AI spending is surging, but a hidden risk is getting overlooked

The main question on Wall Street these days is whether the bull run will continue in 2026.

For 2023, 2024, and 2025, the S&P 500’s calendar-year returns stand at 24%, 23%, and 16%. Analysts are unanimously predicting a solid year in the coming years.

However, there's a catch.

Bloomberg's wide-ranging look at the top predictions for 2026 puts AI at the center of the market, with companies saying that the greatest danger is not being part of the AI scene. But the focus seems to be software, while the main, underlying issue is infrastructure.

And that helps explain why the data center ecosystem is becoming a deal-making machine.

Data centers are becoming the “new oil patch” for AI capital

Bloomberg said AI has led to about $70 billion in data center M&A talks in 2025. This isn't a random number; it's a sign that buyers think size is important right now.

A headline example: SoftBank agreed to buy DigitalBridge for $16 per share in cash. The deal is expected to close in the second half of 2026. The company also said the offer was 15% higher than the previous close and 49% higher than the 30-day VWAP.

DigitalBridge CEO Marc Ganzi called it an AI-driven demand moment and, more crucially, a chance to play offensive with the help of the balance sheet.

The money is real, and the capital expenditure numbers are clear

If you want "market proof" that this isn't just a vibes rally, read the business transcripts right away.

Alphabet revealed in its Q2 2025 earnings call that in 2025, it will have spent about $85 billion on capital expenditures, which is more than the $75 billion it had planned.

Most of this money will go toward technical infrastructure. It also said that capital spending would rise again in 2026 due to client demand and chances for expansion.

More AI Stocks:

• Morgan Stanley sets jaw-dropping Micron price target after event
• Bank of America updates Palantir stock forecast after private meeting
• Morgan Stanley drops eye-popping Broadcom price target
• Nvidia’s China chip problem isn’t what most investors think
• Bank of America sets AI stocks to buy list for 2026

Microsoft announced that in the first quarter of its fiscal year 2026, it spent $34.9 billion on capital expenditures. Half of that was for "short-lived assets" such as GPUs and CPUs, and the other half was for long-lived assets and massive data center sites (including finance leases).

In simple terms, this means that Big Tech isn't simply "interested" in AI; it's spending money to build an entire second economy around land, power, cooling, switches, racks, and fiber.

Chips are no longer the major problem for AI; it's electricity and schedules

The International Energy Agency is very worried. The organization forecasts that data centers around the world will use about 415 TWh of electricity in 2024, which is 1.5% of all electricity use.

The base case says this number could grow to almost 945 TWh by 2030, which is four times as much. That's not an easy fix and poses a problem for developing infrastructure.

Related: Nvidia’s $20 billion Groq play is a blueprint for 2026

The IEA forecasts that by 2030, the U.S. may need roughly 240 TWh more electricity for its data centers than it does now (and China may need about 175 TWh more).

That's why the AI market in 2026 doesn't look like "pick the best model" anymore. It looks more like:

• Who can get power the quickest?
• Who can give approval and build on time?
• Who can afford long-term assets without affecting profits?

Investors can't overlook the new AI risks: depreciation and "short-lived" gear

Alphabet acknowledged that increasing capital spending is hurting P&L, largely because of higher depreciation.

It also revealed that depreciation swelled to $5 billion in the second quarter, which is $1.3 billion more than at the same time last year.

Related: This $40 PC fix might be the quietest MSFT signal for 2026

Microsoft became even more specific: A lot of the money goes to "short-lived assets" including CPUs and GPUs, which is a polite way of expressing that this hardware burns out quickly.

That matters in 2026 because the market will be asking harder questions.

• Are these investments earning money fast enough?
• Are companies building long-term competitive advantages, or are they just spending money on the most expensive upgrade cycle in the world?

Three AI pressure points that can change markets in 2026

• Power becomes a limiting factor: The IEA's base-case arithmetic demonstrates that demand for data centers is expanding quickly enough that grid concerns can become critical on a bigger scale, not only in the news.
• Capex keeps rising, and Wall Street begins to give it a score. Alphabet boosted its forecast for capital spending in 2025 to $85 billion and predicted it would climb even more in 2026. Microsoft spent $34.9 billion on capital goods every three months. That's the kind of spending that makes investors care about more than simply the "AI narrative."
• M&A picks up speed as buyers search for bigger enterprises. The figure for the 2025 data center from Bloomberg is $70 billion. When everyone wants exposure at the same time, things that used to be dull become important.

Related: Novo and Lilly shift GLP-1 strategy abroad: US may be next