Anthropic has secured contracts totalling $65 billion to support Claude models, while Google and Amazon plan around 10 GW of capacity.

Scale of commitments

The announced commitments reflect large-scale investments in data-centre capacity and power consumption, comparable to the electricity needs of a small European country.

Such capacity enables sustained training and inference workloads for large language models, and it raises questions about allocation, cost and user access patterns.

Potential for personalized instances

Part of this infrastructure can be allocated to individual users via subscription access, allowing Claude to run tailored workloads on dedicated resources.

Providers may offer per-user or per-tenant deployments that leverage the same underlying compute fabric while isolating data and performance characteristics.

Pilot: Claude applied to DeFi analytics

In a pilot demonstration, an expert showed Claude parsing on-chain data and proposing decentralized finance strategies based on observed movements and yields.

The experiment emphasised natural-language interaction: questions are posed in plain English and Claude returns strategy suggestions without requiring code from the user.

Technical components used in the pilot

• 3 connectors described as MCP, which link Claude to a server, a database and a browser interface.
• 5 data sources supplying prices, yields and large-holder movements, initially available without fees.
• Configuration and orchestration contained in a single file, with setup time reported as 15 minutes and initial cost at $0.

The pilot illustrates how large-scale infrastructure commitments can translate into end-user functionality, enabling model-driven services that interact directly with public blockchain data.