Meta’s AI spending faces investor doubts as cash flow collapses

Meta raised its 2026 capital-expenditure guidance to a range of $130 billion–$145 billion on July 29, even as the company reported a near-collapse in free cash flow, intensifying investor questions about the payoff from an enormous AI buildout. The move followed management signalling plans to deploy up to 14 gigawatts of compute next year and to monetise spare capacity—an argument that has so far failed to reassure markets. (reuters.com)

The dispute is straightforward: can Meta fund what Morgan Stanley and others expect will be a trillion-dollar industry capex wave while still generating attractive returns for shareholders? Chief Financial Officer Susan Li told investors the industry has “underbuilt historically” for AI demand and that existing capacity is “extremely valuable,” asserting capacity will remain tight “for the foreseeable future.” (reuters.com) But traders and analysts are signalling growing scepticism as Meta's cash generation falters. (wsj.com)

Why Susan Li says capacity is “extremely valuable”

Meta has framed the spending as defensive and strategic. Management argues that the industry underinvested in large-scale AI compute and that scarcity will create a market for both models and the compute that runs them. Zuckerberg has suggested Meta could earn higher margins by selling AI “intelligence” or hosted models rather than raw compute, and the company is exploring selling access to hosted models or spare machine time. (reuters.com)

That pitch has internal logic: supply-constrained markets can sustain elevated margins, and renting out spare capacity resembles Amazon’s cloud business, a profitable model many investors respect. (cnbc.com) Yet the thesis rests on two uncertain bets—demand will outstrip supply for multiple years, and Meta can pivot from consumer ad products to enterprise-grade compute sales—both of which remain largely unproven in public results so far.

Investors push back as free cash flow plunges

Market unease hardened after Meta disclosed a second-quarter free cash flow decline of roughly 91% year‑over‑year to $784 million, a collapse Reuters said “laid bare the financial strain” of the buildout. (reuters.com) Analysts on the Wall Street Journal podcast warned the company could enter negative free cash flow in the latter half of 2026—the first such stretch since Meta’s IPO.

Bloomberg and other outlets describe a “Wall Street test” for Meta: investors want concrete evidence that new products or compute sales will translate into durable margins and cash flow. Bloomberg grouped Meta with Microsoft and Amazon as firms facing renewed investor anxiety over debt‑fuelled capex plans, underscoring that the problem is industry-wide, not unique to Meta. (bloomberg.com)

How Meta’s strategy stacks up against Alphabet, Microsoft and Amazon

Competitors have raised capex guidance too, but the business models differ. Alphabet and Microsoft are tying infrastructure more directly to cloud and productivity ecosystems; Amazon already monetises vast server capacity through AWS. Meta’s differentiator is its pitch to monetise both consumer AI products and the compute that runs them—an approach that would turn a historically ad‑centric company into a hybrid cloud‑and‑AI vendor. (cnbc.com)

Critics note that Amazon and Microsoft earned those monetisation routes over many years and with established enterprise sales channels. Meta lacks that sales footprint today, and investors are sceptical that product-led monetisation alone will close the gap before cash metrics deteriorate further. (bloomberg.com)

Meta's higher capex guidance—now up to $145 billion at the top end—signals escalation rather than prudence, even as management narrows the low end of its range. (reuters.com) That shift explains why stock reaction and analyst commentary have been so negative: the company is amplifying a long‑term, high‑risk bet at precisely the moment its cash cushion is thinning.

Looking ahead, investors will watch two concrete metrics for validation: quarterly free cash flow and any early revenue from hosted models or compute rentals. If Meta can’t show improving cash generation or convincing early monetisation, the company’s argument for continued heavy AI investment will grow notably weaker—and the market is already signalling it may not wait.