How much are hyperscalers spending on AI?
The scale of capital being poured into artificial intelligence by hyperscalers is entirely unprecedented and triggering an intensive infrastructure sprint that is reshaping the global economy. In 2026, total capital expenditure from the top five hyperscalers (Amazon, Alphabet, Microsoft, Meta, and Oracle) is projected to exceed $750 billion.
CreditSights estimates that roughly 75% of this sum, or approximately $450 billion, is being funneled directly into dedicated AI infrastructure, including next-generation data centers, specialized fiber networks, and bleeding-edge AI accelerators like Nvidia’s (NVDA +2.30%) Blackwell systems.
According to an analysis by S&P Global Ratings (SPGI -0.21%), this breakneck investment cycle is expected to peak over the next 12 to 18 months, with aggregate hyperscaler capex projected to cross $1.3 trillion by 2027. Financial institutions are divided over the extreme risk profile of this build-out.
While some bears warn that failing to monetize enterprise AI software immediately could result in the largest misallocation of capital in corporate history, the hyperscalers are operating under a "race with no finish line" mentality. The leaders of these companies are deeply convinced that underinvesting in AI infrastructure poses a far greater existential threat to their long-term market dominance than overinvesting.
