ChipsSeptember 10, 2026via Financial Times Technology

Latham & Watkins buys Nvidia servers to set up in-house AI systems

Why it matters

Enterprise shift from API-first to on-prem AI infrastructure. A marquee professional-services firm is investing in capex and custom models rather than relying on OpenAI/Anthropic — a signal that on-prem economics are winning for regulated, data-sensitive workloads.

Key signals

  • Latham & Watkins purchasing Nvidia servers for on-premise deployment
  • Firm customizing open-weight models instead of using OpenAI/Anthropic APIs
  • Data-sensitive legal work driving shift to on-prem infrastructure
  • Enterprise move from cloud API consumption to capex-based hardware ownership
  • Latham & Watkins purchasing Nvidia servers for in-house deployment
  • Law firm customizing open-weight models instead of relying on OpenAI/Anthropic APIs
  • On-prem AI strategy in legal services sector
  • Alternative to proprietary cloud-hosted models

The hook

Major law firm ditches the cloud: Latham & Watkins buys Nvidia servers to run custom open-weight models in-house.

Law firm is customising open-weight models as alternative to OpenAI and Anthropic

The week's key stories, every Friday.

For practitioners and enthusiasts — free, in your inbox.

Free forever. No spam.