UBS says AI infrastructure companies have overtaken major technology hyperscalers in value creation as profits increasingly shift toward chipmakers and hardware suppliers.

Research from the bank’s Holt unit found that forecasts for cash flow return on investment have risen sharply across semiconductor and technology hardware companies, excluding Apple. Returns among hyperscalers including Amazon, Alphabet and Meta have moved in the opposite direction as spending on AI infrastructure accelerates.

UBS expects economic profit generated by AI infrastructure companies to rise from about $200 billion in 2023 to $1.4 trillion in 2027. Economic profit from hyperscalers is projected to double over the same period to roughly $400 billion.

Memory chipmakers Samsung, SK Hynix and Micron are expected to account for a significant portion of the increase after recovering from losses recorded during the previous semiconductor downturn. Demand for high bandwidth memory used in AI accelerators has tightened supply and supported higher prices and margins across the sector.

The shift is changing the companies that generate the largest share of value across the technology sector. Nvidia, Samsung, SK Hynix and Micron are emerging as leading contributors, replacing a structure previously dominated by companies including Apple and Microsoft.

The change comes as hyperscalers commit a growing share of their cash flow to data centers, chips and other AI infrastructure. UBS estimated that combined hyperscaler capital expenditure could approach $770 billion in 2026, while the companies are on track to direct nearly all of their free cash flow toward investment.