A new analysis from BofA Global Research concludes that heavy investment tied to artificial intelligence could lift total global hyperscale capital expenditure to $1 trillion in calendar year 2027.
The bank's revised forecasts follow first-quarter earnings from several large cloud operators and reflect two concurrent trends cited by analysts: accelerating AI-driven revenue at the hyperscalers and continued tightness in supply for high-end compute hardware through 2026.
Projected capex trajectory
BofA now expects combined hyperscale capex to top $800 billion in 2026 - a 67% increase year-over-year - before moving beyond the $1 trillion mark in 2027. The upgrade in the outlook is linked to the companies' reporting of stronger-than-anticipated AI demand and to upward revisions in their own spending guidance for 2026.
Company-level signals
Public disclosures from several major U.S. technology firms underpin the projection. Alphabet is reported to be generating more than 16 billion Gemini tokens per minute and has seen its search business expand by 19% as AI-infused queries gain traction. Microsoft has disclosed an annualized AI sales run-rate above $37 billion, a 123% increase year-over-year. Amazon Web Services recorded its fastest growth in over three years at 28%, a pace the report ties largely to AI workloads and strategic partnerships.
Higher guidance and component cost impact
To meet surging demand, the hyperscalers have lifted their 2026 capex outlooks. Microsoft raised its 2026 capex projection to $190 billion, up from what had been a market expectation of $154 billion. Amazon is maintaining robust guidance at $200 billion. Alphabet and Meta increased their 2026 spend projections to $185 billion and $135 billion, respectively.
The report emphasizes that a notable portion of the increase reflects rising hardware costs that are being absorbed by the cloud operators. Microsoft, for example, attributes $25 billion of its additional 2026 capex to higher component pricing.
Market dynamics and beneficiaries
BofA notes that semiconductor vendors currently have significant pricing power, enabling them to pass through higher costs for wafers, memory, and substrates. The sustained spending cycle is expected to benefit companies that supply compute hardware and related components - explicitly naming compute chipmakers, memory suppliers, optics vendors, semicaps, and power semiconductor manufacturers.
The research also points out that hyperscalers are balancing purchases of merchant GPUs with deployments of custom silicon. Supply for AI compute is expected to remain constrained through 2026, and the combination of strong customer commitments and improving free cash flow is cited as justification for maintaining historically high investment levels.
Summary
BofA Global Research projects hyperscale capex will exceed $800 billion in 2026 and reach $1 trillion in calendar year 2027, driven by accelerating AI revenues, raised capex guidance from major cloud providers, and ongoing supply constraints for high-end compute hardware.
Key points
- Hyperscale capex is forecast to top $800 billion in 2026, a 67% year-over-year increase, before surpassing $1 trillion in 2027.
- Alphabet, Microsoft, Amazon, and Meta have all reported strong AI-related traction and raised 2026 capex guidance, with Microsoft citing $25 billion of its increase due to higher component prices.
- Semiconductor vendors and suppliers of memory, optics, semicaps, and power semiconductors are identified as key beneficiaries of the investment cycle.
Risks and uncertainties
- Supply constraints for high-end AI compute are expected to persist through 2026, creating uncertainty around timing and delivery of hardware necessary for planned expansions.
- Rising hardware component prices are increasing capex totals - for example, Microsoft attributes $25 billion of its 2026 increase to higher component costs - which could affect margins for hyperscalers and their customers.
- Dependence on continued strong AI revenue growth to justify elevated capex levels represents an underlying risk if demand trajectories change.