Goldman Sachs Sees AI Investment Shifting to Data Centre Infrastructure

TL;DR:

• Goldman Sachs Research forecasts AI workloads could represent 30% of total data centre capacity within two years, with global power demand rising 175% by 2030
• The investment bank describes a “flight to quality” as investors move away from speculative AI software plays and towards companies controlling physical infrastructure
• For UK data centre operators, the shift presents both opportunity and challenge as energy grid constraints become a strategic bottleneck

The initial gold rush phase of generative AI investing is giving way to something more measured. Goldman Sachs Research now characterises the market as entering a “flight to quality,” where investors are increasingly drawn to companies that own and operate the physical infrastructure underpinning AI — data centres, computing hardware, and networking systems — rather than those merely attaching the AI label to their products.

The Numbers Behind the Shift

The scale of projected demand is striking. Goldman Sachs estimates that AI workloads could account for roughly 30% of all data centre capacity within two years, driven by both model training and inference at scale. The energy implications are equally significant: global data centre power demand could rise approximately 175% by 2030 compared with 2023 levels — equivalent to adding the electricity consumption of a top-10 power-consuming country to the global grid.

Hyperscale cloud providers are already spending tens of billions of dollars annually on new facilities and hardware. But the challenge extends well beyond capital expenditure. Large data centre projects involve complex supply chains, grid connections, and long-term energy agreements, with construction timelines measured in years rather than months.

Why This Matters for UK Businesses

The UK’s data centre sector faces a particular version of this challenge. Grid capacity constraints and planning delays have already slowed several proposed facilities. As global demand intensifies, securing adequate power supply and suitable sites will become increasingly competitive. Companies that already control established data centre networks hold a structural advantage — a pattern that echoes earlier computing infrastructure cycles where platform builders captured more stable returns than application developers.

Looking Forward

The Goldman Sachs analysis reinforces a broader pattern: as AI matures from experimental technology to production infrastructure, the value chain is tilting towards those who build the foundations rather than those who build on top of them. For UK investors and technology firms, this points toward data centre capacity, energy infrastructure, and cooling technology as the durable plays in the AI economy — areas where physical constraints, not software innovation, may ultimately determine who wins.