The Silicon Pivot: Cathie Wood’s $18.7 Million Rebalance and What It Says About AI Capex Risk

(SeaPRwire) –   By: Reginald Vance

When OpenAI’s revenue runs twenty billion dollars below earlier estimates, the entire AI hardware capex thesis shakes. Cathie Wood sold twenty two thousand five hundred AMD shares valued at nearly fourteen million dollars on Thursday, October 8. AMD shares dropped almost four percent that same day. Bond yields climbed. Oil prices pushed higher. Yet AMD still sits up about 190 percent year-to-date. The market is pricing fear into the tape while the underlying supply chain keeps signing multi-year chip commitments. Wood’s sell signal reads as a bet that the margin of safety in speculative AI compute infrastructure has thinned beyond comfortable levels. She also unloaded nearly four point seven million dollars in SpaceX shares that same session. This was not isolated selling. The day before, ARK had already sold fifty four thousand eight hundred seventy three SpaceX shares. SpaceX stock fell more than four percent after new regulatory obstacles for Starlink internet service in India emerged. The company had already built much of the infrastructure needed to launch there. Spectrum allocation and other approvals still block operations. Yet after the market closed, SpaceX announced an agreement to acquire a nationwide low-band spectrum portfolio from Grain Management. The deal covers the 800 MHz band across the country. It could let Starlink Mobile operate alongside traditional cell networks. This puts SpaceX in direct competition with AT&T, Verizon, and T-Mobile. Combined sales across both positions totaled about eighteen point seven million dollars for the single day. That is a significant capital reallocation signal from one of Wall Street’s most aggressive growth investors. The underlying panic is not about chip demand. It is about whether hyperscaler revenue can justify the compute infrastructure being built today.

AMD’s pipeline of chip supply deals continues to accelerate despite the stock pressure. The partnership with Anthropic includes plans to supply up to two gigawatts of chips starting in 2027. Oracle also plans to use AMD chips in its cloud data centers. Thirty analysts out of thirty-six covering AMD recommend buying the stock. The average price target sits at six hundred sixty-two dollars and thirty eight cents. That represents about seven percent upside from Thursday’s closing price. SpaceX tells a parallel story of institutional backing. Out of thirty-four analysts, twenty-seven recommend buying the stock. Five hold neutral ratings. Two recommend selling. The average price target for SpaceX reaches two hundred fifteen dollars and seventy two cents. That is about thirty-four percent above Thursday’s closing price. The SpaceX spectrum acquisition from Grain Management could unlock mobile revenue streams that do not depend on satellite broadband subscribers alone. Starlink Mobile would leverage the second-generation constellation to ensure Americans have access to connectivity in underserved areas. The agreement still requires regulatory approval before SpaceX can move forward. Meanwhile, OpenAI’s revenue shortfall casts a shadow over the entire AI infrastructure buildout. Hyperscalers may rein in chip spending if revenue projections continue to disappoint. This is precisely the kind of structural headwind that makes AMD’s forward-looking supply agreements look either prescient or risky. Either way, the volume of commitments signals something important. Companies building frontier AI models are not walking away from the silicon shortage. They are pre-positioning to grab more of it when the market corrects.

Kratos Defense stock has had a brutal run. Shares are down forty-five percent in 2026. Investors raised concerns about the company’s spending levels. They questioned how long unmanned aircraft and hypersonic weapons programs would take to pay off. The second-quarter revenue number tells a materially different story. Revenue grew thirty point five percent year-over-year to four hundred fifty-eight million eight hundred thousand dollars. Management raised the full-year growth forecast to a range between eighteen and twenty-three percent. Sixteen out of seventeen analysts covering Kratos recommend buying the stock. The average price target is ninety-five dollars. That suggests about one hundred twenty-six percent upside from Thursday’s closing price. Wood bought one hundred fifty nine thousand nine hundred and eighteen shares of Kratos for about six million seven hundred thousand dollars. This was not her first Kratos purchase. She has been adding to the position recently. The signal here is unmistakable. Capital is rotating away from compute infrastructure narratives that depend on unproven AI revenue. It is moving toward defense hardware orders backed by sovereign government procurement cycles. Defense budgets do not contract when AI startups miss revenue targets. The Kratos bet assumes that defense spending is counter-cyclical to speculative technology infrastructure spending. The hardware vendor consolidation endgame belongs to whoever can tie silicon output to sovereign budgets rather than venture backstops. Wood is making that bet now, while the stock trades at a steep discount and the analyst community overwhelmingly agrees there is substantial upside ahead. The practical takeaway is clear. In a market where AI capex assumptions are being tested, defense procurement contracts provide a more predictable cash flow stream.

Author bio: Reginald Vance, a venture partner specializing in semiconductor valuation and advanced materials, with over a decade tracking capital flows across the hardware supply chain and frontier defense technology investments.