Nvidia Just Obliterated The Bear Case, And Shares Are Still Undervalued

Nvidia reported record Q2 FY2027 results, with revenue up 105.85% YoY, and guided to 70% growth in FY2028 while pointing to ongoing supply constraints. The company is also running its capital return program ahead of plan, returning 60% of FCF YTD through buybacks/dividends, with $99B in remaining buyback authorization. Data center diversification improved (non-hyperscale customers now 45% of the segment) and management expects Vera Rubin to drive the fastest product ramp in NVDA history.

Analysis

Near term, this is less a demand story than a scarcity story: when compute is the bottleneck, value accrues to the tightest parts of the stack and to vendors with the best allocation. That argues for continued strength in advanced packaging, HBM, networking, and system integrators with supply access, while smaller rivals and custom-silicon efforts remain at a disadvantage because they cannot match the ecosystem depth or shipment cadence.

The broader customer mix matters more than the headline growth rate. A wider base beyond the hyperscalers reduces the risk of a single buyer pausing capex, which should stabilize the AI infrastructure cycle over the next 1-3 quarters. The flip side is that supply-constrained growth can mask a future slowdown: when capacity normalizes, revenue growth can decelerate sharply even if end demand stays healthy, and that is when the multiple is most exposed.

Contrarianly, the market may be underpricing how much of the equity story depends on continued execution across a few checkpoints: hyperscaler capex, export policy, and whether alternative accelerators gain software traction. Capital returns support EPS and help on pullbacks, but they do not neutralize a de-rating if growth steps down from extraordinary to merely excellent. The thesis is falsified first by softer next-quarter data-center momentum or a margin guide down, and over 6-18 months by any sustained evidence that supply catches up faster than demand.

Key Decisions for Investors

• Buy NVDA only on post-print consolidation or a 3-5% pullback; 1-3 month horizon. Best risk/reward is a reload after the first fade, with thesis invalidation on any guide cut or gross margin compression tied to product transition.
• Pair trade: long NVDA / short AMD for 1-3 months. Rationale: NVDA still owns the scarcity premium and ecosystem moat; the spread should widen if customers prioritize guaranteed supply over speculative share-gain stories. Cover if AMD lands a meaningful design-win/revenue inflection.
• Overweight TSM as a second-order beneficiary on a 6-18 month view. The real bottleneck is not just chips but packaging and advanced-node capacity, so the scarcity premium should spill into the foundry/tooling chain even if GPU multiples stall.
• Use a call-spread structure instead of chasing upside outright if entering after the event. This caps theta bleed while acknowledging valuation risk; exit if the stock rips but next hyperscaler capex commentary fails to confirm follow-through.
• Set a watch item on hyperscaler capex and export-control headlines. If a major cloud provider signals a pause or China restrictions tighten materially, trim NVDA exposure quickly because that is the fastest path to multiple compression.