Nvidia Shareholders Should Brace Themselves for 1 Particular Thing in the Months to Come. Here’s What It Means for the Long-term Picture.

Nvidia reported $96B+ revenue and $59B net income in the latest quarter, supported by continued AI GPU demand and the start of production shipments for the Vera Rubin platform (including its first stand-alone CPU). Management guided to gross margin compression to 71%-72% in Q4 from 75% in Q2, citing “extreme pricing conditions in memory,” though gross margin is expected to stabilize around 72%-73% in FY2028. Nvidia also projects $20B in stand-alone CPU sales this year and expects 70% revenue growth in FY2028, with the net takeaway being strong growth offset by near-term profitability pressure.

Analysis

The key mechanism is not “lower margins” in isolation, but a transfer of economic rent from compute vendors to memory suppliers. If AI buildouts stay tight on HBM/GDDR supply, the first beneficiaries are MU and the memory complex; the second-order loser is any accelerator vendor whose system BOM keeps inflating faster than its pricing power. That also pressures hyperscaler ROI math, which can slow incremental capex at the margin and indirectly favor companies that sell picks-and-shovels into the buildout rather than the branded system itself.

For NVDA, the stock only breaks badly if investors decide the margin glidepath is structural rather than temporary. Near term, the market can tolerate low-70s gross margin as long as unit growth stays extreme; the multiple risk shows up if guide quality weakens or if management is forced to choose between share and profitability. Over 6-18 months, a steady rise in memory capacity is the main falsifier of the “scarcity tax,” but if AI demand keeps outrunning supply, the pricing environment can remain tight longer than consensus expects.

The contrarian read is that the margin compression may actually validate demand strength: a company with dominant share and still-elite profitability can absorb component inflation better than peers. That argues against an outright structural short on NVDA, but it does support relative-value expressions where the beneficiary is the upstream input rather than the platform leader. In other words, the trade is not “AI is over,” it is “who captures the spread from AI scarcity?”

Key Decisions for Investors

• Long MU / short NVDA as a 1-3 month relative-value trade into the next round of AI memory pricing data and NVDA commentary; thesis works if memory tightness keeps EPS revisions for MU moving up while NVDA gross margin estimates drift lower. Falsify if NVDA re-accelerates margin guidance above low-70s or MU preannounces weaker pricing.
• Buy a 1-2 month NVDA put spread on strength ahead of the next earnings/guidance window if the stock continues to price in flawless revenue growth; use defined risk because the downside is mostly multiple compression, not business collapse. Falsify above a sustained reclaim of prior highs with margin guidance unchanged or better.
• For a cleaner hedge on AI capex, pair long MU with short a basket of NVDA-call-delta exposure via SMH or SOXX only if you want semiconductor beta neutralized; this isolates the memory rent-transfer theme. Exit if memory lead times normalize faster than expected.
• If you are already long NVDA structurally, consider a collar into the next catalyst: sell upside financed by downside protection. This is the right posture if the fund believes revenue keeps compounding but margin volatility creates event risk rather than thesis risk.