Jensen Huang's AI Capex Pulse Check

Nvidia CEO Jensen Huang projected that the five largest hyperscalers will lift AI capex to $1.3T next year from roughly $800B this year, implying about 60% year-over-year growth despite rising concerns over negative free cash flow and circular financing. Marvell raised its growth outlook to 50% from 45%, while CrowdStrike posted record net-new ARR growth of 51% and forecast ARR growth of roughly 41%, reinforcing an AI-driven cybersecurity spending tailwind. Offsetting the bullish AI narrative, Dick's Sporting Goods fell more than 30% in its worst trading day as Foot Locker weakness, heavier discounting, and deteriorating athletic-footwear demand pressured margins and guidance.

Analysis

AI infrastructure is shifting from a compute bottleneck to a networking, power-delivery, and construction bottleneck. That broadens the earnings pool beyond NVDA/TSM toward MRVL, AVGO, FPS, PSIX, and STRL, but the latter two carry materially higher fixed-cost and backlog-cancellation risk when hyperscaler build schedules flatten. The key near-term read-through is not announced capex, but whether cloud providers sustain lease commitments and whether supplier lead times remain extended through 1H27.

MRVL's customer-warrant structure creates a non-obvious trade-off: it can secure multi-year demand and displace AVGO in custom silicon/interconnect, but economic value is diluted if share grants vest faster than incremental gross profit accrues. The market should reward verified design-win conversion and gross-margin expansion, not addressable-market narratives. Favor MRVL only if the next two reports show AI revenue growth outpacing dilution-adjusted share count and no concentration-driven guidance haircut.

Cybersecurity is becoming more bifurcated: AI-driven threats increase security budgets, but buyers are consolidating vendors rather than funding every endpoint platform. CRWD's bundle-led sales motion can compound operating leverage, while S and OKTA face a higher bar to prove that product growth translates into durable net retention and positive FCF. CRWD's premium multiple leaves it vulnerable to any booking deceleration or renewal friction; the better expression is relative, not outright.

The DKS drawdown may ultimately create value, but the market is now underwriting a longer integration and promotional-margin cycle. A recovery depends less on core comps than on Foot Locker inventory turns, markdown rates, and evidence that procurement synergies offset incremental store rationalization. Avoid treating a low headline earnings multiple as a margin of safety until those metrics stabilize over two quarters.

Key Decisions for Investors

• Initiate a 3-6 month long MRVL / short AVGO pair, sized beta-neutral. Target 15-20% relative upside if custom interconnect and ASIC wins convert; exit if MRVL AI revenue growth decelerates for two consecutive quarters or dilution exceeds management's implied framework.
• Own CRWD versus S as a 6-12 month quality-consolidation pair rather than chase CRWD outright. Upside is further ARR/FCF divergence; stop if CRWD net-new ARR growth falls below 30% or S demonstrates sustained positive FCF with accelerating enterprise retention.
• Build a staged long FPS or PSIX position only after confirming backlog quality, customer concentration, and working-capital conversion at the next earnings release. These are higher-beta beneficiaries of power scarcity; use a 15% downside stop because project-delay risk can re-rate multiples quickly.
• Keep DKS on a 1-3 month watch list, not a fresh long. Enter only after Foot Locker gross margin and inventory turns improve sequentially; a 12x forward multiple is attractive only if normalized operating margin stops compressing. Failure to stabilize within two earnings prints supports a short versus specialty retail peers.