By Asa Fitch and Dan Gallagher
The biggest tech companies are spending a fortune on AI now. They, and their shareholders, will be paying for it for years to come.
Earnings reports from Microsoft, Amazon.com, Meta Platforms and Google parent Alphabet on Wednesday all showed the companies continuing on a path of blowout capital expenditures. A collective $133 billion went out the door during the first quarter, up about 70% from a year earlier. The four are now on track to spend a combined $725 billion this year, according to updated capital-spending projections issued with Wednesday's reports.
The sharp spending growth is driving another important number higher as well. Depreciation charges surged at all four companies, totaling $41.6 billion for the most recent quarter.
When companies make capital investments, they don't count the outlays immediately as expenses. Rather, these capital assets have to be depreciated over a period of time. So the impact on profits is delayed. But a multitrillion-dollar bill will have to wash through in coming years, taking a bite out of reported profits.
And the impact of the past few years' spending is already growing fast -- all the more so since the lifespan of AI chips and servers is relatively short compared with other types of assets. Microsoft, Amazon, Meta and Google all depreciate their server equipment over five to six years. This means the record amount of capex being spent now will be charged against their earnings over the next several years.
Anat Ashkenazi, Alphabet's finance chief, acknowledged on a call with analysts Wednesday that depreciation would put pressure on profits. She said at an investor conference in March that even with Google's scale, it was "not a small number to offset."
That creates new pressure on AI's big spenders to show payoffs from AI investment soon. If they succeed, the big outlays will look smart and depreciation won't be much of a concern.
But it is unclear whether AI can generate returns commensurate with its costs -- a worrisome issue for investors.
Annual depreciation charges for these four companies are currently expected to exceed $430 billion within the next five years, according to consensus estimates from Visible Alpha. These noncash charges will increasingly eat into reported profits if AI services fail to boost earnings at a similar rate. The four reported combined net income of $372 billion last year.
And there is little the companies can do about the growing charges. They have already pulled the one trick they did have: extending their assessments of the useful life of AI servers and related gear to spread out the depreciation waterfall over a longer period.
That shift came in the first couple of years of the AI boom, moving from an earlier cycle of about four years. It would likely be hard for them to justify pushing out further, although Meta is extending the useful life of non-AI servers to seven years.
So the coming depreciation wave is largely unavoidable. The good news for the tech giants is that there is little sign their businesses on a whole are slowing. All of the largest tech companies posted double-digit revenue growth in the first quarter, and analysts' forecasts suggest more upside to come.
Things like cloud computing, software sales and advertising can cover the depreciation charges for the foreseeable future. They can afford it, for now.
That might be a fine trade-off for many investors. In that setup, tech companies effectively give away AI, turning it into a loss-making side hustle with a chance to turn a profit eventually.
But there are also reasons for investor unhappiness with such a structure. It could sap companies of a financial buffer if their core businesses struggle while they are still waiting for the AI bonanza to materialize.
That uncertainty is why markets have sometimes punished tech companies for raising their capital spending -- including recently with Tesla. Its shares fell last week after the company raised spending plans this year to $25 billion.
The depreciation problem is sure to be more acute for some tech companies than others as their paths to monetizing AI diverge.
Google has emerged as a leader in converting AI into revenue. The company said Wednesday that its cloud-computing unit's sales rose 63% in the first quarter, driven by AI services, and that its AI models processed 16 billion tokens -- the computing currency needed to use AI systems -- every minute. That is up 60% from the previous quarter, painting the outlines of a significant and fast-growing business.
Search and advertising are booming too. Alphabet's stock jumped around 7% in after-hours trading Wednesday, following a gain of around 25% over the past six months.
Meta, meanwhile, is making hay with AI in advertising, boosting revenue there. But its struggles in advancing cutting-edge models over the past few years suggest a dimmer long-term outlook for its AI efforts to drive a broader revenue expansion.
The company in its report Wednesday raised its capital-spending target for this year by $10 billion to around $135 billion. The rising spending and a downbeat revenue forecast for the second quarter disappointed investors; shares fell around 7% in after-hours trading.
Microsoft showed a small improvement in growth for its Azure cloud business. It also cited strong AI demand as a driving force for its plan to spend $190 billion in capex this calendar year -- about 23% more than what Wall Street was expecting.
That includes $25 billion to cover "higher component pricing," indicating that Microsoft's future earnings are also being eaten up by skyrocketing memory prices.
Most tech executives have approached the AI boom with calculated irresponsibility. They know the current returns can't justify their spending, but their faith in a future in which AI drives the global economy means they won't rein it in.
The executives are, in a way, the corporate equivalents of graduate students running up credit-card debt, certain their lucrative careers will pay it off. They just better not drop out -- or else end up working at Starbucks.
Write to Asa Fitch at asa.fitch@wsj.com and Dan Gallagher at dan.gallagher@wsj.com
(END) Dow Jones Newswires
April 30, 2026 05:30 ET (09:30 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments