admin Microsoft, Amazon, Alphabet and Meta remain at the centre of the spending wave, with outlays directed toward data centres, advanced chips and cloud infrastructure. S&P Global Visible Alpha’s Melissa Otto said persistently elevated oil prices could force companies to revisit capital expenditure plans in the first and second quarters, a move that could act as a trigger for a wider correction in equity markets if earnings fail to absorb the added strain. The scale of the build-out is striking: the projected 2026 total is up from $383 billion in 2025 and just $80 billion in 2019, underscoring how quickly AI has moved from growth story to core corporate strategy.
That spending surge is no longer being judged only on demand for AI services. Investors are increasingly focused on whether enough electricity can be delivered, at a viable cost, to keep the expansion on track. Reuters reported from CERAWeek in Houston that Google president and chief investment officer Ruth Porat warned the United States may not be scaling energy supply quickly enough to meet the needs of AI data centres. She said the country would need to embrace every available source of energy, reflecting a broader industry shift from talking about compute and chips to talking about power, transmission and interconnection queues.
Grid stress is already shaping how the industry operates. Reuters reported on March 26 that utilities and regulators are pushing technology groups to make data centres more flexible during peak-demand periods, a step that would have been hard to imagine when hyperscale facilities were designed to run continuously. The same report cited a study by the Electric Power Research Institute showing electricity use from data centres could more than quadruple by the end of the decade and consume as much as 17% of U. S. power supplies. That prospect has raised pressure on policymakers to balance industrial expansion against household power bills and grid reliability.
The financing backdrop is adding another layer of risk. A Reuters market analysis this week said Big Tech’s AI-related capital expenditure could exceed $630 billion this year and rise above $800 billion next year, while cumulative spending across 2026 and 2027 may approach $1.4 trillion. Analysts cited in that report said hyperscalers are burning through cash faster, with debt issuance expected to rise materially as borrowing costs climb. Even companies with formidable balance sheets now face the possibility of a double squeeze: higher rates on one side and rising energy-linked operating costs on the other.
Physical infrastructure constraints are also becoming harder to ignore. Reuters, citing the International Energy Agency, reported in February that more than 2,500 gigawatts of projects worldwide remain stuck in grid-connection queues, putting about a fifth of global data-centre build-out at risk of delays. The IEA said meeting electricity demand through 2030 would require annual grid investment to rise by 50% from the current $400 billion. That has given fresh weight to the argument that the AI race may be limited less by software breakthroughs than by transformers, transmission lines, substations and gas turbines.
Companies are responding by moving deeper into energy planning. Alphabet has invested in advanced nuclear and struck demand-response arrangements with utilities, while Meta has backed large-scale power development in Louisiana tied to its data-centre ambitions. Reuters reported last week that Entergy’s revised agreement with Meta would support new energy infrastructure, including seven gas-fired plants with more than 5,200 megawatts of capacity, transmission lines, battery storage and nuclear upgrades, with Meta covering the full cost of service. The arrangement illustrates how tech groups are increasingly willing to finance energy assets directly when the grid cannot keep pace.
