Nvidia GPUs to serve GPU-demanding customers.

Custom AI chips

Trainium2/Trainium3, Inferentia

Anthropic investment

~$8B+ total; Project Rainier cluster

Custom CPU

Graviton4 (Arm-based)

Market position

#1 cloud by revenue

How it fits the stack

Amazon Web Services (AWS) with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.

usesusesusesused byhostsdepends onsuppliesdepends onhostsdepends ondepends onpartners withpartners withpartners withinvests inAmazon Web Services(AWS)CloudAMDArm HoldingschokepointArm Neoverse / CSSserver IPchokepointAWS Trainium /InferentiaDigital RealtyElectricity Grid &UtilitieschokepointInventecAmazon AGI (Nova)AnthropicPerplexity AIEquinixPalantir AIPPalantir Technologies
Amazon Web Services (AWS)Depends on ↑Feeds ↓Related

Amazon Web Services (AWS) in the AI stack. Amazon Web Services (AWS) with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.

Graph data (text) — 14 entities, 15 relationships
  • Amazon Web Services (AWS) —uses→ AMD
  • Amazon Web Services (AWS) —uses→ Arm Holdings
  • Amazon Web Services (AWS) —uses→ Arm Neoverse / CSS server IP
  • Amazon Web Services (AWS) —used by→ AWS Trainium / Inferentia
  • Amazon Web Services (AWS) —hosts→ Digital Realty
  • Amazon Web Services (AWS) —depends on→ Electricity Grid & Utilities
  • Amazon Web Services (AWS) —supplies→ Inventec
  • Amazon AGI (Nova) —depends on→ Amazon Web Services (AWS)
  • Anthropic —hosts→ Amazon Web Services (AWS)
  • Anthropic —depends on→ Amazon Web Services (AWS)
  • Perplexity AI —depends on→ Amazon Web Services (AWS)
  • Amazon Web Services (AWS) —partners with→ Equinix
  • Amazon Web Services (AWS) —partners with→ Palantir AIP
  • Amazon Web Services (AWS) —partners with→ Palantir Technologies
  • Amazon Web Services (AWS) —invests in→ Anthropic

Depends on ↑ · 12

usesAMDOffers AMD Instinct MI300-series instances as a Nvidia alternativeusesArm Holdings⬦Graviton CPUs and Trainium/Inferentia are built on Arm IP/architectureusesArm Neoverse / CSS server IP⬦AWS Graviton built on Arm coresused byAWS Trainium / InferentiaDeployed on AWS EC2; Project Rainier with AnthropichostsDigital RealtyHyperscale capacity leased to AWSdepends onElectricity Grid & Utilities⬦Gigawatt-scale datacenter load strains regional gridssuppliesInventecHGX AI servers to hyperscalersusesLiquid CoolingLiquid cooling for Trainium and GB200 racksdepends onNuclear Power (incl. SMRs)⬦Signed nuclear PPAs (e.g. Talen/Susquehanna) to power AI datacentersusesNvidia⬦Buys massive volumes of Nvidia GPUs (H100/H200/GB200) for EC2 GPU instanceshostsQTS Data CentersHyperscale pre-leases to cloud operatorspowersX-energyPlanned SMR capacity (incl. Energy Northwest, WA) intended to power AWS data centers
← Back to the AI Systems Map