Nvidia Reports Triple-Digit Revenue Growth; Now Supply Is The Constraint

Nvidia reported Q2 revenue of $96.2B, up 106% year over year, marking a return to triple-digit growth as supply constraints ease back into focus (memory, advanced packaging, and power). Management emphasized that the broader AI supply chain remains constrained, with Edge Computing still only 7.5% of revenue today. The company reframed its platform narrative around agentic AI and continuous inferencing, supporting a bullish read-through for demand visibility.

Analysis

The market should treat this less as a simple NVDA print and more as a signal that the AI buildout has moved from chip design scarcity to infrastructure scarcity. When the bottleneck shifts into HBM, advanced packaging, and power delivery, the incremental dollar of demand migrates down the stack: memory vendors, foundry packaging capacity, and thermal/electrical infrastructure can see faster earnings translation than the headline platform name. That also means NVDA’s incremental upside is increasingly gated by suppliers, not by end-demand, so the next leg of outperformance may be in bottleneck enablers rather than the leader itself.

Agentic inference changes the cadence of demand: it is less spiky than training and more like a recurring utility load, which should lengthen the cycle for memory and networking spend. In the next 1-3 months, the key catalyst is whether customers validate that inference tokens are scaling fast enough to justify sustained capex rather than just reorder timing. If that shows up in supplier lead times or HBM pricing, the easiest relative-value expression is long the bottlenecks, not long the obvious winner.

The contrarian risk is that the market may still be underestimating how much of NVDA’s demand is already pre-committed, which turns supply tightness into deferred revenue rather than incremental upside. For NVDA, the question is less "can it sell?" and more "can it convert demand into shipments fast enough to justify a further multiple re-rate?" If packaging capacity, memory supply, or power delivery improve faster than expected, the scarcity premium can unwind quickly and rotate back to the broader AI complex.

Key Decisions for Investors

• Prefer a relative-value long in MU or TSM over outright adding NVDA here: the bottleneck suppliers have clearer earnings sensitivity to the current constraint and may capture the next 1-3 quarters of upside if AI inference demand is real.
• Use NVDA strength to fade marginal upside with a small tactical short against a basket long of AI bottlenecks (MU/TSM/AMAT) if you want exposure to the cycle but believe the leader is now supply-capped rather than demand-capped.
• Set a watch item on HBM pricing and CoWoS lead times; if either tightens again over the next quarter, it supports a further leg in memory and packaging names, but if they normalize, the supply-scarcity thesis is weakened.
• Look at SMH/SOXX as a hedgeable way to stay long the AI capex theme while reducing single-name execution risk in NVDA; the best risk/reward may be in the basket, not the platform leader.
• Falsifier: if NVDA guides to materially faster supply ramp or inventory/lead-time metrics show the constraint easing within 1-2 quarters, reduce exposure to bottleneck beneficiaries and reassess the scarcity premium across the AI supply chain.