The markets bounced back a bit this week driven by the US Supreme Court's ruling that Trump’s sweeping tariffs were unlawful. This eased some fears in the market. There’s still some strong debate over the expected return on investment this AI spending will yield and over how disruptive and beneficial AI will be. The overarching uncertainty is why the market boom has slowed significantly.

Billions continue to be spent, investors are asking more questions, and we can’t seem to get the answers. Though big tech is still performing well, the fact that we can’t see a clear picture of the future is making it hard to justify these high stock prices. The hard truth is we don’t really know what we’re looking at on their financial statements.

Companies like Alphabet ($GOOG ( ▲ 3.74% )) (Google's parent company), Amazon ($AMZN ( ▲ 2.56% )) , and Meta ($META ( ▲ 1.69% )) are pouring tens of billions of dollars into building out massive data centers to power the AI revolution. That spending shows up on their financial statements in a category called Property, Plant & Equipment (PP&E) — and more specifically, in a sub-bucket called "Construction in Progress" or "Assets Not Yet in Service" (same category, just depends on the wording the company chooses).

Think of this bucket like a holding area. When a company is in the middle of building something like a data center, a server farm, a building — the costs sit here until the project is finished. Once it's "in service" (meaning it's up and running and generating value), the cost moves out of this bucket and the company starts depreciating it. Depreciation is how a company gradually expenses the cost of an asset over its useful life, spreading that hit to profits over many years.

The problem is, these balances are growing at a jaw-dropping rate, and investors have no way to know what's actually inside them. Also an important note, this balance is lumped into one number with no break out of what it contains. This is in line with accounting rules. With relatively small, normalized numbers, this isn’t an issue. When this number grows by billions each year, we have a problem.

Companies like Alphabet (Google's parent company), Amazon, and Meta are pouring tens of billions of dollars into building out massive data centers to power the AI revolution. That spending shows up on their financial statements in a category called Property, Plant & Equipment (PP&E) — and more specifically, in a sub-bucket called "Construction in Progress" or "Assets Not Yet in Service."

Think of this bucket like a holding area. When a company is in the middle of building something — a data center, a server farm, a building — the costs sit here until the project is finished. Once it's "in service" (meaning it's up and running and generating value), the cost moves out of this bucket and the company starts depreciating it. Depreciation is how a company gradually expenses the cost of an asset over its useful life, spreading that hit to profits over many years.

So far, so normal. The problem? These balances are growing at a jaw-dropping rate, and investors have no way to know what's actually inside them.

Not all assets depreciate the same way. Here's the key distinction that's keeping analysts up at night:

If you're an investor trying to forecast a company's future earnings, this distinction is enormous. A dollar spent on chips will reduce profits much faster than a dollar spent on brick and mortar. But right now, these companies aren't required to break out that detail and they aren't volunteering this information either.

Let's look at how fast these balances have grown:

Alphabet (Google) $GOOG ( ▲ 3.74% ) went from $50.5 billion in assets not yet in service to $78.5 billion in just one year — a 54% increase. The year before that, it grew 44%. That's roughly $28 billion added in a single year sitting in a bucket we can't see inside.

Amazon $AMZN ( ▲ 2.56% ) saw its construction-in-progress balance jump from $46.6 billion to $71.7 billion — a 54% increase, following a 62% jump the year before.

Meta $META ( ▲ 1.69% ) is the most striking case. Their balance went from $26.8 billion to $50.5 billion — a 92% increase in one year. And it may actually be much larger than that.

Meta is facing scrutiny for what its own auditors flagged as a financial engineering move. The company began building a massive data center called the Hyperion project, estimated at $27 billion. To keep that cost off its books, Meta created a joint venture with a firm called Blue Owl Capital, which holds 80% of the venture while Meta holds 20%.

On the surface, that structure might seem like it justifies keeping the project off Meta's balance sheet but Meta will be the sole user of that data center. They bear all the operational risks. They get the benefit. They control the asset. Under accounting rules, that kind of arrangement almost always means the asset has to be reported on your books, regardless of how the ownership is structured.

If that $27 billion does get added back — which I expect — Meta's construction-in-progress balance balloons from $50.5 billion to roughly $77 billion. A huge amount of spending whose future impact on earnings remains unclear.

When you can't clearly forecast a company's future expenses, you can't confidently value it. And right now, analysts are staring at balance sheets with massive, fast-growing, cloudy amounts of spending and being forced to make educated guesses about what it all means for earnings two, three, or five years down the road.

What we’re seeing is the response to uncertainty, not panic, and not a fundamental collapse in these businesses. Investors are de-risking their portfolios. It's rational caution in the face of incomplete information.

This doesn't mean the AI trade is over, or that these companies won't generate great returns. Until there's more transparency about what's being built, what it costs to maintain, and how fast those costs will flow through to earnings, investors are wisely demanding a discount for the uncertainty.

The question is, when these expenses hit the income statement, how much will the pressure be on earnings? Right now, nobody knows for certain. And in investing, uncertainty has a price.

Vertiv provides power management, thermal cooling, and infrastructure systems used in data centers and AI compute environments. The company benefits from rising server density and AI infrastructure spending, positioning it as a “picks-and-shovels” provider rather than a semiconductor risk play. Data centers can’t thrive without these services.