Latham & Watkins is taking artificial intelligence in-house - literally!
Why Latham Wants Its Own AI Infrastructure
Most firms access generative Al through third-party platforms and cloud providers. Latham, however, recently announced an investment in Nvidia-powered servers to build a hybrid system: using commercial models like Harvey where they make sense, while running or adapting open-weight models (models whose architecture but not training data is public) internally for other tasks.
With sensitive information at stake, greater control over data processing gives Latham more flexibility in meeting confidentiality and client expectations. Owning computing capacity also reduces reliance on external vendors and gives the firm strategic flexibility as the market evolves.
The Legal Teams Behind the Machines
Building private AI infrastructure creates legal work inside the firm itself.
Technology and commercial lawyers will be central to negotiating hardware, software, data-centre and vendor agreements, including liability, service levels and intellectual-property rights.
Data privacy and cybersecurity specialists must determine how client and personal information can be processed, stored and protected. Professional-responsibility and risk teams will need controls governing confidentiality, privilege, supervision and acceptable AI use.
IP lawyers may also address licensing restrictions surrounding open-weight models, training material and internally developed tools. Employment teams could become relevant where AI systems affect workflows, monitoring, recruitment or staff responsibilities.
The result is unusual: Latham is simultaneously the legal adviser, technology operator and regulated professional-services business.
What Happens Next?
Latham may be the first major firm publicly known to buy its own AI servers, but Big Law is already moving towards a much more expensive battle over who controls the technology behind legal work.
The clearest evidence is Kirkland & Ellis. The world’s highest-grossing law firm has set aside $500mn to develop its own proprietary AI platform, including more than $100mn of spending this year and hundreds of millions more over the next three to four years. With Kirkland generating around $10bn in annual revenue, building technology in-house is becoming a serious strategic investment rather than an experimental IT budget.
Not every firm can afford that game. But when Kirkland is prepared to commit half a billion dollars, and Latham is buying the hardware itself, the direction is becoming clearer. Big Law’s next competitive divide may not be between firms that use AI and those that do not. It could be between firms that rent their intelligence and those wealthy enough to build their own.