[Photo: Shutterstock]

U.S. law firm Latham & Watkins has directly purchased Nvidia GPU servers and started fine-tuning AI models, the Financial Times (FT) reported on Thursday.

The report said Latham & Watkins is seen as the first major law firm to buy its own AI hardware and fine-tune models.

Latham & Watkins, which posted revenue of $8.3 billion last year, bought multiple servers equipped with high-performance GPUs for running AI models. Its engineers are currently adjusting existing open-weight AI models to fit the firm’s needs.

By building its own servers, Latham & Watkins must handle cybersecurity and server operations itself. In return, it can store sensitive client information on its own systems without exposing it to external cloud services, FT reported.

Latham & Watkins engineers are adjusting existing open-weight AI models to fit the firm’s needs.

Rene Mendoza (르네 멘도자), Latham & Watkins' chief information officer, said, "There are times when you don’t want to entrust very sensitive client information to any cloud provider." He added, "We also wanted to secure flexibility, as we expect the cost burden to grow depending on how AI usage fees are charged."

Legal work involves large amounts of information and many repetitive tasks, making it one of the fields where AI-driven change is among the fastest. Wall Street banks are also demanding fee cuts from major law firms, saying AI speeds up work.

Law firms have relied on software subscriptions and cloud infrastructure rather than AI hardware, saying upfront investment reduces partner payouts. Latham & Watkins' move runs counter to that trend, FT reported.

Building its own technology does not mean Latham & Watkins is breaking entirely with AI products and services from major technology companies. It plans to use both its own models and tech companies’ AI. Latham & Watkins said it employs more than 900 technology specialists.

Keyword

#Latham & Watkins #Nvidia #GPU servers #open-weight AI models #Financial Times
Copyright © DigitalToday. All rights reserved. Unauthorized reproduction and redistribution are prohibited.