Big Tech Is Being Asked to Justify Its AI Spending as Chip Stocks Post Their Worst Week Since 2025

--- headline: "Big Tech Is Being Asked to Justify Its AI Spending as Chip Stocks Post Their Worst Week Since 2025" slug: big-tech-ai-spending-selloff category: business story_number: "01" date: 2026-07-19 ---

For three years, the question Wall Street asked about artificial intelligence was how much. Last week it changed to for what.

A brutal five-session stretch ending Friday, July 17, sent the Philadelphia Semiconductor Index down roughly 10% and erased more than $1 trillion in combined market value across the chip complex — the sector's worst week since 2025, and enough to push semiconductors into bear-market territory. The Nasdaq Composite fell 2.9% on the week. The VanEck Semiconductor ETF, the retail proxy for the AI trade, has now dropped more than 17% in July alone, its third weekly decline in four. On Friday the Dow sank 480 points as the selling widened out of chips and into the broader tape.

Nothing about the AI story technically broke. What broke was investors' willingness to keep funding it on faith.

Three cracks in the narrative

The triggers were mundane and cumulative. Samsung's quarterly results in early July missed a bar set implausibly high by AI demand assumptions. Reports emerged that SK Hynix was preparing to slow its expansion of high-bandwidth memory capacity — the most supply-constrained input in the AI stack — and reallocate toward lower-margin commodity DRAM. That is not the behavior of a company expecting insatiable demand eighteen months out.

Then came the competitive shocks. A wave of custom AI silicon designed in-house by the hyperscalers began shipping in late June and early July, the clearest evidence yet that Nvidia's near-monopoly on training compute is contestable. And Chinese startup Moonshot released a model it says performs on par with the frontier systems from OpenAI and Anthropic — a reminder that if inference token prices keep collapsing, the revenue side of the capex equation collapses with them. Applied Materials, Lam Research, Intel, KLA and Arm each fell about 4% in a single session; Nvidia and Micron dropped more than 2%.

The result is a market that has stopped treating announced spending as a growth signal. Higher capital plans only produced outperformance last earnings season when paired with upgraded revenue estimates. Capex funded by debt rather than operating cash flow was punished outright.

The arithmetic that spooked everyone

The numbers are genuinely without precedent. Amazon has guided to roughly $200 billion of capital expenditure in 2026. Alphabet has signalled $175 billion to $185 billion. Meta is at $115 billion to $135 billion, and Microsoft is running at an annualized $145 billion. Aggregate Big Tech capex for the year lands somewhere between $650 billion and $725 billion, with sell-side forecasts now putting 2027 above $1 trillion.

Against that, the cash-flow math has inverted. AI capital spending consumed about 33% of hyperscaler cash flow from operations in 2023. In 2026 that figure is an estimated 93%. Torsten Slok, chief economist at Apollo Global Management, put it bluntly in a July 13 note: free cash flow for the hyperscalers has "literally [dropped] down over the next six, 12 months towards zero. And it might even begin to go negative."

Slok's broader warning is what gives the selloff its macro edge. AI capital spending now accounts for roughly half of US GDP growth. "A slower payoff wouldn't just be a sector problem," he wrote. "It would risk tipping the economy into recession and the S&P 500 into a correction." On the competitive threat he was equally direct: "If Chinese models keep gaining and token prices keep falling, the hyperscaler cash flows expected may prove too optimistic."

Why this matters

Three weeks before the selloff, the Bank for International Settlements — the central bank of central banks — used its flagship Annual Economic Report to name an AI bust as one of the "pressure points" that "demand attention." The BIS noted that the five largest hyperscalers are on pace to spend over $1 trillion on AI-related capital expenditure across 2025 and 2026 combined, a sum already outrunning their earnings and free cash flow and forcing several to issue debt to close the gap.

The Basel institution's sharper concern was structural. It described the sector as bound together by "a complex web of private arrangements," most notably circular financing: hyperscalers take equity stakes in AI labs, which in turn commit to multi-year purchases of chips or computing power from those same hyperscalers. Private credit lending to AI-related companies has grown from roughly $3 billion in 2010 to over $40 billion in 2025, and collateral in some circular deals may be "pledged multiple times." Disappointing returns, the BIS warned, "could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust."

BIS General Manager Pablo Hernandez de Cos was more pointed still: "The race to capture market share may have led to overinvestment."

That is the crux. The bear case is not that AI is useless — enterprise adoption is still strong, and Google Cloud is carrying a $240 billion backlog. The bear case is a timing mismatch: the spending curve has run so far ahead of the monetization curve that markets will reprice the difference before the payoff arrives. Depreciation on the 2025 and 2026 buildout hits income statements now. The revenue from those assets does not.

And if financing dries up, the pain does not stay in tech. It travels through private credit and through the vendor-financing chains linking chipmakers to labs to clouds. A sector correction and a credit event are not cleanly separable outcomes.

What to watch

Second-quarter results from Alphabet, Microsoft, Amazon and Meta land over the next two weeks, and the market's tolerance for capex guidance unaccompanied by revenue guidance is visibly exhausted. Watch three things: whether any hyperscaler trims — or merely "reprofiles" — its 2027 plan; whether cloud backlog growth decelerates; and whether the debt-funded share of capex keeps climbing. The awkward part is that discipline and disappointment now look identical from the outside. A CFO signalling restraint could be read as prudence, or as quiet confirmation that the demand was never going to be there.