NVIDIA beat everything again. This time the stock actually went up. Here's what was different.

Four straight quarters NVIDIA beat revenue, beat EPS, raised guidance, and the stock sold off anyway. Average decline was -4.9% five days after the print. Every single time. Didn't matter how big the beat was.

Q2 FY2027 was the fifth. $96.2B revenue, up 106% YoY. Stock dropped about 2% on the initial release after hours. Same pattern loading up. If you've been holding through these you know exactly what that feels like.

Obviously I've been long this the whole way down through all four of those selloffs, so factor that bias in.

Anyway here's why I don't think this was just a one-day pop:

Jensen did something he's literally never done before. He gave a full-year forward revenue guide. 70% growth for FY2028. Supply-constrained, meaning that's the floor, not the ceiling. Kress said customer demand implies a doubling.

The stock reversed during the call. The -2% became +8.7% by the next close.

70% on \~$401B FY2027 consensus = $682B implied FY2028 revenue. The Street consensus was at $601B before the call. It jumped to $637B within 24 hours as the fastest analysts repriced. That's $36B of revision in one session. There's still a $45B gap between where consensus sits and where management just told you the floor is.

Could be wrong on timing, wouldn't be the first time I got the mechanism and then sat through three more months of nothing happening before the multiple actually moved.

LSEG SmartEstimate (which overweights the most accurate analysts) is already at $689B. Predicted surprise: 8.16%. Individual estimates range from $438B to $879B across 57 analysts. That dispersion alone tells you most haven't updated their models yet.

The prior post I wrote explained why the stock kept selling off despite massive beats. It was a combination of everyone who can own it already owning it (78% active institutional ownership, 8.3% of the S&P 500) and sell-side models not being able to build discrete segment models for new revenue lines like Vera CPU because those lines hadn't appeared in a 10-Q filing yet. Jensen just solved the second problem by giving them the total. The blank cells in the analyst models don't need to be filled individually anymore.

Meanwhile the stock trades at 14.81x NTM EV/EBITDA. That's down from 32.44x a year ago and 16.71x from my prior post. It got cheaper while the thesis got stronger.

Curious if anyone thinks Jensen's just sandbagging the 70% to guarantee himself beats for the next four quarters.

Wrote the full thing up with the options flow data if anyone wants the numbers behind it: [https://substack.com/home/post/p-213088654](https://substack.com/home/post/p-213088654)