Big Tech keeps pouring money into AI data centers as Wall Street and voters push back

Amazon, Microsoft, Alphabet and Meta are still pouring vast sums into AI infrastructure even as investors question the returns and political pressure grows around new data centers. Together, the four companies plan at least $720 billion in capital spending this year, mostly for AI buildouts, while also sitting on more than $1 trillion in already arranged future commitments, according to The Verge.

The result is an uneasy standoff: the biggest names in tech are accelerating a spending race they believe they cannot afford to lose, but markets are punishing them for it and regulators may slow the next wave of construction.

What is driving the AI spending surge?

The short answer is that the hyperscalers still see AI infrastructure as the foundation of their next phase of growth. Data centers, chips and networking gear are the physical backbone of large models, and the companies leading the AI race are building aggressively to keep up with demand for training and running those systems.

That spending is no longer theoretical. The Verge reported that Amazon, Microsoft, Alphabet and Meta are on track to spend at least $720 billion combined this year on capital expenditures, a category that includes long-lived assets such as data centers and the hardware inside them. The companies also have at least $1 trillion in spending commitments already lined up for the years ahead.

Those numbers underscore how quickly AI has become a company-defining investment cycle. For investors, the problem is not whether the buildout is real; it is whether the returns will arrive quickly enough to justify the scale.

Why it matters

This matters because the AI arms race is no longer just a story about model performance or product launches. It is increasingly a story about balance sheets, debt markets, power grids, local permitting and political backlash.

If the spending pays off, the companies building the most infrastructure could lock in a durable advantage in cloud services, search, advertising and enterprise AI. If it does not, the industry could be left with enormous sunk costs and less room for flexibility just as competition intensifies.

For readers, the issue reaches beyond Silicon Valley. Data centers affect electricity demand, land use and local development decisions, while the financing behind them can ripple through credit markets. The Verge’s reporting suggests the stakes are broad enough that Wall Street is no longer granting the AI boom a free pass.

How did investors respond?

They responded sharply. Last week’s tech selloff reflected growing skepticism that all of this spending will translate into profits fast enough. Alphabet took the brunt of it, falling 13% in a single day on Thursday after earnings, its worst daily drop on record.

The market reaction was not just about raw spending levels. It was also about evidence that the money is not yet producing clean monetization. Alphabet said AI Mode query growth reduced monetization in Google Search, while Google Search ads revenue growth slowed to 12% in the third quarter from 17% a year earlier.

Alphabet also told investors it plans $195 billion to $205 billion in capital expenditures this year, with a significant increase expected in 2027. That outlook helped fuel doubts that the company’s AI strategy will boost near-term returns without pressuring the core advertising business first.

Alphabet’s profitability challenge

The Verge’s figures point to the central tension facing Alphabet: AI products are changing user behavior, but not always in ways that make money immediately. Philipp Schindler, Alphabet’s senior vice president and chief business officer, said that “strong growth in AI Mode queries had an impact,” a comment that captures the company’s struggle to balance usage growth with ad monetization.

That problem is not unique to Alphabet, but it is especially visible there because Search remains the company’s core profit engine. If AI features pull traffic or searches into formats that generate less revenue per query, the company may have to spend more just to defend its existing business.

What are the biggest companies spending?

The scale differs by company, but the direction is the same: more infrastructure, more capacity and more money committed far into the future.

Amazon is the most aggressive of the four in absolute annual spending, at $220 billion this year, according to The Verge. CEO Andy Jassy has signaled that the company sees the opportunity as “once in a lifetime opportunity” and expects spending to be “meaningfully more” in future periods.

Meta is also spending heavily, with expected capital expenditures of $135 billion to $145 billion this year. But Meta has been unusually explicit about the performance hurdle: the company says compute efficiency needs to rise tenfold next year if agentic AI is going to be profitable.

Microsoft’s spending story is a little different. It reported $45.2 billion in quarterly CapEx, which was below expectations because third-party capacity delays constrained its pace of buildout. Shortly after, Microsoft disclosed a $25 billion, 10-part bond sale to help fund its AI expansion.

How is Microsoft financing the buildout?

Microsoft is using the debt markets as one route to keep AI spending moving. The company’s $25 billion bond sale, disclosed after its earnings call, suggests that even a cash-rich hyperscaler is willing to tap public debt to support the scale of the infrastructure race.

That matters because Microsoft’s financing decision may become a model for others. The company had told investors during its Wednesday night earnings call that debt markets were available, but not yet tapped. By the next morning, the bond sale was public.

The move also highlights a broader shift in the AI boom: spending is becoming so large that operating cash flow alone may not be the only answer. Companies are already arranging future commitments and exploring financing structures that can support multi-year construction cycles.

What does Anthropic’s update say about the broader AI economy?

Anthropic’s latest S-1 update offers a rare window into how expensive frontier AI can be. The company reported third-quarter revenue of $17 billion and compute costs of $17.5 billion, a strikingly close match that underscores how much infrastructure can swallow even rapidly rising sales.

The company said revenue has increased by more than 1,000% versus last year, yet it still has nearly $900 billion in future compute and infrastructure commitments. It also plans to start building its own data centers, another sign that dependence on outside capacity may no longer be enough for major AI developers.

Anthropic’s figures are especially important because they show that even explosive growth does not automatically solve the economics of AI. The revenue line can rise dramatically while the cost base rises too, leaving the industry still searching for a durable profit model.

Background: how we got here

The current spending wave did not appear overnight. Over the past several years, hyperscalers have been racing to secure compute capacity, chips and power for a new generation of AI systems. The companies have treated infrastructure as a strategic moat: whoever controls the most compute can train, deploy and iterate models faster.

At the same time, investors have been asking a more basic question: when does the money come back? That tension has grown as capital spending rose, AI features began to reshape core products and the cost of scaling large models stayed stubbornly high.

Regulation has entered the picture too. Data-center permitting and local opposition have become political issues in some states and municipalities, where concerns over land, water, power consumption and community impact have made large projects harder to move forward.

That broader backdrop helps explain why the latest earnings season felt different. The market was not just reacting to quarterly figures; it was reacting to the possibility that AI spending may need to keep rising long after the first wave of enthusiasm.

Why Wall Street is getting impatient

The selloff last week showed that investors are less willing to reward spending for its own sake. In previous phases of the AI boom, large budgets could be framed as strategic necessity. Now, the burden is shifting toward proof of monetization.

Alphabet is the clearest example. Its stock drop after earnings, combined with slower ad growth and reduced monetization from AI Mode, gave traders a concrete reason to question whether the company’s AI strategy is strengthening or diluting its near-term economics.

The same scrutiny applies to the rest of the group. Amazon, Microsoft and Meta may have different business models, but each is trying to justify enormous AI-related spending before markets turn even more skeptical.

For now, the spending continues. That suggests the companies believe underinvesting would be more dangerous than overinvesting, even if the latter unsettles shareholders.

How politics is entering the data-center debate

The pressure is not just coming from Wall Street. The midterms brought in candidates favoring moratoriums and other data-center restrictions, adding a political layer to what had mostly been a corporate and financial story.

That development matters because data-center projects are local as much as they are national. They require land, power, permits and community approval. If elected officials decide to slow or block construction, the AI buildout could face delays that have little to do with demand and everything to do with local constraints.

Those restrictions would not necessarily stop the AI race, but they could reshape it. Companies might need to spread investments across more jurisdictions, adjust timelines or spend more to secure sites and energy access.

By the numbers

The scale of the AI infrastructure push becomes clearer when the key figures are laid out side by side.

What happens next?

The next phase will hinge on whether the companies can show that AI spending is turning into durable revenue rather than just larger depreciation bills and heavier balance sheets.

Several questions remain open:

• Whether voters and state or local officials will materially restrict new data-center projects.
• How quickly Big Tech can convert AI usage into profits.
• Whether Alphabet can monetize AI Mode without further hurting Search.
• How much each company will ultimately spend in 2027 and beyond.
• Whether Microsoft’s bond sale becomes a template for other hyperscalers.

Alphabet has already signaled that spending will rise significantly in 2027. Amazon’s Andy Jassy has also indicated more investment ahead. Meta has not disclosed next year’s spending, but it has made clear that efficiency gains will need to improve sharply for agentic AI to pay off. Microsoft, meanwhile, has shown it is willing to use debt to keep the buildout moving even when internal capacity is constrained.

The broader industry message is plain: the AI infrastructure race is still on, but the financial and political costs are becoming harder to ignore. The companies are spending like winners-in-waiting; the market is asking them to prove it.

Source notes

This report is based on The Verge’s coverage of recent earnings calls, company filings and related market reaction.