According to Woofun AI, the U.S. stock market experienced rare volatility on October 22, with the Nasdaq closing down approximately 2.6%, the Philadelphia Semiconductor Index plunging about 6.7%, Meta (META.US) shares dropping around 7%, and the Dow Industrial Average remaining nearly flat. At first glance, this appears to be a retreat from crowded AI trades, but the underlying pricing logic has fundamentally shifted: over the past two years, the market focused on "how many more chips need to be bought"; now, the focus has turned to "who will pay for AI."
This shift was triggered by Meta's earnings report released the previous day, causing market attention to quickly move from demand to financing.
Meta’s third-quarter financial results showed an adjusted EPS of $7.40, excluding $9.89 billion in legal expenses, compared to a GAAP EPS of just $3.83. However, free cash flow turned negative at $2.12 billion, down from a positive $10.63 billion in the same period last year—the first quarterly net outflow in over a decade. Capital expenditure guidance for 2026 has been revised to $135–145 billion, with the lower end raised from the previous $13 billion. The board approved a 1:1 distribution of non-voting Class C shares in January 2027. Capital expenditures for the quarter totaled approximately $35 billion, with advertising revenue growth insufficient to cover server and data center costs. The Class C share structure is designed to preserve Zuckerberg’s control while preserving future fundraising flexibility. Market selling pressure is not directed at the stock split itself, but rather at concerns over potential equity supply. Valuation is based on assumptions of perpetual cash flow growth; changes in financing methods directly impact the discount rate, not merely quarterly earnings.
Woofun AI’s aggregated data shows significant disagreement between upstream and downstream sectors regarding the causes of the downturn. NVIDIA (NVDA.US) CEO Jensen Huang stated in a CNBC interview that demand for neocloud is “exploding,” with the current bottleneck being working capital—capital demand far exceeds supply, but “this issue will be resolved.” His remarks pointed to liquidity constraints rather than disappearing orders, yet the market interpreted them as evidence of financing bottlenecks. If upstream suppliers acknowledge that customers lack funds, chip order timelines will depend on financing progress rather than demand strength.
On the other hand, Bernstein (AB.US) analyst Madison Rezaei lowered CoreWeave’s (CRWV.US) price target from $74 to $51, maintained a Underperform rating, and estimated that fulfilling its committed orders would require approximately $85 billion in financing. Market reactions were more immediate: CoreWeave’s bond yields rose to around 15.8%, and its five-year credit default swap spread exceeded 1,100 basis points. When borrowing costs surpass 15%, the pace of GPU rental business expansion is determined by creditor credit limits rather than market demand.
Jensen Huang believes demand is secure and funding issues are solvable, while Bernstein views rising funding costs as a signal of weakening effective demand. On October 22, the market showed its first large-scale shift toward the latter view. Chip stocks declined more than cloud providers, with NVIDIA down about 4% and CoreWeave down approximately 10%. Lam Research (LRCX.US) intensified sentiment volatility: although its guidance for the December quarter midpoint slightly exceeded analyst consensus, it fell short of buy-side private expectations, while also warning that a decline in revenue from Chinese customers and supply chain bottlenecks could lead to shipment delays.
Despite Lam’s emphasis on strong AI demand, global wafer fab equipment spending expected to exceed $200 billion by 2027, and record-breaking gross and operating margins for the quarter, the stock fell by approximately 10%. Lam’s guidance can explain the pullback in equipment stocks themselves but not the broader sector’s 6.7% decline; however, as a specific case, it clearly illustrates how financing concerns have transmitted from cloud providers’ balance sheets to visibility into equipment orders. The market’s sensitive reaction to “marginal misses” confirms that valuations remain highly dependent on the slope of capital expenditures.
From a macro perspective, Meta has become the first visible example of a shift in capital structure. Morgan Stanley (MS.US) estimates that next year, hyperscale cloud providers and neo-cloud companies may spend up to $1.58 trillion on capital expenditures, with a free cash flow gap—including leases—potentially reaching $540 billion. Goldman Sachs (GS.US) estimates a lower figure, projecting that top cloud providers’ capital expenditures will total approximately $800 billion by 2026. Although the two investment banks’ estimates differ by nearly a factor of two, both point to the same conclusion: the scale of AI infrastructure construction has surpassed the capacity of any single company’s internal cash flow to fund it. The funding structure is shifting from internal cash flow to external capital, and the market is beginning to price in dilution and spread risks.
Current evidence does not support the notion that "AI demand has peaked" or that "the funding window has closed." Although CoreWeave’s bonds are expensive, buyers still exist, and Meta’s Class C shares have only reserved a channel without immediate issuance. The key observation lies in when "expected dilution" transforms into "realized dilution." If leading companies actually issue large amounts of equity or new debt, valuation pressure will shift from discounted assumptions to actual equity supply. If credit spreads in the tech sector continue to widen, companies beyond neo-cloud will also be required to offer higher risk premiums, and financing constraints will no longer be isolated cases. Conversely, if capital expenditures continue to be revised upward next quarter and order revenue follows suit, this correction is more likely a temporary repricing. What remains from October 22 is not a conclusion of "AI’s end," but a new equation: demand persists, but behind every GPU, the question of funding source must first be answered.