NewsNTech
Running an open-weight language model requires dedicated GPU compute. That hardware requirement is what commercial API arrangements sidestep, and it is now where Latham & Watkins is putting capital.
The law firm is buying Nvidia servers to stand up in-house AI systems, customizing open-weight models as a direct alternative to commercial services from OpenAI and Anthropic.
Open-weight models, unlike the closed systems that OpenAI and Anthropic offer through their APIs, can be downloaded and adapted by the operator. The adapted weights still need somewhere to run.
Latham & Watkins is choosing owned infrastructure over third-party API dependency. The firm moves from consuming AI as a service to operating it directly, with the Nvidia hardware as the enabling compute floor.
Keep reading