Nvidia Q2 Highlights: Double Beat, Huang Says ‘AI Infrastructure Buildout is at Full Steam'

Nvidia reported Q2 revenue of $96.22B (+106% YoY), beating the $92.18B consensus, alongside EPS of $2.22 vs $2.10 expected and gross margin of 75.0%. The company guided Q3 revenue to $105.84B–$110.16B (+85.7% to +93.2% YoY), below the Street’s $104.20B midpoint, while noting it does not assume China Data Center compute revenue and China Hopper shipments were <1% of Data Center revenue. Nvidia also had $99.0B remaining under its buyback authorization, supporting a bullish read-through even as shares were down ~1.8% after hours.

Analysis

The market’s real question is no longer whether NVDA is winning share; it is whether the growth rate is still high enough to justify a premium multiple once AI spend shifts from novelty to budgeting discipline. Margin stability is the important tell: if gross margin holds while growth decelerates, earnings power remains intact, but the stock can still compress because investors are paying for scarcity, not just profits.

Relative winners are the GPU-consuming ecosystems that can monetize inference and hosting, especially MSFT, GOOGL, and ORCL. The full-production ramp implies the AI infrastructure stack is moving from pilot to deployment, which should favor cloud capacity owners and GPU lessors with pricing power; second-order, that can also squeeze smaller software vendors that still lack clear token-level monetization and may see AI capex compete with other spend.

The contrarian risk is that the consensus is treating the slowdown in growth rate as a warning sign when it may just be comping against an extreme base and excluding China. The bigger 6-18 month threat is custom silicon and efficiency gains reducing incremental GPU demand, not a near-term demand collapse. Near term, the stock likely trades off hyperscaler capex commentary over the next 4-8 weeks; if MSFT/GOOGL keep raising AI spend, pullbacks should be bought, but if cloud budgets flatten, multiple compression can outrun earnings growth.

Key Decisions for Investors

• Buy NVDA only on post-earnings weakness, preferably via a 3-6 month call spread rather than stock; use a pullback toward the low-$200s as entry and exit if the stock loses the prior breakout area or if the next hyperscaler capex guide softens.
• Initiate a relative-value long ORCL / short NVDA pair for 1-3 months if you want AI exposure with less single-name execution risk; ORCL has more direct operating leverage to GPU hosting demand while NVDA is more exposed to any multiple reset from growth deceleration.
• Overweight MSFT and GOOGL on any AI-capex follow-through, but treat them as the better diversified beneficiaries, not pure AI beta; thesis fails if their next guidance shows AI spend growth slowing or margin dilution from infrastructure buildout.
• Set an alert on NVDA around the next 4-8 week hyperscaler earnings cycle; if cloud capex keeps accelerating, cover any short exposure immediately, because the market will re-rate the whole AI supply chain higher.