Nvidia by the Numbers: 3 Remarkable Facts in Its Q2 Report

Nvidia reported Q2 revenue of $96.2B (+106% YoY) versus the $92.2B consensus, with gross margin expanding to 75% (from 72.4%) and adjusted EPS rising 120% to $2.22 (vs. $2.09). CEO Jensen Huang guided for at least 70% revenue growth next year, far above Wall Street’s ~45% model, helping reverse early selling and lift the stock about 4.4% after hours. Data center revenue jumped 117% to $89B, reinforcing continued AI GPU demand and the risk being more about whether the AI boom sustains than near-term execution.

Analysis

This print matters less as a one-day beat and more as a proof point that AI capex is still supply-constrained at hyperscale. That keeps the near-term revenue pool intact for MSFT, AMZN, and GOOGL, but it also means their cash flow statements will keep showing heavier depreciation and capex before monetization catches up. In other words, the immediate winner is NVDA, but the second-order winners over the next 1-3 quarters are the infrastructure bottlenecks that turn chips into usable clusters.

The market is likely underpricing how quickly the scarcity premium can migrate away from NVDA itself. Once supply starts to catch up, the multiple won’t be anchored by growth alone; it will be anchored by how much of the AI stack remains proprietary versus commoditized. That argues for relative-value exposure to the broader AI buildout rather than an unhedged chase in the largest name, especially as networking, advanced packaging, memory, and power/cooling vendors absorb more of the incremental spend.

The main falsifier is a pause in hyperscaler spend, not a single quarter miss. Watch for commentary on capex moderation, backlog conversion, or gross margin slip back into the low-70s; that would signal the growth runway is becoming more normal and the rerating should stall. Over 6-18 months, the real structural risk is custom silicon and in-house accelerators, but those only matter if performance parity becomes credible and software portability improves enough to reduce switching costs.

Key Decisions for Investors

• Do not chase NVDA after the print; wait for a 3-5% post-earnings fade to add. Risk/reward is better on pullbacks because the market has already priced in a near-flawless guide.
• Pair trade: long NVDA / short a basket of lower-quality AI narrative names or high-multiple software that still lacks clear AI monetization. Horizon: 1-3 months; thesis is that capex spend remains real while application-layer revenue proof remains thin.
• Accumulate MSFT and AMZN on dips as second-order AI infrastructure beneficiaries, but treat them as capex-duration trades rather than immediate margin expansion stories. Falsify if cloud capex slows or operating margin guidance tightens.
• Watch semis infrastructure exposure rather than just GPU beta: consider adding any weakness in networking, memory, or power/cooling suppliers tied to AI cluster buildouts. These names can lag the headline but often capture the next leg of orders.
• Set an alert on NVDA growth guide: if forward growth drops materially below the current ~70% narrative or margin trends reverse, reduce exposure; that would signal the scarcity trade is maturing faster than the market expects.