‘AI Still in the Early Innings’: Big Tech Accelerates Spending Plans

U.S. technology giants are not easing up on artificial intelligence spending. If anything, they are stepping it up. This earnings season, faster cloud growth and rising AI demand led the biggest hyperscalers to increase their capital-expenditure plans for the year, underscoring how quickly Big Tech is ramping up investment to defend its position in the AI era.

According to industry data as of Aug. 11, the four largest hyperscalers — Amazon, Google, Meta and Microsoft — raised their combined 2026 capital-spending plans to $720 billion to $745 billion from a previous $695 billion to $725 billion. Amazon lifted its full-year capex forecast to $220 billion from $200 billion, while Google increased its outlook to $195 billion to $205 billion from $180 billion to $190 billion.

Meta also raised the low end of its forecast to $130 billion from $125 billion, while maintaining the upper end at $145 billion. Microsoft’s figure declined to $175 billion from $190 billion, but that reflected some data-center lease contracts being classified as operating leases rather than capex. Its actual level of investment remains unchanged.

Big Tech Cloud Revenue Growth Accelerates

This earnings season, U.S. tech companies made clear that AI demand is rising fast. Cloud growth accelerated at Google, Microsoft and Amazon. What began as a competition over AI models is rapidly shifting toward enterprise AI, AI agents, and the cloud and data centers needed to run them.

Google was the first to show tangible results. Google Cloud reported second-quarter revenue of $24.8 billion, up 82% from a year earlier. Remaining performance obligations grew to $51.4 billion. The company attributed the growth to demand for AI infrastructure and enterprise AI. About 90% of Fortune 100 companies are using enterprise Gemini, Google said. Data processed through Google’s proprietary model APIs climbed to about 22 billion tokens a minute from 16 billion in the previous quarter.

Chief Financial Officer Anat Ashkenazi said computing supply remains constrained as AI demand continues to rise.

Microsoft showed a similar pattern. Azure, its core cloud platform, posted 43% quarterly revenue growth. On a fiscal-year basis, annual revenue surpassed $100 billion for the first time. After spending last year testing generative AI, companies are now moving to directly integrate AI into workplace software, data and cloud systems.

Amazon also kept pace. Second-quarter revenue at Amazon Web Services rose 37%, its fastest growth in 18 quarters. Meta reported second-quarter revenue of $60.8 billion, up 28%, while advertising revenue increased 27% to $59.4 billion. The performance suggests AI improved content recommendations and ad efficiency, boosting profitability in its core business.

Earnings Back Up AI Demand

Before this earnings season, some questioned whether AI spending was sustainable. Big Tech’s latest results answered those concerns with stronger numbers. Alphabet, Meta, Amazon and Microsoft all posted double-digit year-on-year revenue growth of 24%, 28%, 20% and 18%, respectively.

Growth also accelerated across cloud businesses, which are directly tied to AI demand. “AI is still in the very early innings across multiple areas,” Alphabet Chief Executive Officer Sundar Pichai said on the earnings call. “If anything, I’m more optimistic about the opportunity than I was a year ago.”

Developing and running large language models requires more graphics processing units and more data centers. Infrastructure is also needed to improve advertising efficiency in businesses such as YouTube and search through AI.

Google said it would use neocloud providers — companies that rent out graphics processing units — to address shortfalls until its own infrastructure buildout is complete. Other Big Tech companies are facing the same issue.

“Big Tech used to be an asset-light platform business where revenue grew much faster than investment,” said Shay Boloor, chief market strategist at Futurum Equities. “Now it is becoming a hybrid model that depends on spending on data centers and AI infrastructure.”

Park Han-shin, Hankyung.com reporter, phs@hankyung.com