Matt Bloxham 3.1 6 ideas

US proposing new rules requiring licenses to export AI chips abroad. While demand is high, regulatory friction is increasing. The US wants "control" over AI proliferation. This creates a bureaucratic ceiling on how fast these companies can ship to non-US markets. WATCH. Fundamental demand is strong (see Marvell), but regulatory headlines create volatility. Strict bans on sales to key high-growth regions.

Hyperscalers (Amazon, Microsoft, Google) announced ~$650B in CapEx plans for 2026 a few weeks ago. Nvidia's forward guide points to 80% revenue growth. The massive, committed capital expenditure from Hyperscalers guarantees near-term order flow for Nvidia's chips, regardless of immediate enterprise adoption. The "infrastructure build" phase is fully funded. LONG. The cash flow is locked in via Hyperscaler budgets. If enterprise demand for "scaled AI" (not just pilots) doesn't materialize, Hyperscalers may cut CapEx in 2027.

Meta is deepening its relationship with Nvidia, and Meta was Nvidia's second-largest customer last year. Despite fears of hyperscalers (like Meta) building their own chips to replace Nvidia, the deepening partnership signals that Nvidia's GPUs remain indispensable for the immediate future of AI model training. Long NVDA (continued demand) and META (securing necessary compute). Regulatory crackdowns on Big Tech capex; diminishing returns on AI spend.