Three of Four Hyperscalers Now Burn Cash as AI Capex Hits Legacy Businesses

Meta joined Alphabet and Amazon in reporting negative free cash flow this quarter, leaving Microsoft as the only hyperscaler (the small group of giant cloud providers — Amazon, Microsoft, Alphabet and Meta) still generating positive cash. The milestone marks a new level of strain on the legacy businesses — online advertising, business software, e-commerce — that fund the tech giants' multitrillion-dollar artificial intelligence buildout.

The numbers are stark. Meta posted negative free cash flow of $2.12 billion. Amazon's outflows swelled to $42.4 billion. Alphabet's reached $9.07 billion. Microsoft stayed positive, but its cash generation dropped 27% to $18.63 billion, a second straight quarter of steep decline, and the company said it will now turn to the bond market as it continues to ramp up AI infrastructure spending.

Earnings over the past two weeks have coincided with intensifying concerns about the overall health of the broader AI sector. Rising interest rates are making it costlier for companies to raise debt, putting pressure on the cash-burning neoclouds — smaller, specialized GPU cloud providers — that have become critical to the global buildout. The tech giants hold a cushion the neoclouds lack. But investors eventually need the historic spending rates to convert into massive new revenue streams.

The market reaction was immediate. Alphabet suffered its steepest stock drop ever on Thursday. Amazon slid in extended trading after its report. Meta shares fell 7% after earnings last week even though the company beat on revenue.

Why does Alphabet's report worry investors most?

Alphabet's core search business recorded a third straight quarter of slowing growth and missed analysts' projections, a trajectory that underscores the competitive pressure from ChatGPT, Claude and other AI-powered bots and agents. YouTube-related advertising revenue also came in light.

Mark Mahaney, an analyst at Evercore, told CNBC's "Closing Bell Overtime" on Wednesday that while Google's strong cloud revenue was "good news" for the company, its decelerating digital advertising business represented "the overwhelming bad news."

Forrester analyst Tracy Woo said in an email that "Alphabet's massive capex investment in infrastructure buildout has been tolerated for now," but the company "is walking a fine line" because of the online ad sales misses.

"Weakness in Search and YouTube revenue performance relative to expectations surfaces growing concern about AI displacement of the company's largest revenue generators," Woo said. "The issue is less the current earnings miss and more the possibility that future AI spending grows faster than the cash generation of the franchises expected to support it."

That framing captures the central question of this earnings season: not whether the current quarter disappointed, but whether the businesses funding the AI buildout can keep growing while AI itself erodes them.

How bad is Amazon's slowdown?

Amazon said growth slowed in both its retail and advertising businesses. Online sales gained 2% year over year, decelerating from 10% a year ago and falling short of analysts' estimates. The midpoint of Amazon's revenue guidance came in lower than the $244.5 billion analysts were expecting — a particularly disappointing figure for the quarter that includes the holiday shopping season.

The spending continues regardless. Amazon expects capital expenditures this year to reach $220 billion, and analysts project that number will top $280 billion next year, according to FactSet, as the company continues to produce negative free cash flow.

Can Meta's AI products offset the cash burn?

Meta offers the clearest test of whether new AI revenue can arrive fast enough. Investors have responded favorably to the company's early success with its Muse AI agent app and related foundation models, which present potential new revenue sources outside of advertising.

But significant impact on the overall business will take time. Meta's capex will reach as high as $145 billion this year, the company says, and analysts see it passing $210 billion in 2027, alongside a deepening cash flow hole. Meta followed its earnings report by announcing on Monday that it plans to raise a total of $95 billion in stock to support its AI investments.

Is the spending sustainable?

Not everyone is concerned. Bullish investors point to the tech leaders' long operating histories and legacy businesses dominant enough to prop them up for the foreseeable future.

Christopher Ballard, a managing director at Check Capital Management, noted that Alphabet's negative free cash flow of $9 billion is tiny relative to the company's overall business. Its operating income topped $40 billion for the quarter.

Unlike AI model companies OpenAI and Anthropic, which remain unprofitable and spend billions of dollars a quarter on infrastructure, Alphabet carries a strong balance sheet, Ballard said.

"They're borrowing from the for-now in order to invest in the future," Ballard said. "I think it's OK, and they're doing it in a responsible manner."

The hyperscalers did not provide specific capex projections for next year, but all of them implied the data center spending spree will continue to grow — even if it means reporting uncomfortable metrics along the way.

"Would it be more convenient if this type of acceleration and opportunity happened over 25 years, and you can spread the capex over 25 years?" Amazon CEO Andy Jassy said on Thursday's earnings call. "Yeah, it might be more convenient, but that's not the world that we live in, and that's not the opportunity."

What happens next?

The bet embedded in these earnings is explicit: trillions in infrastructure spending will produce many years of profitable growth. That payoff sits in the future. For now, the equation depends on legacy franchises holding up under simultaneous pressure — decelerating ad revenue, rising interest rates, and the cash demands of the largest capital expenditure cycle in the industry's history. The coming quarters will test whether new AI revenue streams like Meta's Muse can arrive before the cash generation of the old businesses, in Woo's words, grows slower than the spending they are expected to support.