Who really controls the GPUs?

This article was produced by Gilbert + Tobin, a member of IPFA

As demand for artificial intelligence (AI) computing capacity accelerates, GPU-as-a-Service (GPUaaS) is becoming an increasingly important part of the digital infrastructure landscape. However, the separation between asset ownership, operational control and contractual responsibility raises important questions for investors, financiers and infrastructure providers.

In this article, Gilbert + Tobin Partner Ross Phillipson and Special Counsel James Frixou examine the cybersecurity, operational and export-control risks associated with GPUaaS arrangements, and their implications for project bankability. They explore why understanding who controls the infrastructure supporting AI compute, rather than simply who owns the assets, is critical to managing risk, protecting revenue streams and securing investment.

The insights highlight key considerations for providers, sponsors, financiers and data-centre operators as they navigate the evolving commercial and regulatory landscape for AI infrastructure.