Law firm Latham & Watkins has purchased its own Nvidia GPU servers and begun fine-tuning open-weight AI models, marking the first public example of a major law firm owning AI hardware. The US's second-largest firm by revenue ($8.3bn in 2025) aims to keep highly sensitive client data entirely within its own systems and avoid escalating cloud consumption costs, according to Chief Information Officer Rene Mendoza. The firm leases secured data center space accessible only to its staff and employs over 900 technology specialists, including machine learning engineers customizing Nvidia Nemotron 3 models.

This capital-intensive approach contrasts with rivals: Kirkland & Ellis partners with Palantir, A&O Shearman works with Harvey, and Freshfields collaborates with Anthropic. Latham's strategy shifts AI from operating expense to upfront capital expenditure, which reduces partner profits — a traditional disincentive in Big Law. The investment likely runs into tens of millions annually. Joy Heath Rush, CEO of the International Legal Technology Association, called the move unprecedented. Wall Street banks are pressuring firms to cut fees as AI accelerates legal work, but Latham bets that infrastructure control offers long-term flexibility and security advantages.