- Nvidia reported Q2 revenue of $96.2 billion — beating the $92.5 billion analyst consensus — and guided for $108 billion in the current quarter (plus or minus 2%), nearly double a year ago; its fiscal 2028 revenue growth forecast of approximately 70% obliterated the 45% analyst estimate, with CFO Colette Kress saying: “Incredibly, we are seeing demand acceleration even at our scale. Customers’ forecasts point to our growth doubling next year” — and that Nvidia “would grow faster if it had access to more supplies.”
- Nvidia’s all-important data center segment posted $89 billion in Q2 revenue versus an $85.8 billion estimate, with hyperscalers including Amazon and Alphabet’s Google accounting for the majority of those sales; Vera Rubin, the company’s latest chip architecture, is now “in full production,” and CEO Jensen Huang declared that “the AI infrastructure build-out is at full steam” — a direct rebuttal to investors who had been watching for signs of demand deceleration.
- The margin picture is more nuanced: gross margin will be approximately 74% this quarter but is expected to narrow to 71-72% by Q4 (January) as surging memory costs — driven by a DRAM and HBM shortage that Micron, SK Hynix, and Samsung don’t expect to resolve for years — compress profitability; Nvidia has already notified customers of price hikes above 15% on some products, which Kress says should allow margins to settle back to 72-73% in fiscal 2028.
- Three structural headwinds are worth watching alongside the blowout results: (1) a stampede of would-be competitors including customer-developed in-house chips — OpenAI this week claimed its new Jalapeno processor outperforms Nvidia’s current lineup in tests; (2) circular financing concerns, as Nvidia’s investment pacts with AI companies have put it on the hook for tens of billions in commitments that critics argue foster artificial demand; and (3) China market access, where Washington’s limited authorization to sell H200 chips is complicated by Beijing’s own controls on purchases.
What Happened?
Nvidia reported Q2 results (quarter ended July 26) that beat on every major metric: revenue of $96.2 billion versus $92.5 billion estimated, EPS of $2.22 versus $2.09 estimated, data center revenue of $89 billion versus $85.8 billion estimated. The company guided for $108 billion in Q3 revenue and forecast approximately 70% sales growth for fiscal 2028 — roughly double the 45% analyst consensus. CFO Kress said demand is accelerating and that supply, not demand, is the binding constraint. CEO Jensen Huang touted the rollout of Vera Rubin, now in full production, as purpose-built for “this moment.” The stock surged 7.4% in pre-market trading and was on pace to add approximately $370 billion in market capitalization. Nvidia has now beaten Wall Street estimates for 16 consecutive quarters.
Why It Matters?
Nvidia’s quarterly earnings have become the most important single data point for the AI investment thesis — functioning, as Bloomberg puts it, as “a barometer on the state of the wider industry.” A 70% growth forecast for fiscal 2028, combined with a CFO saying customers’ own demand forecasts imply growth doubling, is the most direct refutation available of the AI bubble thesis. When the company supplying the GPU infrastructure says it cannot keep up with demand, the question of whether hyperscaler AI spending is real or circular is answered empirically rather than analytically. The margin compression from memory costs is the most important nuance: rising DRAM and HBM prices are structural, not cyclical, and Nvidia’s ability to pass those costs through to customers (via 15%+ price hikes) while maintaining demand demonstrates pricing power that very few companies in any industry possess. The OpenAI Jalapeno chip claim is the long-term competitive risk worth monitoring — not because it threatens Nvidia today, but because it establishes the pattern by which the largest customers may eventually reduce their Nvidia dependency.
What’s Next?
The margin trough in Q4 (January) is the most watched near-term financial metric: if memory costs continue rising faster than Nvidia’s price increases, the 71-72% floor could prove optimistic, and any miss would give renewed ammunition to bubble-thesis investors. Watch for follow-on procurement announcements: Amazon has separately confirmed purchasing 2 million Nvidia chips for its data center buildout, and further hyperscaler procurement deals will serve as confirmation that the Q2 demand levels are sustainable. The China H200 situation is the geopolitical wildcard — Nvidia’s ability to expand sales there represents meaningful upside that is currently underpriced in estimates, but any escalation of US-China semiconductor restrictions could reverse that access rapidly. Jackson Hole’s macro implications for growth multiple valuations remain the near-term risk overlay for the entire sector.
Source: Bloomberg
