Big Tech AI spending tops $700B

Yahoo Finance reported on June 4 that Big Tech’s AI spending is on track to top $700 billion this year as the largest cloud and internet companies keep raising capital budgets for data centers, chips and networking gear. The spending wave is widening the list of companies investors are watching beyond Nvidia and other semiconductor names to electric utilities, cooling suppliers and site operators that can help bring new capacity online. Moody’s and other analysts have said the build-out is being driven by demand for AI training and inference, even as the cost is cutting into free cash flow and pushing some companies toward more borrowing. ### Which companies are driving the $700 billion figure? Moody’s said in March that six U.S. hyperscalers — Microsoft, Amazon Web Services, Meta Platforms, Alphabet, Oracle and CoreWeave — are expected to spend about $700 billion in 2026. Moody’s said that would be nearly six times 2022 levels and about 81% above last year, with further growth possible in 2027. (finance.yahoo.com) CNBC reported in February that Alphabet, Microsoft, Meta and Amazon alone were projected to spend close to $700 billion combined this year. Bloomberg reported in April that the biggest U.S. tech firms could spend as much as $725 billion in 2026, primarily on AI data center equipment. Yahoo Finance’s related May 1 report detailed some of the quarterly pace behind the total. Alphabet spent $35.67 billion on capital expenditures in the first quarter, Amazon spent $44.2 billion, Microsoft spent $30.88 billion in its fiscal third quarter, and Meta raised its full-year capex guidance to $125 billion to $145 billion. (datacenterknowledge.com) ### Why are investors looking past chipmakers now? (cnbc.com) Yahoo Finance said the next trade tied to AI may be electric utilities, as hyperscalers pour record sums into data centers that need large and reliable power supplies. Moody’s said the spending surge is lifting demand across the data center supply chain, including semiconductors, IT hardware, power generation, construction and cooling equipment. (finance.yahoo.com) Data Center Dynamics, citing Moody’s, said the lack of readily available electricity and the time needed to build data centers are likely to constrain AI capacity through 2027. That has made grid access, power delivery and facility build-out more central to the investment case around AI infrastructure. ### What is forcing the focus onto power and facilities? (finance.yahoo.com) Moody’s said demand for computing capacity to train new AI models and support growing inference workloads exceeds current supply. The ratings firm said the industry remains capacity-constrained and that electricity availability is one of the main bottlenecks. (datacenterdynamics.com) Yahoo Finance’s June 4 report framed utilities as an “overlooked” way to play the boom because the spending is no longer limited to processors and servers inside the data center. New AI capacity also requires land, transmission, substations, cooling systems and physical construction that can take years to complete. ### Are investors fully comfortable with this pace of spending? (datacenterknowledge.com) CNBC reported that the spending surge is taking a visible toll on cash generation. The network said Amazon could turn free-cash-flow negative this year, while the four biggest U.S. internet companies already generated less combined free cash flow in 2025 than in 2024. Moody’s said the gap is widening between how hyperscalers view AI spending and how some investors view it. (finance.yahoo.com) The agency said the build-out is accelerating revenue growth, but it is also eroding historically strong free cash flow, increasing borrowing and raising the risk that creditworthiness could be reassessed if profit growth does not materialize. (cnbc.com) ### What comes next in the build-out? Moody’s said hyperscaler capital spending could rise to $820 billion in 2027, with potential for further upward revisions. The firm said electricity shortages and construction timelines mean capacity is likely to lag demand through 2027 even as companies stage projects and align some expansion with contracted demand. (datacenterknowledge.com) The next milestones are likely to come through company earnings, capex guidance updates and utility or data center project announcements from Microsoft, Amazon, Meta, Alphabet, Oracle and CoreWeave. Yahoo Finance’s June 4 report said investors are already shifting attention to the companies that can supply the power and physical infrastructure needed to keep that spending cycle moving. (finance.yahoo.com) (datacenterdynamics.com)