July 23, 2026, (Inside AI) — Alphabet just posted its first cash burn on record, a $5.9 billion outflow in the second quarter that jolted investors and sent its stock sliding 5% before Thursday’s open. The Google parent’s aggressive AI buildout is now consuming more cash than its operations generate, a stark reversal for one of the world’s most profitable companies.
The cash drain came even as Google Cloud revenue surged 82%, a record pace that hints at market share gains against larger rivals Amazon Web Services and Microsoft Azure. But the growth wasn’t enough to offset the soaring capital expenditures required to train and serve AI models. Alphabet now expects to spend $15 billion more in 2026 than previously planned, with another increase forecast for next year.
This marks a fundamental shift in Big Tech’s financial model. Once prized for fat margins and reliable cash gushers, the group is increasingly leaning on debt and share sales to fund AI infrastructure. Combined capex for the four hyperscalers—Alphabet, Microsoft, Meta, and Amazon—is projected to top $700 billion this year, outstripping their cash flows. The capex-to-revenue ratio, a key metric, is set to nearly double across the board. Meta is expected to hit 54.9% from 35.9%, Alphabet 41% from 23%, Microsoft 45% from 31%, and Amazon 25% from 18%.
Investors are bracing for more pain. Next week’s earnings from Microsoft, Meta, and Amazon will be scrutinized for similar spending hikes. Shares of all three fell between 2% and 4% in sympathy with Alphabet’s slide.
“The risk is tilted towards further increases, particularly while Microsoft and others remain capacity-constrained,” said Charu Chanana, chief investment strategist at Saxo Markets.
“But investors will increasingly focus on how much of that cash must be reinvested simply to remain competitive—and whether AI revenue can grow faster than capital expenditure, depreciation and operating costs.”
Cloud Growth Masks a Deeper Capital Strain
Google Cloud’s performance adds pressure on Amazon and Microsoft, whose cloud divisions have grown more slowly. AWS is expected to report 31.04% growth for the quarter, up from 28.4% in the prior period, while Azure’s growth is seen at 39.98%, roughly flat with the previous quarter’s 40%. That relative stagnation could weigh on Microsoft shares, already the worst performer in the “Magnificent Seven” this year with a decline of nearly a fifth.
Alphabet’s AI demand is so intense that executives plan to lease data-center capacity from other companies to serve clients, even though it will compress margins. At least 20 brokerages raised their price targets on Alphabet after the results, lifting the median to $430, nearly 26% above the last close. Citizens was most bullish at $515, while TD Cowen was most bearish at $240.
“Google Cloud was an absolute blow out,” said Richard Clode, Portfolio Manager of Janus Henderson Investors' Global Technology Leaders. “Alphabet has competitive advantage running all the way through the stack from their own custom AI chips through to distribution to billions of users.”
Analysts expect Alphabet and Amazon to burn cash in 2026, while Meta’s cash flow is likely to shrink 95.7% to just $1.85 billion. Microsoft, whose fiscal year ends next June, is forecast to generate $25.39 billion in cash, less than half the estimated $58.74 billion from the previous year.
Competition is deepening as Meta engages in talks to rent computing power to Anthropic, joining an industry that already includes AI cloud firms like CoreWeave. “As compute becomes more available and models become cheaper, cloud capacity may look increasingly interchangeable. That could force providers to spend more while accepting lower returns,” said Lale Akoner, global market strategist at eToro.
The cash burn underscores a broader reckoning: AI’s infrastructure demands are testing the limits of even the richest tech firms. With no clear endpoint to the spending cycle, the question is whether revenue can ever catch up—or if investors will lose patience first. Research on the economics of large-scale AI training, such as this analysis of compute costs, suggests that model scaling remains deeply capital-intensive, and efficiency gains may not offset the appetite for ever-larger systems.
