The four largest US technology companies are on course to spend close to $700bn on artificial intelligence infrastructure this year, and the cost is starting to show up in the one figure that is hard to dress up: free cash flow. An analysis by Reuters found that combined capital spending across the major cloud operators is set to overtake the cash their core businesses generate. Wall Street’s consensus for this year’s AI capex has climbed from roughly $485bn in January to about $730bn by July, on the figures Reuters compiled.
The sharpest early warning came from Amazon, whose free cash flow fell to $1.2bn on a trailing 12-month basis in the first quarter, down from about $26bn a year earlier, even as operating cash flow rose 30% to $148.5bn. The mechanics are straightforward. Capital spending is growing far faster than the money coming in. Epoch AI estimates that hyperscaler capex is expanding at roughly 70% a year while operating cash flow grows about 23%, which puts aggregate spending on track to overtake operating cash flow around the third quarter of 2026.
Reuters put the gap in blunt terms. Between 2025 and 2027, capital expenditure across Microsoft, Alphabet, Amazon, Meta, and Oracle is expected to rise by about $534bn, against a roughly $340bn increase in operating cash flow. That works out at $1.57 of capex for every additional dollar of cash the businesses generate. Not every company is under the same strain. Microsoft reported $37.5bn of capital spending, including finance leases, in its fiscal second quarter, against $35.8bn of operating cash flow, and has guided towards roughly $190bn for the year. Alphabet and Meta are still producing enough cash to cover dividends and buybacks, at least for now.
The guidance keeps climbing regardless. Meta has said it will spend up to $145bn this year, and Alphabet has reset its own bar higher for a second consecutive quarter. Asked about the return on all that outlay, Meta chief executive Mark Zuckerberg told analysts in April that it was “a very technical question,” which is roughly the answer the market has been given across the sector. Oracle is furthest down the road. Its capex reached 174% of operating cash flow in fiscal 2026, up from 47% four years earlier, and its free cash flow has turned negative.
The company’s credit rating sits one notch above junk, and it has signalled plans to raise between $45bn and $50bn to keep building. That points to the wider shift. Most of the hyperscalers have already turned to external financing, whether cash reserves, bond issuance, or equity, to fund the build-out rather than pay for it out of operations. Amazon, Alphabet, and Meta have all tapped the bond markets in recent months, at a scale that has begun to reshape corporate debt issuance on both sides of the Atlantic.
The question investors keep returning to is whether the spending pays off. “AI is pushing them toward a hybrid model where software, advertising and cloud economics increasingly depend on enormous physical infrastructure spending,” Shay Boloor of Futurum Equities told Reuters. Others are more cautious about the timeline. David Russell of TradeStation warned that “earnings growth may not be enough to justify investment if capex is depleting cash,” while Freddy Lavric of Winthrop Capital said companies would need two to three years to show the spending translates into incremental revenue and improving margins.
This massive capital outlay is reminiscent of earlier technology investment cycles, such as the fiber-optic buildout of the late 1990s or the rapid expansion of data centers during the cloud computing boom of the 2010s. However, the scale of AI infrastructure spending is unprecedented. The shift from traditional data centers to AI-specific hardware, including high-performance GPUs from Nvidia, custom AI chips, and advanced cooling systems, has dramatically increased costs. Moreover, the competitive pressure to deploy large language models and generative AI services has forced companies to invest heavily even before clear revenue streams are established.
For Amazon, the steep decline in free cash flow is particularly notable because the company’s core retail and AWS businesses have historically generated strong cash flows. AWS itself is a major beneficiary of AI spending, as many enterprises run their AI workloads on its cloud platform. Yet the capital required to expand AWS's AI capabilities is so large that it is straining even Amazon’s financial resources. In the first quarter, Amazon’s free cash flow dropped to $1.2 billion, compared to $26 billion a year earlier, driven by a surge in capital spending on data centers and AI infrastructure.
Microsoft’s situation is more nuanced. The company reported $37.5 billion in capital spending in its fiscal second quarter, against $35.8 billion in operating cash flow, meaning it spent slightly more than it generated from operations. Microsoft has guided toward roughly $190 billion in capital spending for the full fiscal year, which would exceed its operating cash flow by a wide margin. However, Microsoft benefits from a large cash reserve and strong revenue growth from its Azure cloud and Office 365 products. The company is also embedding AI into its Copilot products, which could eventually drive significant incremental revenue.
Alphabet and Meta are still in a relatively comfortable position. Both companies generate substantial free cash flow and have net cash positions that allow them to cover dividends and share buybacks. Alphabet reported $69 billion in free cash flow for 2023, while Meta generated $43 billion. Yet both are rapidly increasing their capex guidance. Alphabet’s capital spending rose to $32 billion in the first quarter of 2024, up from $28 billion in the same period last year, and the company said it expects to spend significantly more in the coming quarters. Meta’s capex guidance for 2025 is as high as $145 billion, more than triple its 2023 level.
Oracle is the extreme case. The company’s capex has risen to 174% of operating cash flow, and its free cash flow has turned negative. Oracle has been aggressively building out its cloud infrastructure to compete with AWS, Microsoft Azure, and Google Cloud. The company has also invested heavily in AI services, such as its partnership with Nvidia to offer GPU clusters. However, Oracle’s relatively smaller scale and higher debt load make it more vulnerable to cash flow pressures. Its credit rating is just above junk status, and it has signaled plans to raise $45–50 billion in additional funding to continue its build-out.
The use of external financing is a key theme across the industry. Amazon, Alphabet, and Meta have all issued bonds in recent months, taking advantage of relatively low interest rates despite the Federal Reserve’s tightening cycle. Amazon raised $10 billion in a bond sale in February, while Alphabet issued $5 billion in debt in March. Meta tapped the bond market for $6 billion in April. This wave of corporate debt issuance has reshaped the investment-grade bond market, with technology companies now accounting for a larger share of issuance than in previous years. Some analysts warn that if interest rates remain high or the economy slows, the companies could face higher refinancing costs.
Investors are closely watching the upcoming quarterly earnings reports for signs that the AI spending is beginning to pay off. The key metrics will be not only revenue growth but also margins and free cash flow. If capital spending continues to outpace cash generation, the companies may need to cut dividends, reduce buybacks, or raise additional capital. The pressure is most acute for Oracle and Amazon, but even Microsoft and Alphabet could face scrutiny if their AI investments do not translate into measurable returns within the next few years.
For now, the accounting cushions the blow. Because capital spending is depreciated over years rather than booked upfront, all of the big spenders remain profitable, and increasingly so. Free cash flow is simply where the pressure lands first. The next test comes with the quarterly earnings due in the coming weeks. Investors will be watching capex guidance at least as closely as revenue, and for once the two numbers may well pull in opposite directions.