The $760 Billion AI Risk That Nobody Is Talking About. But Should.

$760 billion. That's roughly how much Amazon (NASDAQ: AMZN), Alphabet, Microsoft, and Meta Platforms are expected to spend on capital expenditures in 2026.

The number is staggering. It's also growing rapidly. The four companies spent roughly $413 billion in 2025, meaning their combined capital spending could jump by more than 80% in just one year.

Much of that spending is going toward the infrastructure needed for artificial intelligence: data centers, servers, GPUs, custom chips, networking equipment, and power capacity.

The bull case is easy to understand. AI demand is exploding and cloud providers can't build capacity quickly enough. But there is another possibility investors should consider. What if the industry is building too much capacity?

That's the risk I think investors should be watching.

Image source: Getty Images.

The numbers are difficult to ignore

The explosion in demand for AI is real. We don't have to look far to understand that -- just think about how much we engage with ChatGPT to answer our questions, or how much we rely on Gemini or Claude to solve our work problems.

So, there is no doubt that companies must invest to provide these services. But $760 billion for four companies is staggering. In fact, Goldman Sachs forecasts AI investments will exceed $1 trillion in 2026.

And strategically, it makes sense for these big tech companies to invest heavily to gain market share in the AI computing industry. From management's perspective, sitting still could be far more dangerous than spending aggressively. So the companies are racing to build. And that's where the risk begins.

Customer commitments don't eliminate the risk

One of the strongest arguments supporting this spending is that much of the infrastructure already has customers waiting for it.

Amazon, for example, says customer commitments back a substantial portion of its 2026 AWS capital expenditure. Microsoft similarly points to its enormous customer demand.

That's reassuring. But investors should be careful about what these commitments actually mean: Customers want the capacity today. They don't necessarily tell us how much capacity those customers will want five years from now.

That's an important distinction. AI is developing extraordinarily quickly. Models are becoming more efficient. Hardware is improving. Businesses are still figuring out which AI applications will generate genuine economic value. A company can sign a multiyear commitment today and still find that its computing requirements change dramatically over time.

The same thing happened throughout previous technology cycles: Demand forecasts made during periods of rapid adoption can look very different once the technology matures.

A five-year contract may provide visibility. It doesn't provide certainty about the ultimate economics of the AI industry.

The biggest question is asset utilization

Imagine a hotel owner building 1,000 rooms because demand is growing rapidly. If occupancy reaches 95%, the investment looks brilliant. If occupancy eventually falls to 50%, the same building becomes a capital-allocation problem.

AI infrastructure works similarly. Companies are spending enormous amounts of money today to ensure they have enough computing capacity tomorrow. The crucial variable is how intensively that capacity gets used -- and at what price.

If AI demand continues to scale sustainably in the next decade, the returns could be spectacular. But if computing becomes dramatically cheaper, AI models become more efficient, or customers discover that they don't need as much infrastructure as expected, the economics could change.

But the bull case is still powerful

None of this means the AI infrastructure boom is a fake. In fact, there is considerable evidence that the demand is real.

For instance, AWS reported a 37% rise in revenue in the second quarter of 2026 to a $169 billion annualized revenue run rate, the fastest growth in 18 quarters. Similarly, Google Cloud saw revenue surge 82% to $24.8 billion in the same quarter.

In other words, these big tech companies aren't making these investments from weak financial positions. They are reinvesting their profits into the capex, expecting to generate even more profits in the future.

The only question left is whether the investment will earn a sufficiently high return over time.

What does it mean for investors?

I wouldn't bet against Amazon, Microsoft, Alphabet, or Meta simply because they're spending hundreds of billions of dollars on AI. These companies have some of the strongest balance sheets, technology platforms, and customer relationships in the world.

But I also wouldn't assume that every dollar of AI capital expenditure will automatically generate an attractive return. In other words, this is ultimately a capital-allocation issue.

Investors should ask the most important question here: Will the incremental dollars invested in AI infrastructure generate an attractive incremental return five or 10 years from now? That's the number investors can't see yet, but should closely track in the coming quarters.

Lawrence Nga has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Meta Platforms, and Microsoft. The Motley Fool has a disclosure policy.