NVIDIA Just Reported the Biggest Quarter in Chip History — How to Check Your AI Exposure

NVIDIA reported $96.2 billion in revenue for Q2 FY2027 on August 26, up 18% from the prior quarter and up 106% from a year ago — the largest quarterly revenue number ever reported by a chipmaker. Data Center revenue was $89.0 billion, up 117% year over year, and both GAAP and non-GAAP gross margins held at 75%, a level that would have been unthinkable for a hardware company a few years ago. The company guided Q3 to roughly $108 billion. For anyone who has followed the AI trade this summer, the print is the confirmation that the story the market split into chapters is real.

The significance is not just the size of the number, though the size is absurd — it is what the number says about the AI demand curve. This is the first earnings call since the market spent August segmenting the AI trade into compute, memory, and embodiment. Compute showed up in NVIDIA's $89 billion Data Center line and in Microsoft's $30B Azure AI run-rate. Memory showed up in the high-bandwidth-memory shortage that lifted Micron past $1,000. Embodiment showed up in the Physical AI pipeline NVIDIA showcased at Actuate 26. NVIDIA just posted a quarter that makes the compute chapter undeniable and guided up, which is the strongest possible signal that the demand is front-loaded rather than fading.

What the print actually confirms

Three things, and all three matter for how you think about your AI exposure. First, the growth is accelerating, not decelerating: revenue up 106% year over year on top of a 62% year-over-year quarter a year earlier, and yet NVIDIA still guided Q3 higher to $108 billion. That is a demand curve that has not rolled over. Second, the center of gravity is the data center, not gaming or the consumer business — $89 billion of the $96.2 billion is Data Center, so the AI infrastructure buildout is driving essentially all of the growth. Third, the gross margin held at 75%, which means NVIDIA is not having to discount to move products; demand is outrunning supply, the same supply tightness that showed up in the memory chapter.

The through-line to the posts this blog has run all summer is that the AI trade keeps discovering new phases, and each one re-rates a different slice of the market. Compute, memory, embodiment. NVIDIA's print is the anchor for the compute chapter, and the reason the memory and embodiment chapters matter is that they are the same demand showing up in different ledgers. When the whole complex moves together — as it has through August — it feels like one trade. The lesson of this print is that it is not: a memory bust and an embodiment stumble are different risk events, and a portfolio that blends them into one "AI" line is the one that gets caught flat-footed when they diverge.

How to audit your AI exposure in one sheet

The first step is seeing what you actually hold across every brokerage, instead of logging into each one separately and adding up the same ticker by hand. That is exactly what InvestSheet removes: it syncs positions from Robinhood, Fidelity, Schwab, and 35+ other brokerages into a single Google Sheet, so each AI name shows up once with live formulas no matter which account holds it:

With the raw values in one place, bucket each position by chapter and total the buckets, counting the overlap honestly — NVDA is in every chapter, so if you hold it plus a robotics fund, your true embodiment exposure is the NVDA value plus the robotics weight inside the fund. Then compare each bucket against your position-sizing rules. NVIDIA's print is a reminder that these are separate risk events that happen to run together when the demand is this strong; the day they diverge, the investor who knows exactly what they hold in each slice is the one who gets to rotate deliberately instead of reacting to the tape.

Frequently asked questions