Nvidia Q2: The $10 Trillion Thesis

Nvidia’s Q2 results beat expectations with $96.2B revenue and a 66% operating margin, and management guided Q3 revenue to $108B, indicating continued strong AI infrastructure demand. Profitability is still expanding with 66% operating margin and a valuation described as compelling at a low-20s P/E, alongside upside optionality from new platforms like Vera Rubin and Hyperion. Overall, the earnings beat plus robust guidance should support NVDA’s near-term momentum and reinforce its wide-moat position.

Analysis

The near-term setup is still favorable, but the market question is shifting from “can demand exist?” to “how much of the AI capex stack does NVDA capture before customers diversify?” In the next 1-3 months, the cleanest winners remain the adjacent picks-and-shovels: advanced packaging, HBM memory, networking, and power infrastructure. The second-order risk is that every upside beat encourages hyperscalers to accelerate custom silicon and inference-efficiency work, which can keep the overall AI spend pie growing while capping NVDA’s share and multiple.

Consensus is likely underestimating how quickly the bottleneck can move from GPUs to deployment absorptions: data center power, cooling, and rack-level integration. That supports a broader AI infrastructure basket, but it also means NVDA can remain “too good” for its own multiple once growth normalizes, even if revenue keeps rising. In 6-18 months, the main falsifier is not a demand miss but evidence that gross margin or operating leverage stalls as pricing power gets competed away by alternative architectures.

The contrarian view is that this is a quality growth story, not a free option: if the stock is already discounting several more quarters of exceptional execution, the risk/reward on outright longs becomes less attractive than relative-value exposure. The more interesting trade is to own the ecosystem where capacity additions are still underappreciated, while expressing skepticism that NVDA can keep expanding multiple and margins indefinitely. Watch for any guide-down in gross margin, commentary on customer concentration, or signs that lead times normalize faster than backlog converts.

Key Decisions for Investors

• Buy NVDA on 3-5% post-print pullbacks, but prefer a 1-3 month call spread over stock to limit multiple-compression risk; thesis breaks if gross margin or next-quarter guidance stops improving.
• Pair trade: long SOXX or SMH vs short a single-name AI hardware winner only if valuation re-rates further; if not, keep exposure diversified across the semiconductor stack rather than outright chasing NVDA.
• For a more contrarian expression, buy NVDA Jan-2026 call spreads instead of shares if you want upside participation with defined downside; this is the cleanest way to own continued AI spend without paying for perfection.
• Overweight beneficiaries of the buildout that are less exposed to architecture substitution risk over the next 6-18 months: AMAT, ASML, MU, ANET; these names capture the capex wave even if NVDA share of wallet moderates.