UBS is putting a number on the AI infrastructure boom, and it’s a big one. The bank’s research team projects value creation in the AI infrastructure sector surging 600% over four years, dwarfing the 100% gains expected from the hyperscalers actually writing the checks.

The capex tsunami

UBS forecasts AI-related capital expenditures hitting approximately $820 billion in 2026. That figure climbs to nearly $990 billion in 2027.

The hyperscalers, meaning Amazon, Microsoft, Google, and Meta, are the ones driving the spending bus. UBS estimates their combined capex will land around $602 billion in 2026, with approximately 75% of that earmarked specifically for AI infrastructure. Over 85% of total AI-related capex is expected to come from major technology companies.

NVIDIA, Microsoft’s Azure, Amazon’s AWS, AMD, and Arista Networks are among the names UBS identifies as key AI exposure plays.

The overheating problem

Certain AI infrastructure stocks have recorded short-term gains exceeding 500% within a single year, according to the bank’s analysts. The analysts emphasize that while growth expectations for companies like NVIDIA, Amazon, and Azure remain robust, the risk of unrealistic growth scenarios being priced into current valuations is real.

Where crypto fits into the picture

UBS is developing tokenization capabilities and planning to provide crypto trading access to select wealth management clients.

Investors watching this space should pay attention to actual capex deployment figures as they come in quarter by quarter. If hyperscaler spending meets or exceeds UBS’s $602 billion 2026 projection, the infrastructure thesis holds. If companies start trimming budgets or pushing timelines, the 600% value creation forecast becomes a ceiling rather than a floor.