The four hyperscalers — Microsoft, Google, Amazon and Meta — collectively raised their 2026 capital expenditure plans to roughly $710 billion, on top of $416 billion spent in 2025. 

Nearly every executive on their respective April 29 first-quarter earnings calls said the same thing: Demand is outpacing supply, and the cycle is accelerating. 

“We’re just in a stage where there’s just not enough capacity for the amount of demand,” Amazon Chief Executive Andy Jassy said on the earnings, adding that rising CapEx might not be enough to meet customer and internal demands. 

The four leaders of AI CapEx spent nearly $517.5 billion in 2023, 2024 and 2025 combined. 

The four-year arc in the chart above shows how fast the AI infrastructure build has scaled. In 2023, the four companies spent about $149 billion combined on property, equipment and finance leases. 

Last year, that figure nearly tripled to $416 billion. The midpoint of 2026 guidance points to a roughly 70% year-over-year increase to more than $700 billion.  

The forecasts 

Microsoft

“For calendar year 2026, we expect to invest roughly $190 billion in capital expenditures, which includes approximately $25 billion from the impact of higher component pricing,” Chief Financial Officer Amy Hood said during the company’s earnings call.  

The tech giant is increasing its capital expenditure by 61% YoY, Hood said, adding that, “even with these additional investments, and continued efforts to bring GPU, CPU and storage capacity online faster, we expect to remain constrained at least through 2026.” 

Alphabet 

Google raised its full-year 2026 CapEx guidance to $180 billion to $190 billion, according to its Q1 earnings report. The previous guidance, released on Feb. 4, was $175 billion to $185 billion for 2026. 

Q1 CapEx was $35.7 billion, more than double the $17.2 billion the company spent in Q1 2025, according to the earnings report. 

“We expect our 2027 CapEx to significantly increase compared to 2026,” Google CFO Anat Ashkenazi said during its earnings call. 

Google CEO Sundar Pichai was explicit about the supply problem. 

“We are compute-constrained in the near term,” Pichai said on the call. “Our cloud revenue would have been higher if we were able to meet the demand.” 

Google Cloud revenue grew 63% YoY to $20.03 billion in Q1, according to the company’s earnings report. 

Amazon 

Jassy reiterated the company’s $200 billion 2026 CapEx spend mentioned in February during the company’s Q4 earnings.  

“Of the AWS CapEx we intend to spend in 2026, much of which will be installed in future years, we have high confidence this will be monetized well, as we already have customer commitments for a substantial portion of it,” Jassy said. “As we have been sharing, the faster AWS grows, the more short-term CapEx we will spend.” 

In Q1, Amazon’s CapEx was $43.2 billion, up from $24.3 billion in Q1 2025, according to its earnings report. 

Meta 

Meta also raised its full-year guidance to $125 billion to $145 billion, up from $115 billion to $135 billion shared in February when the company reported Q4 earnings. 

“We are increasing our infrastructure CapEx forecast for this year,” Mark Elliot Zuckerberg, chief executive officer, said during the company’s earnings call. “[What] we are seeing in our own work and across the industry gives us confidence in this investment.” 

Demand is high, supply restricted 

Despite pouring half a trillion dollars in AI infrastructure and pledging to pour in an additional $700 billion this year, hyperscalers say they can’t keep up with demand. 

“We are seeing unprecedented internal and external demand for AI compute resources. … These strong results reinforce our conviction to invest the capital required to continue to capture the AI opportunity,” Google’s Ashkenazi said. 

Meta CFO Susan Li said that the company has been underestimating its compute needs even as it has added significant capacity. 

“We’re going to continue building out our infrastructure with flexibility in mind,” she said. 

A major supply issue is the lack of raw materials required to build chips and GPUs, a problem that has been exacerbated by the Iran war, according to a Moody’s Ratings report on April 22.  

“Hyperscalers are committing [more than] $650 billion to U.S. AI infrastructure this year alone,” David Pan, director and AI industry practice lead at Moody’s, told FinAi News. “And, that investment assumes the supply chain holding it together remains intact.” 

One ingredient in the AI supply chain that doesn’t get much attention is helium, Pan said. It’s essential for cooling wafers during chip etching, and there is no viable substitute at scale. 

However, helium isn’t manufactured. It accumulates over millions of years through radioactive decay and is captured only as a byproduct of natural gas processing, Pan said. 

“The AI economy runs on tokens, tokens run on GPUs and GPUs depend on Qatari helium, Israeli bromine and LNG tankers with a single, 21-mile-wide exit from the Persian Gulf,” Pan said. 

Amazon’s Jassy agreed, saying the trickle-down effect of supply chain restriction will not only delay projects but also push the price of AI infrastructure buildout higher. 

Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas.