GPU-as-a-Service

AI Demand Reshapes Data Centers and Fuels NeoCloud Growth

Artificial intelligence is changing how data centers work and creating space for GPU-as-a-Service providers, often called NeoClouds.¹

These firms buy Graphics Processing Units (“GPUs”) , place them in data centers and sell computing power on demand. Clients avoid upfront costs, long contracts and heavy infrastructure investments whilst gaining flexible and scalable access to high-performance compute.

Even hyperscalers now use NeoClouds to meet AI demand. Deals like Microsoft and Google tapping CoreWeave show how this model is moving into the mainstream. Their success lies in lowering barriers to entry for AI workloads, speeding time to market and offering the flexibility to scale capacity in line with fast-changing demand.

Investors looking at data centers and GPUaaS² must weigh two central issues: counterparty risk and the impact of rapid hardware evolution. For data center investors, this means scrutinizing tenant contracts, identifying the ultimate GPU capacity off-takers and potential capex burden of keeping pace with chipset advances. For investors in the GPUaaS space, key questions include the long-term competitive advantage of NeoCloud against hyperscalers, the scale of price erosion with each hardware cycle and the ability to repurpose servers to preserve asset value. Addressing these questions early allows investors to position themselves in a market segment that sits at the core of AI adoption, with potential for outsized returns as demand for compute power accelerates.

Slide