Latham builds private Nvidia capacity and a 100-person AI team

On September 11, Bloomberg Law reported that Latham & Watkins had activated multiple Nvidia H200 GPUs in a third-party data center restricted to firm personnel and had roughly 100 employees dedicated to AI within a 900-person technology workforce. The firm declined to disclose the hardware program’s cost, while describing security, integration and model optionality as strategic benefits. This moves AI economics from per-seat subscriptions into capital planning, specialist payroll, utilization and infrastructure refresh cycles. Firms considering private capacity should compare total workload cost and control benefits with managed services, because underused hardware and scarce technical talent can erase savings from lower marginal inference costs.