Grid capacity and power supplySecuring stable, reliable multi-gigawatt power supply, especially from renewable sources, is the primary constraint for scaling data centers.
Operational expertiseAnthropic has been a research company; building and running hyperscale data centers requires different operational skills and culture.
Capital efficiencyRecouping $50B in capital requires generating enough revenue from model inference and services; costs must stay competitive with cloud pricing.
Physics limits
Power density limitsPhysics of cooling and power distribution limit how densely GPUs/AI accelerators can be packed per facility, capping throughput.
Transmission lossesMoving multi-terabit data between locations has latency and power costs; a single centralized facility is more efficient than distributed compute.
How it works
4 parts
Custom design
Optimized for Anthropic's workload
Fluidstack builds data centers for Claude training and inference, not generic cloud.
Ownership
Anthropic control
Unlike AWS/Google Cloud, Anthropic owns and operates these facilities, giving it direct control over capacity, maintenance and hardware.
Scaling strategy
Gigawatt-scale ambition
Anthropic plans to grow these facilities to multi-gigawatt capacity, enabling training of increasingly capable models at the frontier.
Financial backing
$50B committed capital
The $50B outlay aligns with Anthropic's internal projections: $70B revenue and $17B positive cash flow by 2028.
The team behind Anthropic US Data Centers (with Fluidstack)
Anthropic
Claude models, Claude Code, AI safety
Anthropic is an AI safety company that makes the Claude models. Former OpenAI researchers founded it in 2021. It studies how models work inside, and Claude is widely used for coding and business agents.
Raised $65B at a $965B valuation in May 2026, after a $30B round in February.
Claude Code, launched in 2025, made Claude a leading model for AI coding agents.
Trains on Google, Amazon and NVIDIA chips instead of relying on one supplier.
