There's a classic scene in The Wolf of Wall Street where Jordan Belfort looks straight at the camera and says: "I'm not leaving." That's pretty much what Jensen Huang just did with Nvidia's fiscal Q4 2026 earnings report.

The numbers? Revenue of $68.13 billion, up 73% year-over-year. The Street was expecting $66.2 billion. Adjusted earnings per share of $1.62, versus the $1.53 consensus. Everything above. Again. For the umpteenth consecutive time.

But hold on, because the best part is coming.

The guidance that shut the bears up

Look, beating revenue estimates by $2 billion is nice. Cool. But what really matters is that the guidance for the current quarter came in $5 billion above what analysts projected. Five. Billion. Dollars. More. Than what the suit-wearing crowd on Wall Street, with their 47-tab Excel models, managed to forecast.

And here's the kicker: analysts had already revised their estimates upward after the hyperscalers (Microsoft, Amazon, Google, Meta) dropped their jaw-dropping capex budgets for the year. In other words, the gang tried to fix the model. They just didn't fix it enough.

CFO Colette Kress then put the cherry on top: she said Nvidia expects sequential revenue growth throughout all of 2026, exceeding the $500 billion revenue opportunity Jensen had mentioned last fall for the Blackwell and Rubin chip generations.

For those lost in the jargon: "sequential growth" means each quarter will be bigger than the last. It's not just "oh, it grew compared to last year." It's quarter after quarter climbing higher. That's rare. That's monstrous.

The margin question (and why Jensen isn't worried)

Everyone wanted to know: with memory costs shooting up like a rocket, will Nvidia's gross margins — sitting in the 75% range — hold up?

Huang's answer was elegant in form and brutal in substance: Nvidia's best lever is delivering generational leaps in performance. Translated from Jensen-speak into plain English: "If our chip is still the best on the market and nothing else comes close, we pass the cost along and the customer pays with a smile."

It's Taleb's logic applied to hardware: when you have a practical monopoly on an essential technology, you have pricing power. Simple as that.

The story of old chips that nobody throws away

Now, here's a golden nugget that flew under the radar: Kress mentioned that even the Ampere chips, from six years ago, are sold out in the cloud. Hopper? Sold out. Ampere? Sold out.

Why does this matter? Remember when everyone piled on the hyperscalers for extending their chip depreciation cycles? They called it "financial engineering." The Wall Street Journal ran a story. CNBC ran a story. It was the usual circus.

Well then. If chips that are six years old are still running and in demand, maybe — just maybe — extending depreciation wasn't accounting trickery but simply a reflection of reality. The chips last. They generate revenue. They remain useful.

This gives buyers confidence — both the giants like Microsoft and Amazon, and the neoclouds like CoreWeave — that the hardware they buy today will generate returns for years. And when the buyer has confidence, they buy more. And when they buy more...

Well, you already know who wins.

And the stock?

Shares dipped a measly 50 cents in after-hours trading, to $195.35. They had touched $200 right after the numbers dropped. Those insane +15% post-earnings moves from the early days of the AI boom are behind us. Nvidia has become such a massive company that even spectacular results are already partially priced in.

But don't confuse short-term price action with fundamentals. The fundamentals here are crystal clear: demand for AI infrastructure is not decelerating. It's accelerating.

Now, the lingering question: if after all of this the stock barely moved, how much more does Nvidia need to deliver to satisfy this market? Or is it time to stop waiting for the perfect pullback and accept that sometimes the train doesn't come back to the station?