Nvidia reported after the close. As we discussed yesterday, if they follow the history of beating guidance, we'd likely be looking at a company running at a $400 billion annual rate, and growing at a triple-digit rate (again!).
We got it.
Nvidia reported $96.2 billion in revenue this afternoon, up 106%.
Just four years ago it was a $7 billion quarter. Then $13 billion. Then $30 billion. Then $47 billion. Now $96 billion. Back to triple-digit growth.
Keep in mind, that's doubling off an already enormous base. It's unprecedented at this scale.
That said, the CFO led the call by saying they expect 70% revenue growth in fiscal 2028. That's not demand slowing to a 70% growth rate, it's a supply issue (again).
The supply constraint is back.
It's memory. It's advanced packaging. It's power. Jensen says the entire supply chain is constrained.
We've watched this pattern for three years now. The buildout looks like it's found its limit, then it shifts into a higher gear, and then it runs into a new physical wall (chips, storage, power, memory).
And it's severe enough that Nvidia has committed $279 billion to secure future supply and manufacturing capacity, up from $119 billion just three months ago.
Let's talk about two other key takeaways from the call that tell us about, not just Nvidia, but about the state of the AI boom and outlook.
1) The demand is broadening.
Everyone sees the AI spending from the big hyperscalers (Google, Meta, Amazon, Microsoft, Oracle) . That business at Nvidia grew 102%.
The other business is the AI clouds, enterprises and sovereign governments. That business grew 138%, and it's about to become the larger of the two. Jensen's framing was that the hyperscalers were the beginning, not the destination.
2) The next key takeaway: They changed the way they report on the businesses within Nvidia.
Five product lines are now two "platforms."
There's Data Center. And there's Edge Computing.
Edge is only 7.5% of revenue today. Yet Nvidia now considers it one of only two "platforms." Why? Jensen's answer throughout the call was agentic AI. This is continuous inferencing from, ultimately, billions of AI agents running around the clock.
By making Edge Computing a one of only two platforms in Nvidia, is this Jensen telling us where he thinks the rest of this decade goes?
It seems that way.
He says "when the world goes to fully agentic, you're going to have agents running all the time, working with other agents running all the time." That's constant compute usage. And the meter is always running — billable compute.
And this works because, as he said, AI is doing productive work, which generates profitable tokens, which becomes more productive work and more profitable tokens with the addition of more computing capacity. That's the self-reinforcing loop Jensen has been talking about all year — what we've called the “boom loop.”
