Investors pivot to GPUs as long-term AI infrastructure

Nuway Capital has expanded its focus on GPU and data-centre infrastructure after new research conducted with KPMG highlighted a shift in investor interest towards physical AI computing assets.

The research draws on a survey of 120 investors across 10 international markets and positions GPUs as a long-term infrastructure asset class rather than a short-term investment theme. The report links this shift to rising demand for AI computing and ongoing supply constraints.

Nearly 80% of respondents identified generative AI as the primary driver behind their interest in GPU investments. More than 70% of high-net-worth investors, and 60% of family offices and wealth managers, ranked GPUs ahead of blockchain and quantum computing. Investors also placed GPUs ahead of renewable energy in the survey results.

The research highlighted limitations across chip manufacturing, power availability and land suitable for new data-centre developments. These constraints were cited as reinforcing the view of GPU capacity as infrastructure-like, with characteristics shaped by long-term demand and physical bottlenecks.

The report pointed to a growing gap between demand for computing capacity and available infrastructure. It referenced House of Commons Library projections that global data-centre capacity will increase by around 15% year-on-year through to 2027, a pace the research said would fall short of demand driven by AI and cloud adoption.

Rising demand for GPUs continues to be constrained by limited chip production, while power availability and land scarcity restrict where large-scale AI systems can operate.

Nuway Capital said these dynamics are shifting investor attention away from individual AI vendors and towards infrastructure purpose-built for AI workloads.

"Demand for GPUs is no longer episodic - it's structural," said Colin Bosher, Founder, Nuway Capital. "The economics are becoming more infrastructure-like: sustained demand, physical bottlenecks, and a growing emphasis on operational efficiency and asset selection."

The survey focused on high-net-worth individuals, family offices and wealth managers, with findings showing greater confidence in GPU-linked investment themes relative to other emerging technologies.

More than 75% of respondents described GPU-as-a-Service business models as either promising or highly innovative. The report identified GPU-as-a-Service as one of several models emerging to provide access to compute capacity.

Nuway Capital said it is expanding its advisory and platform activity around GPU capacity and high-performance computing infrastructure, including GPU-as-a-Service offerings. The firm described these structures as lowering barriers for organisations unable to finance dedicated GPU infrastructure, while providing investors with exposure to recurring, usage-based revenues.

KPMG said access to physical infrastructure is becoming an increasingly important factor in how investors assess AI exposure. The firm also pointed to constraints in power supply and data-centre development as key drivers of investment risk and return assumptions.

One section of the report examined how physical infrastructure constraints could reshape the economics of AI deployment, with implications for data-centre development pipelines and procurement of specialist compute capacity.

"The next phase of artificial intelligence will be shaped less by software innovation alone and more by access to physical infrastructure. GPUs, data centres and power availability are becoming the real constraints, and that shift is fundamentally changing how investors assess risk, return and long-term value in the AI ecosystem," said Chris Brown, Partner and Head of KPMG Strategy in Ireland, KPMG.

Nuway Capital launched in July 2024 and positions itself as an alternative investment and infrastructure advisory firm focused on AI and GPU-linked digital infrastructure.

KPMG operates as a global professional services network across audit, tax and advisory services. The research, titled "Investing in GPUs – Paper II", examines supply-side constraints, investor appetite and business models across the GPU and data-centre market.

The findings come as developers and operators compete for power connections and suitable sites for new data-centre builds, while chip supply remains concentrated among a small number of manufacturers. The research highlighted these factors as central to investor interest in GPU-related assets over the medium term.

Nuway Capital said it plans to continue expanding its activity around GPU capacity and related infrastructure models as demand for AI computing grows.