Is the AI Trade Over? The Next Phase of AI Investing Has Begun

Is the AI trade over?

I don’t think so.

But the easy-money version of the artificial intelligence boom may be.

At some point, spending a few hundred billion dollars each year starts to sound almost normal.

At least when Amazon, Alphabet, Meta, and Microsoft are doing it. Together, those four hyperscalers spent roughly $165 billion in one quarter building AI data centers, cloud capacity, chips, networking, and related infrastructure.

Current estimates have quarterly spending approaching $225 billion by the end of 2026, while one forecast puts total hyperscaler capital spending above $1.2 trillion over the following 12 months.

Needless to say, the AI buildout is not slowing down anytime soon.

But investors are becoming increasingly focused on a much less exciting question:

Who is going to pay for all of it?

The Return Question

For the last few years, investors have treated rising AI capital spending as an unquestioned positive for AI stocks.

A company announces another data center, orders more Nvidia chips, or raises its AI budget and the conclusion is simple: More AI!

And from a business perspective, some of this makes a ton of sense.

The companies making these investments clearly believe generative AI will reshape search, advertising, cloud computing, software, and much more. If they’re right, underinvesting today could prove far more costly than overspending.

Still, these are not pretend dollars.

Amazon’s free cash flow turned negative by $7.6 billion in the second quarter. Alphabet’s turned negative by $5.9 billion. Meta generated only $1.7 billion of free cash flow while spending roughly $30 billion on AI and other capital projects.

Microsoft looks stronger than the rest, but even its capital spending has more than doubled from a year ago.

Alphabet recently completed its largest bond sale ever. Bank of America expects hyperscalers to issue $175 billion of investment-grade debt in 2026 to help finance the AI buildout.

These companies can afford it.

But “they can afford it” and “this will produce an attractive return” are not necessarily the same thing.

And that distinction is beginning to matter more and more.

Nvidia remains the face of the AI stock trade. It is also not the only player in the game.

Intel and AMD have significantly outperformed Nvidia this year. Qualcomm has fallen. Semiconductor stocks are no longer moving as a single AI basket simply because they have exposure to artificial intelligence.

Even Amazon appears motivated to reduce its dependence on outside suppliers. Its in-house chip business reportedly carries an estimated value comparable to a stand-alone company generating more than $25 billion in annual revenue.

Apparently, buying enormous amounts of expensive hardware has a way of encouraging companies to find cheaper alternatives.

This is why I think the more interesting AI investment opportunities may be outside the companies writing the biggest checks.

The Next Phase of AI Investing

Every AI data center needs networking equipment, electrical systems, power, construction, high-bandwidth memory, and cooling. Those infrastructure companies get paid whether the eventual AI application is revolutionary or merely helpful.

Of course, that doesn’t mean every company with AI exposure is a good investment.

Arista Networks, Dell Technologies, and Caterpillar are up strongly as AI spending drives demand for networking, servers, power generation, and construction equipment. Meanwhile, Oracle, IBM, Accenture, and Tesla have lagged as investors question whether AI spending will produce attractive returns, pressure existing businesses, or both.

Adding “AI” in an investor presentation no longer guarantees a higher stock price.

The AI infrastructure opportunity is spreading globally too. Danfoss, a Danish industrial company supplying data center cooling and power systems, expects that business to at least double its share of sales. Samsung and SK Hynix are benefiting from surging demand for AI memory, while earnings expectations in Taiwan have climbed with semiconductor demand.

So, is the AI trade over? No.

AI spending is accelerating. Demand for computing power is real. The data centers, chips, power systems, memory, networking, and cooling infrastructure still need to be built.

But the easy money may be over.

For a while, investors could buy almost anything connected to artificial intelligence and feel smart. Chips. Software. Power. Cooling. Data centers. Close enough!

The next phase of AI investing will require more judgment. Who is spending the money? Who is receiving it? Who has pricing power? And most importantly, which companies are earning an acceptable return on AI investment?

I believe artificial intelligence will reshape nearly every business.

I also believe investors can overpay for a trend.

Both can be true at the same time.

Do you think the biggest AI winners will be the companies spending the money, or the ones getting paid to build the infrastructure?