Four of the largest U.S. technology companies expect their combined capital expenditures to reach approximately $650 billion by 2026, in a massive spending wave aimed at building new data centers and acquiring the long list of equipment required to operate them — including artificial intelligence chips, network cables, and backup power generators.

The spending planned by Alphabet, Amazon, Meta, and Microsoft, as part of the race to dominate the still-nascent market for AI tools, represents an unprecedented investment boom in this century.

According to Bloomberg data, each company’s projected spending for the current year is on track to mark the highest level of capital expenditures ever recorded by a single company in any year over the past decade.

A Historic AI Spending Boom

To find a historical comparison for the scale of these ambitious spending projections — announced alongside the four companies’ earnings reports over the past two weeks — one would have to look back at least to the telecommunications bubble of the 1990s. Possibly even further, to the construction of U.S. railroad networks in the 19th century, the massive federal investments in the interstate highway system after World War II, or the stimulus programs of the New Deal era.

The soaring figures, which collectively point to an increase of roughly 60% compared with last year, reflect a renewed acceleration in the global build-out of data centers.

The race against time to construct these massive facilities — housing rows of constantly operating servers powered by high-cost processors — has placed additional strain on energy supplies, raised concerns about higher prices for other users, and led to growing friction between developers and local communities worried about competition for electricity and water.

This expansion also heightens the risk of distorting macroeconomic indicators, as construction spending becomes increasingly concentrated among a small number of wealthy companies that already account for a growing share of U.S. economic activity.

Gil Luria, an analyst at DA Davidson, said the four companies view the race to provide AI computing capacity as “the next market where a single winner could dominate or capture the largest share,” adding that “none of them wants to end up on the losing side.”

Pressure on Tech Stocks

Meta announced last week that its annual capital expenditures could rise to as much as $135 billion, reflecting a potential increase of around 87%.

On the same day, Microsoft disclosed that its capital spending in the second quarter jumped 66%, exceeding expectations. Analysts estimate its total spending could reach $105 billion for the fiscal year ending in June. The news triggered a broad sell-off, resulting in the second-largest single-day drop in market value for any publicly listed company.

Alphabet, founded in a garage in Southern San Francisco in 1998, unsettled investors on Wednesday after announcing capital expenditure projections that not only surpassed analysts’ estimates but also exceeded the spending levels of a wide swath of U.S. companies, with plans to invest up to $185 billion. The following day, Amazon raised the bar by unveiling plans to spend $200 billion in 2026, pushing its stock lower in extended trading.

By contrast, the combined capital expenditures of the largest U.S. automakers, construction equipment manufacturers, rail companies, defense contractors, wireless telecom firms, parcel delivery companies — along with Exxon Mobil, Intel, Walmart, and the 21 companies spun off from General Electric — are expected to total just $180 billion in 2026, according to Bloomberg estimates.

The Expanding Physical Footprint of Tech Giants

While each tech giant has charted a somewhat different path toward earning returns on its investments, their spending rests on a shared assumption: that generative AI tools — such as OpenAI’s ChatGPT and its competitors, capable of generating text and displaying human-like reasoning abilities — will play an increasingly important role in people’s work and home lives.

Developing the advanced software models that enable this shift, however, is extremely costly. Running them requires linking thousands of chips, each selling for tens of thousands of dollars, helping explain the ballooning expenditure. The expansion also hinges on a bet that the resulting products will drive steadily rising revenues in the future.

These investments have also reshaped companies that until recently had little physical presence, despite their digital services reaching billions of users. For much of their history, Meta and Alphabet — Google’s parent company — viewed their sleek headquarters and office space as the most significant part of their tangible assets, while most spending went toward salaries and stock grants for engineers and sales teams.

That chapter has now closed. Last year, Meta spent more on capital projects than on research and development — which largely consists of engineers’ salaries — for the first time in six years. By the end of last year, the value of property and equipment at the owner of Facebook and Instagram had reached about $176 billion, nearly five times its level at the end of 2019.

Funding and Resource Bottlenecks Threaten Ambitions

As the numbers continue to swell, questions remain about these companies’ ability to execute their vast ambitions. Since the pace of data-center construction accelerated, competition has already intensified for scarce resources, including teams of electricians, cement trucks, and Nvidia chips produced by Taiwan Semiconductor Manufacturing Company (TSMC). Luria noted that “bottlenecks already exist today and will persist.”

Questions are also being raised about how this expansion will be financed. Meta and Google, which derive most of their profits from digital advertising; Amazon, the world’s largest e-commerce and cloud computing company; and Microsoft, the largest provider of business software, all hold dominant positions in their sectors and possess large cash reserves. Yet their willingness to channel vast portions of that liquidity into an AI-driven future means those reserves — and investors’ patience — will be put to a real test.

Tomasz Tunguz, an investor at Theory Ventures and a former Google employee, said these companies were once seen as “cash-generating machines.” He added: “Now, suddenly, they need that cash — and more of it — which is why they are turning to borrowing.”

Tunguz, who published a blog post last year comparing the AI boom to previous investment manias, noted that such surges do not always end well. Still, he stressed that they “serve as powerful catalysts for economic activity” during their ascent.

Rising Spending Unsettles Investors

It is increasingly clear that investors who rushed to buy tech-giant shares over the past year have grown more cautious in the face of the sharp rise in capital expenditures.

In some cases, they have moved to sell despite the stability of companies’ core businesses — from digital advertising and search engines to e-commerce and productivity software — and despite revenues exceeding expectations.

Steve Lucas, CEO of Boomi, a company specializing in enterprise data and software integration, said: “What worries investors? It’s clearly the narrative — and analysts’ messaging — around the speed at which AI will fundamentally transform business models.”

He added: “I don’t doubt the potential of AI. But I certainly doubt the timeline — and its economic viability.”