Meta Platforms might not be the first Big Tech company to rein in AI spending. But the argument that it should be is growing stronger.

Meta is the most financially stretched of the biggest tech companies. Its ability to generate returns from artificial intelligence hinges almost solely on ad sales. And it has no record of building successful new businesses beyond its core social-media franchise.

Those flaws are coming into sharper focus as the AI spending boom pushes into overdrive. Meta is anticipating capital expenditures of around $137.5 billion this year -- a total that analysts expect to put it into negative free cash flow territory in the latter half of the year for the first time since its IPO in 2012. The company will likely fall deeper in the hole next year, when projections via FactSet suggest it will burn through more than $20 billion in cash.

Investors have given Big Tech a long leash to invest in AI over the past few years. That made sense given the capital intensity of what looked like a potentially world-changing boom.

But shareholder patience is being tested like never before, as tech giants take AI ambitions to the next level even as the cost of computer memory and other AI-infrastructure components rises. Meta's stock fell 10% in after-hours trading Wednesday after it raised the midpoint of its 2026 capital-spending range and fell short of Wall Street's second-quarter profit expectations.

Of course, Meta isn't the only company that is bleeding money to fund AI. So are Amazon.com and Alphabet. Google parent Alphabet's stock fell nearly 7% last week after its earnings included a $10 billion hike in capital spending this year.

But Meta's revenue is more modest than that of its competitors and its cash burn could easily get much worse. Meta's anticipated negative free cash flow next year hinges on a projection of about $174 billion in capital spending. But that might be a vast underestimate.

Following reports that the company might double its computing capacity from 7 gigawatts this year to 14 gigawatts next year, Deutsche Bank analysts recently forecast up to $215 billion in spending next year. Raymond James analysts are projecting a headier $280 billion.

In Meta's favor, its core ad business is healthy and growing, and analysts' assessments of its position in the market suggest it is getting stronger. The company on Wednesday reported 28% higher second-quarter revenue, driven almost entirely by ads, and it forecast more strong growth ahead. Meta's use of AI to improve ad performance is unmatched in the industry.

Meta has long relied on that golden goose to fund its other ventures, from smartglasses to videoconferencing devices. But the spending now is reaching a level that is testing the company's core.

As it bleeds cash, Meta will need to borrow more money to keep up spending. Borrowing has already changed the dynamics of the company's balance sheet in the AI era: Meta had minimal debt until 2022, but its long-term borrowings reached $83.7 billion at the end of the second quarter. That total doesn't include tens of billions of dollars in off-balance-sheet debt tied to massive data-center projects in Louisiana and Texas. Meta has agreed to lease those sites and is committing to various financial guarantees over them.

And the cost of raising new funds is likely to go higher. The 40-year bonds Meta sold in May have already fallen substantially in price, pushing their yields near 7%. If AI bond supply continues to flood the market, the interest rates Meta pays on future debt will surely rise.

Meta executives seem to understand that they need to show more financial returns to justify the spending. Meta has been pushing subscriptions for premium social-media features to bring in new revenue. This month, it gave outside developers paid access to its advanced Muse Spark model. And it is selling AI services such as chatbots to businesses large and small.

Meta could also become a player in cloud computing by leasing out computing infrastructure it doesn't need -- a possibility Chief Executive Mark Zuckerberg said was on the table on Wednesday. Meta was already getting offers for computing power at a "significant premium" over what it paid, he told analysts.

Those could be revenue goldmines. In a recent note, Raymond James analysts suggested cloud revenue could grow from $4 billion this year to $120 billion by 2030. The company's AI model and business offerings could go from close to zero this year to a combined $59 billion in 2030.

But those businesses are nascent and could just as easily fizzle. The subscriptions Meta is offering don't look all that promising. The company's lack of corporate-software options and its reputation for getting in hot water politically might keep many businesses from fully embracing it.

A push into cloud computing holds more promise, but would likely be tactical in nature and limited to selling large chunks of computing power to big AI labs like Anthropic. Competing fully with Amazon.com, Google and Microsoft by drawing in millions of customers large and small would be a gargantuan effort that Meta can ill afford to undertake.

Meta Chief Financial Officer Susan Li on Wednesday hinted that Meta could eventually blink on spending -- but only in 2028 and after. Meta, she said in a call with analysts, was focused on maximizing computing capacity this year and next, but spoke of "flexibility" and "being able to evaluate our actual needs in '28 and beyond."

But investors won't wait that long. Meta needs spending on that capacity to start paying off, and soon.