Latham & Watkins buys Nvidia-powered AI servers in what is reported as a BigLaw first, building bespoke in-house models to keep sensitive client data off external cloud systems

Latham & Watkins has purchased its own Nvidia-powered servers, each fitted with multiple GPUs (graphics processing units, the chips that provide the computing power to run large AI models), to build and customise AI models entirely in-house. The move is described as a first for a major law firm. Rather than routing client data through external cloud providers such as OpenAI, Google or Anthropic, Latham's own engineers are using the hardware to adapt so-called open-weight AI models, which can be downloaded and fine-tuned on private infrastructure. The firm continues to use familiar tools including ChatGPT, Claude and Gemini alongside the in-house capability. Latham's chief information officer Rene Mendoza explained the rationale: 'Sometimes we may have information that is so sensitive, client information that we really want to protect, we don't want to put it to any cloud vendor.' The firm employs more than 900 technology specialists, and the servers are housed in locked data-centre space accessible only to Latham staff. No acquisition cost has been disclosed, though a set-up of this scale, including hardware and specialist staff, can run to tens of millions of dollars annually. Michael Rubin, a Latham partner and chair of its AI strategy committee, noted that the number of technology specialists, innovation lawyers and AI-focused roles has grown significantly in recent years. The move deliberately reduces dependency on any single AI vendor. Other elite firms have taken different routes: Kirkland & Ellis is working with Palantir, A&O Shearman has partnered with Harvey, and Freshfields announced a partnership with Anthropic earlier in 2026.