Nvidia printed $81.6 billion and the stock went down.

Not a lot. About 1% in after-hours. But still – 85% revenue growth, $1.87 EPS against a $1.76 consensus, a Q2 guide of $91 billion with zero China data center revenue baked in, an $80 billion buyback authorization, and a dividend hike from a penny to $0.25 per quarter. All of that, and the tape shrugged. That reaction is either the market being rational about expectations, or it’s a tell about where positioning was heading into the print. Probably a bit of both, honestly.

Here’s the thing most people won’t sit with long enough: on a forward basis, this stock isn’t expensive.

Trailing P/E is around 45.6x. Tech sector average is roughly 34x. That spread sounds bad and most headlines will stop right there. But trailing earnings on a company growing 85% year-over-year are already obsolete the moment they’re published. You’re looking at where the car was, not where it’s going. The forward P/E – based on current consensus estimates, which haven’t been revised up yet after tonight’s guide – comes in around 26.9x. The semiconductor industry median is roughly 36x. So Nvidia, after its biggest data center quarter ever, is trading at about a 25% discount to its own peer group on a forward earnings basis. At a $5.4 trillion market cap. That’s the number that keeps nagging at me.

Those consensus estimates move higher in the next two to three weeks. Analysts update models after prints, not before. A $91 billion guide with 75% gross margins doesn’t give the Street much room to stay conservative. Which means the forward multiple compresses further even if the stock doesn’t move at all.

The PEG is 0.67. Below 1.0 has historically flagged undervaluation relative to growth. At 85% revenue growth, a reading that low is genuinely unusual – the kind of number that shows up when price hasn’t caught up to earnings, not when it’s gotten ahead of them. Worth pausing on.

Slight tangent: NVDA’s 10-year average trailing P/E is around 66x. The stock right now is 31% below that long-run average. And the people paying 66x in prior cycles were buying a company with a fraction of today’s revenue, margin, and free cash flow. The business now carries $62.6 billion in cash against $11.4 billion in debt, generates $96.7 billion in trailing free cash flow annually, and runs a 126% return on invested capital. Historically, investors have been willing to pay far more for far less.

The counterweight to all of this is size. $5.4 trillion is a number that creates its own gravity. The stock touched an all-time high of $236.54 on May 14 – a full week before the print – and consensus price targets cluster around $275, which is only about 23% from current levels. Not a wide margin for a stock with this much embedded expectation. Jensen Huang has laid out a roadmap to $1 trillion in combined Blackwell and Vera Rubin revenue across 2026 and 2027. If hyperscaler capex holds, if enterprise AI deployments accelerate, if sovereign AI spending keeps flowing – that roadmap looks conservative. If any of those legs wobble, 26x forward on a decelerating story reprices fast. That’s not a prediction. Just the math on the other side of the argument.

What’s interesting is that the muted after-hours move might actually be the most useful signal from tonight. Not because anything is wrong. Because the market literally didn’t know how to react to a perfect quarter anymore. Whether that exhaustion is a short-term ceiling or just noise before the next leg – that part isn’t clear yet.