Prediction: Sept. 10 Will Be a Big Day for Nvidia Shareholders. Here’s What to Watch.

Nvidia reported fiscal Q2 2027 revenue of $96 billion, up 106% year over year, and profit of $59 billion, up 126%, with data-center revenue contributing $89 billion. At its Sept. 10 Goldman Sachs conference presentation, investors will focus on the Vera Rubin rollout—expected to generate roughly 20% of Q3 data-center revenue—and on potential CPU-market forecasts. Nvidia projects 70% revenue growth in fiscal 2028, though constrained memory supply, rising memory prices and unresolved China sales remain material headwinds.

Analysis

The relevant question for NVDA is not whether demand remains strong, but whether the Rubin ramp preserves the company’s system-level pricing and gross-margin architecture as memory costs rise. A faster transition to integrated CPU-GPU systems can pressure AMD and INTC’s standalone CPU economics by making the CPU a lower-value attach component, but it also raises execution risk: any rack-level qualification, networking, or HBM availability issue delays recognition of a much larger revenue unit than a chip-only sale. MU is the cleaner second-order beneficiary if HBM tightness persists, although its upside is increasingly contingent on supply discipline rather than incremental AI demand alone.

The conference itself is unlikely to create durable upside absent a quantifiable change in shipment cadence, supply availability, or forward gross-margin commentary; optimistic product rhetoric is already consensus. The near-term asymmetry is negative if management concedes that memory inflation is being absorbed rather than passed through, because NVDA’s premium multiple depends on confidence that platform mix offsets component cost pressure. Over the next 1-3 months, the key catalyst is customer acceptance and production scaling of Rubin systems; over 6-18 months, the more material issue is whether custom silicon from hyperscalers captures lower-end inference workloads and limits NVDA’s incremental share gains.

Contrarian view: investor rotation into AMD and INTC may be treating NVDA’s CPU adjacency as a direct revenue transfer, when customers may instead retain heterogeneous CPU fleets and buy NVDA systems only for accelerated clusters. That makes a broad short of AMD or INTC premature. The better relative expression is to own the component bottleneck beneficiary against the system vendor only if evidence emerges that memory pricing is outpacing NVDA’s realized system ASPs.

Key Decisions for Investors

• Maintain NVDA core exposure into the event only with a defined hedge: buy 1-2 month NVDA put spreads financed by selling upside calls above the pre-event implied move. The risk case is a supply or gross-margin reset rather than a demand reset; remove the hedge if management quantifies stable gross margin despite higher memory costs.
• Initiate a 1-3 month long MU / short SOXX relative-value position on evidence of continued HBM allocation tightness or upward memory contract pricing. Thesis fails if MU signals capacity additions sufficient to normalize HBM pricing faster than demand growth, or if NVDA indicates it has secured supply at fixed pricing.
• Do not chase AMD or INTC solely on CPU-incumbent displacement headlines. Establish an alert for disclosed customer CPU attach rates, Rubin system qualification delays, or AMD/INTC data-center guidance cuts; only then consider a 3-6 month short AMD or INTC versus long NVDA.
• For NVDA, treat any post-event rally without raised shipment or supply visibility as a trim opportunity rather than a breakout signal. Add only after independently corroborated evidence of volume system deployments, with gross-margin guidance holding despite HBM cost inflation.