AI infrastructure spending puts free cash flow levels in focus

AI spending is pressuring cash flow at major technology companies, bringing the sustainability of the sector’s momentum into focus

The release of ChatGPT in 2022 ushered in the AI era. Since then, technology stocks have emerged as a key driver of market performance. The extraordinary gains have naturally sparked questions about whether the momentum can continue, particularly as technology companies invest heavily in AI infrastructure. At the heart of the AI spending debate is free cash flow (FCF).

Here’s the formula:

Net cash from operating activities, or operating cash flow, measures the cash a company generates from its regular ongoing business, while capex represents cash invested in future growth.

In today's AI race, that investment is flowing primarily into massive data centers, particularly hyperscale facilities that power the next generation of AI applications. At the center of this buildout are the largest hyperscalers¹: Alphabet, Amazon, Meta, Microsoft and Oracle. Together, these companies account for the vast majority of the more than $750 billion expected to be spent on data center infrastructure in 2026.

Investors are increasingly focused on a key question: Can this pace of spending continue? Until recently, the group of hyperscalers generated enough cash to comfortably fund its investment plans. However, by the second quarter of 2026, aggregate capex began to exceed operating cash flow, pushing free cash flow into negative territory. Importantly, these companies remain extraordinarily profitable, so the issue isn’t profitability. It's that AI investment has become so large that even their substantial cash flows no longer fully cover it.

Chipmakers like NVIDIA, Broadcom, AMD and Micron sit on the other side of the AI value chain. Rather than building data centers, they supply the critical hardware that goes inside them. Every dollar hyperscalers spend on AI infrastructure is revenue for chipmakers. While these companies have their own capex – building or expanding their factories – the surge in hyperscaler investment has augmented their free cash flow.

Negative FCF isn’t necessarily a bad thing. While it can reduce a company's flexibility to pay dividends, repurchase shares, pursue acquisitions or pay down debt, negative FCF is often normal for companies making heavy investments to fuel future growth. The tradeoff is that they must find other sources of funding to bridge the cash flow gap.

For hyperscalers, that has largely happened through two channels:

• Debt issuance: Hyperscalers are relying more heavily on debt markets. In 2025, the five largest hyperscalers issued a combined $136 billion of debt, financing roughly one-third of their capex needs. In the first half of 2026 alone, issuance climbed to nearly $200 billion across multiple currencies. As leverage increases, borrowing costs are rising as well, reflected in higher credit default swap prices on their bonds.
• Reduced stock buybacks: Simultaneously, hyperscalers are scaling back share repurchases after years as some of the largest buyers of their own stock in the S&P 500. Aggregate buybacks among the five largest companies peaked at $50 billion in the fourth quarter of 2021 but have since fallen to just $5 billion in the second quarter of 2026.

Many tech companies beyond chipmakers, including software firms and service providers, continue to generate positive free cash flow. In fact, the sector remains FCF positive overall, with consensus expecting FCF to grow 47% in 2026.

Looking across all 11 S&P 500 sectors, technology's FCF yield (FCF divided by market cap) is roughly in line with the S&P 500 average of 4%. It's also important to remember that FCF is not static and capital spending is cyclical. With AI adoption still in its early stages, data center investment is surging. Over time, as more data centers come online, hyperscalers should be able to moderate capex while benefiting from the additional revenue those facilities generate.

Several hyperscalers reported accelerating demand for AI-related services, particularly cloud computing, with demand still exceeding available capacity. That helps explain the need for continued investment today. As revenue grows and capex eventually normalizes, FCF should improve, although the timing will vary by company.

Ultimately, the investment case depends on AI spending generating sufficient future revenue. If adoption or monetization falls short, the recovery in FCF could take longer, which helps explain investor concerns. Bottom line: Negative FCF is a reflection of where hyperscalers are in the AI investment cycle, not a sign that the cycle is ending.

¹hyperscaler is a large-scale cloud service provider that offers massive computing, storage, and networking infrastructure. All expressions of opinion are those of Investment Strategy and are subject to change. This information should not be construed as a recommendation. The foregoing content is subject to change at any time without notice. Content provided herein is for informational purposes only. There is no guarantee that these statements, opinions or forecasts provided herein will prove to be correct. Past performance is not a guarantee of future results. Indices and peer groups are not available for direct investment. Any investor who attempts to mimic the performance of an index or peer group would incur fees and expenses that would reduce returns. No investment strategy can guarantee success. Economic and market conditions are subject to change. Investing involves risks including the possible loss of capital. The information has been obtained from sources considered to be reliable, but we do not guarantee that the foregoing material is accurate or complete. Source: FactSet