The four largest hyperscalers have now poured more than $1.1 trillion into AI infrastructure since the start of 2023, and the pace is accelerating rather than plateauing. According to the Financial Times, which aggregated the companies’ latest earnings reports, Amazon, Google, Meta, and Microsoft are expected to add roughly $745 billion to that cumulative figure in 2026 alone, a single-year total that approaches what the group spent in the three preceding years combined.
Company-level guidance filed in the most recent earnings season shows the scale: Amazon raised its full-year capex forecast to about $220 billion, Microsoft is guiding toward $190 billion, Alphabet widened its range to $175-205 billion, and Meta has allocated $115-135 billion, with most of the money flowing into data centers, accelerators, and the power infrastructure required to run them.
Wall Street analysts see no near-term relief. RBC Capital’s Rishi Jaluria said after the earnings reports that investors now expect these companies to walk a tight line between funding AI capacity and protecting the profit engines that made them cash-rich enough to attempt this in the first place. The tension is visible in market reactions: record earnings have repeatedly been met with selloffs whenever spending guidance outpaces revenue visibility.
The cost of the buildout also extends beyond the four hyperscalers. U.S. utility companies have been forced into billions of dollars of grid upgrades to serve data center campuses, and those costs are flowing through to residential ratepayers. The White House’s Ratepayer Protection Pledge, signed by hyperscalers, utilities, and state officials, has so far not been codified into law by any state. Oregon’s POWER Act, enacted in 2025, imposes a 30 percent surcharge on power bills for users drawing more than 20 megawatts while cutting residential electricity costs by 1.3 percent.
On the hardware side, the hyperscalers’ willingness to pay premium prices for high-bandwidth memory has pushed Micron, Samsung, and SK hynix to prioritize HBM production over conventional DRAM, deepening a consumer memory shortage that began last year. The ripple effects now extend into automobiles, where GM has warned of vast cost increases and BYD raised driver-assistance package prices by 20 percent, and into budget smartphones, where memory can account for up to 64 percent of total bill-of-materials cost. Apple, despite its historic leverage over suppliers, raised Mac and iPad prices last month, with the entry-level MacBook Pro climbing $400 to $1,999.
The central question is whether AI revenue can eventually catch the capital being deployed. Bulls point to platform-shift precedents like cloud computing, which took years to monetize; skeptics note that none of the four hyperscalers has yet demonstrated positive returns on its AI infrastructure at scale. What is no longer in dispute is the scale itself: at current guidance, the trillion-dollar mark is not a milestone but a baseline.
Sources: Big tech spends more than $1 trillion on AI infrastructure (Tom’s Hardware, July 31, 2026); Financial Times coverage of hyperscaler capex (Financial Times); Big Tech Capex Surge: Amazon, Microsoft, Alphabet, Meta to Spend $750B in 2026 (Gate News, July 31, 2026); Tech AI spending approaches $700 billion in 2026 (CNBC)