Like "bottomless espresso martinis," "free cash flow" is a magical phrase that gets better with every word. But investors who shy away from companies that burn cash may be making a mistake.

That's the takeaway from a new paper written by Michael Mauboussin of Morgan Stanley. Since free cash flow (FCF) is the money that a company can return to its investors, investors obviously want it to be positive-eventually. But that last word is key. A successful company can use its cash to generate higher returns than investors can on their own, which means that when it pays a dividend or buys back stock, everyone loses.

"A company with opportunities that promise sufficient returns should invest in all of them because those investments create value," Mauboussin writes.

This debate has become newly relevant. The five hyperscalers-Amazon.com, Alphabet, Microsoft, Meta Platforms, and Oracle-are spending some $1 trillion a year to build out artificial-intelligence infrastructure, and their annual free cash flow is expected to show a decline of $440 billion from the first quarter of 2024 to the third quarter of 2027, Mauboussin shows. But from a trough of negative $265 billion, it is expected to soar phoenix-like to $505 billion by 2030.

Without making a specific call on the hyperscalers, Mauboussin argues that if a company can finance its operations, and generates returns on investment that exceed its cost of capital, then "negative FCF is not only fine, but desirable."

Investors increasingly believe it is. Most don't care about the spending-they care about what the money is being spent on, says Sung Cho, co-head of public technology investing at Goldman Sachs Asset Management, who makes a distinction between creating value and just experimenting without a plan to monetize the results. "Investors don't necessarily worry about negative free cash flow in isolation," Cho says. "The market is saying that because use cases are now proliferating, it's a worthy cause for you to go free-cash-flow negative to accrue terminal value in the future."

The numbers back him up. Analysts expect the hyperscalers' collective return on invested capital to rise a bit from here, after bottoming at 24% in the second quarter of 2026. Clearly, that's far more than investors should expect to get from their money if investing it on their own.

We remain a bit more skeptical. When you're building massive physical infrastructure to help unprofitable companies serve up an ever-changing product, there's a lot that can go wrong between here and free-cash-flow Shangri-La. But we're not scoffing at the forecasts either, particularly since the hyperscalers are already showing strong returns from their data-center investments. We simply hope to give investors a sense of the strange landscape they're wading into when they buy the hyperscalers right now.

Because it's one thing to be OK with falling free cash flow. But when it plummets by half a trillion bucks, you'd better love it.