A friend messaged me asking whether I’ve already "made a killing and kept it low-key." Not pretending—people tracking me on other platforms probably know this already. This round, I’m going 4x leveraged long on Nvidia, betting on the earnings report. How much I’m up today—you all can do the math yourselves. If I share the numbers, it would sound like I’m showing off.
Once the earnings report is out, I’ll listen to the call—I'll share five points that I find interesting.
1. Jensen said firsthand: AI has reached a turning point on the business side.
All companies' AI business models in the market are rapidly closing the loop. Tokens are shifting from "spending" to "cash flows and assets." AI is truly starting to generate revenue. There’s supporting evidence for this: Google’s paid user growth for last quarter’s Gemini was up 40%. This isn’t just a concept—it’s beginning to monetize.
2. Demand hasn’t stopped; it’s actually accelerating.
At this time last year, it was basically a laboratory driving infrastructure. Today, we’ve entered the "golden era of new AI labs and startups." Multiple cutting-edge labs are expanding in parallel, and open-source model ecosystems are thriving. Especially the growth momentum of physical AI has even surprised Jensen himself.
3. Vera Rubin is ramping into full production—scaling the revenue opportunity per GW to $40 billion.
During the meeting, Huang confirmed that the new-generation Vera Rubin platform is in a phase of full-scale production ramp-up, calling it the product platform Nvidia has pushed forward fastest in its history. As the Rubin ecosystem (Vera CPU + Rubin GPU + NVLink + networking system) rolls out, the system revenue opportunity per GW continues to rise: $18 billion for Hopper → $25 billion for Blackwell → $40 billion for Rubin.
What’s even more aggressive is that management provided rare forward guidance: the FY2028 growth anchor is set at around 70%. But note—70% is not the ceiling of the AI market; it’s the ceiling based on Nvidia’s ability to match upstream storage vendors’ production capacity. Demand is at the ceiling; supply is on the floor.
4. Gross margin hits the bottom—then keeps lifting over the next few quarters. And the pressure is coming precisely from storage.
Storage components are experiencing industry-wide shortages because AI infrastructure is expanding at a frenzy, driving up unit costs. There will be some mild, phase-specific pressure in Q3/Q4, and we expect gross margin to bottom in the 71%–72% range. As the pricing mechanism gets sorted out and the transition to new platforms progresses, the impact can be absorbed steadily.
The subtext behind this is crucial: storage OEMs like SK Hynix, Samsung, and Micron still hold very strong pricing power even when dealing with a top-tier giant like Nvidia. Their capacity is tightly bound to HBM and high-bandwidth storage due to the AI wave, which directly translates into storage OEMs’ striking gross margins and earnings elasticity. Moreover, downstream consumption of high-end storage shows no slowdown whatsoever. With Vera Rubin ramping into production, requirements for HBM capacity and number of layers (from HBM3e to higher tiers) are rising geometrically—high unit price and high shipment volume on both ends. They get both.
5. CPU shipments surge, and Vera CPU becomes an independent growth driver.
The previous-generation Grace CPU’s cumulative revenue over the past 12 months has already exceeded $5 billion. After the new-generation Vera CPU fully ramps into production, management expects that FY2028 CPU-related standalone and bundled revenues will grow by more than double year over year. They also open up a brand-new TAM as large as $200 billion, breaking the previous pattern that heavily relied on the combination of "GPU + traditional x86 CPU."
Why is this point important? Nvidia’s CPU performance essentially reflects enterprises’ AI-driven demand for improved long-dialogue modes. The core evidence of the surge in enterprise demand is CPU volume—because CPUs are key for enterprise AI long-form Q&A and Agent development, and they’re also the fundamental barometer of overall demand and the market.
To wrap it up: in this Nvidia earnings report, all five lines—"AI business loop + demand acceleration + supply constraints + storage pricing power + enterprise CPU volume"—have been confirmed at once. As for leverage, making money is a combination of skill and luck. Don’t use my position as your template.