By Stuart Kerr, Technology Correspondent, LiveAIWire
AI infrastructure spending strategies reached a genuinely new scale in 2026. Google, Amazon, Microsoft, and Meta collectively plan to spend 725 billion dollars on capital expenditure this year, according to first-quarter earnings compiled by the Financial Times, up 77 percent from 2025’s already record-breaking 410 billion dollars. Jefferies analyst Brent Thill summed up the market’s read on the numbers bluntly: the AI economy is healthy, and the bear thesis is garbage.
The four companies are taking meaningfully different paths to that combined figure, and those differences reveal as much about each company’s underlying strategy as the raw dollar amounts do. Alphabet raised its 2026 capex guidance to a range of 175 to 185 billion dollars, confirmed by CEO Sundar Pichai in the company’s own fourth-quarter earnings release, up from earlier forecasts as Google Cloud demand accelerated. Microsoft set its own 2026 figure at 190 billion dollars, well above the 152 billion dollar average analyst estimate, with CFO Amy Hood attributing 25 billion dollars of that increase directly to rising memory chip and component costs. Amazon is projecting roughly 200 billion dollars for the year, and Meta lifted its range to as high as 145 billion dollars, a 10 billion dollar increase driven by the same memory pricing pressure hitting Microsoft.
Why the Same Story Reads Differently for Each Company
Google’s spending increase came with the clearest evidence that it is converting into revenue. Google Cloud revenue grew 63 percent year over year to roughly 20 billion dollars in the first quarter, the fastest growth in the group, and Alphabet shares climbed after the print. Amazon Web Services grew 28 percent to 37.6 billion dollars, the fastest pace in fifteen quarters, giving Amazon’s spending increase a similarly strong revenue justification. Meta’s position was different. Shares dropped roughly 6 percent after the company raised its capex guidance, with SLC Management’s Dec Mullarkey telling the Financial Times that investors are increasingly uneasy about whether Zuckerberg’s historically capital-light business is morphing into something far more capital-intensive without matching proof of return.
What This Spending Actually Means for You
For anyone outside the industry, the practical stakes of this spending show up in two places: energy and jobs. Data centre electricity demand is on course to more than double by 2030, and communities near new data centre construction are already seeing real pressure on local grids and water systems as a direct consequence of exactly this capital spending. On the employment side, the picture is more mixed than the headline numbers suggest. Meta has cut roughly 8,000 jobs even while raising its AI capital spending, illustrating that the money flowing into infrastructure and the money flowing into headcount are not moving in the same direction inside every one of these companies.
The Custom Silicon Race Nobody Outside the Industry Notices
A meaningful share of this capital is going toward proprietary chips rather than simply buying more Nvidia GPUs. Google Cloud’s Thomas Kurian has specifically credited the company’s strategy of building its own AI chips, foundation models, and products in-house for giving Alphabet a cost and research advantage its rivals lack. Memory itself has become a genuine bottleneck across the entire industry. Analysts now estimate memory will consume roughly 30 percent of hyperscaler data centre spending in 2026, a fourfold increase from 2023, which is the specific factor both Microsoft and Meta cited when explaining why their capex guidance rose mid-year rather than holding steady.
Where the Next Frontier of This Spending Is Heading
The scale of terrestrial spending is also pushing some of this capital toward genuinely unconventional territory. Google has outlined Project Suncatcher, a plan to fly solar-powered satellites carrying its own AI chips into orbit, with prototype launches targeted for 2027, part of a broader industry response to the reality that ground-based grids cannot absorb this build-out fast enough on their own. Whether space-based compute becomes a meaningful share of the 725 billion dollar figure within the next several years, or remains a niche experiment, the fact that serious money is now being allocated to test it signals how genuinely constrained the terrestrial build-out has become.
Alphabet’s cloud contract backlog reaching 460 billion dollars, roughly double where it stood at the end of 2025, is probably the single clearest evidence in this entire earnings cycle that the demand side is scaling alongside the spending side rather than lagging behind it. Whether that holds for the rest of 2026, and whether Meta’s more capital-intensive path eventually shows the same revenue proof Google and Amazon have already demonstrated, is the question that will determine whether this 725 billion dollar figure looks like visionary infrastructure investment in hindsight, or the moment the AI capital cycle finally outran what the market was willing to fund without seeing the returns first.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.