We will tell you about Nvidia earnings. They crushed. Highly anticipated as always. Nvidia is permanently holding up the US economy, continues to hold up the US economy, the global markets, but it's sold off even though they blew out earnings and it was a very good quarter. So earnings dropped yesterday after we got off the phone with Doug Olaflin from Semi Analysis. Some analysis did have some good breakdowns here, but the top line revenue came in at 68.1 billion dollars, up 73% year-over-year and 20% quarter-on-quarter. And this beat consensus estimates by nearly 3%. The stock price popped around 3% immediately but then sold off 5% after the market open this morning, making it the stock's worst day since last April. They're now just a tiny $4.5 trillion company.
Last April. Last April was DeepSeek.
Yeah, that makes sense. People were saying you won't need a whole bunch of Nvidia chips because you'll be able to inference cheap commodity hardware and the models.
DeepSeek was climbing the app store charts in America totally organically.
Yeah, there was a lot of weird stuff going on.
They weren't using a massive bot farm at all.
Yeah. I mean, I guess I sort of steelman the DeepSeek story. There was this interesting takeaway which was that even though the numbers were sort of misreported.
We had it wrong. I think it was January.
Oh, Liberation Day tariffs. Yeah. Chip sanctions. But yeah, the DeepSeek lesson was like you can distill models. You can get, you know, a certain level of intelligence at a much lower cost. And that is real, but the demand on the frontier is just incredible because people are like, 'Oh, I have a strong opinion about 5.3 versus 4.6.' Like people really care about being on that perfect leading edge for exactly what their use case is. You see this with people having favorite models, favorite flavors all over the place. And so the demand in the Jevons paradox really just powered right through.
What do you have to say? I was going to say, like, I think when the DeepSeek moment happened, the narrative was not that they were distilling though. It was that they had built, like, they had trained a brand new model with very few resources. There was no sense that they were just distilling.
There were rumors about it pretty.
There were rumors, but the main narrative was not that they were being distilled. I think what I mean about distilling is that the truth that came out over the news cycle was that distillation does get you somewhere near a particular frontier capability and allows you to reduce cost, but there is still another order of magnitude of demand for the next generation and the next level of the frontier, and we have yet to see a plateau emerge for demand for whatever the next frontier is.
Yeah, and overall the story of the last year is that there's very, very, very little demand for number three. There's some demand for number two, and there is an exceptional amount of demand for the most performant model for a specific task.
Yeah. So, going back to our newsletter today, our earnings recap, which you can subscribe to at tbpn.com. After the earnings call, Semi Analysis called the results staggering. I love it. Clean bill of health on gross margin, revenue, and the guide is firmly above buy-side bogeys. Incredibly clean, really. They also pointed out what Nvidia said about supply commitments with 21.4 billion of inventory on hand, up from 19.8 billion. They have 95.2 billion of supply locked in with chip manufacturers, also stating, 'We have strategically secured inventory and capacity to meet demand beyond the next several quarters.' In other words, we're not running out of chips anytime soon. That's good news for Nvidia.
The second-order effect that I think was interesting that Doug Olaflin flagged for us was that even if TSMC does scale up magically, or they figure out how to build another fab, they increase capacity in Arizona, even if Nvidia has sharp elbows and is able to push out other demand and get all the line time they need to build all the accelerators that they need, soak up all the CPU demand, etc., etc. Well, then you could still wind up in a weird scenario where Nvidia has the chips, they're ready to sell them, the hyperscalers want to buy them, the AI labs want to inference them, but there's just not enough energy. And so there's this question of when does the chip bottleneck shift to the energy bottleneck? That could be part of what is sort of worrying people because you can see in the timeline we have this chart, 'Are you not entertained?' from Nathan Benake over at Airstre, and he's taking a screenshot from the Financial Times, and it is just one of the most incredible graphs I've ever seen. And so as you see this chart, you have to wonder if there are any bottlenecks that they will run into. TSMC capacity.
Yeah. And I asked Doug about this. I asked Doug about this. What does Nvidia do if their customers do have an energy bottleneck?
I feel like they'll find a way to still ship the chips.
I agree. And truthfully, as big of a story data centers are, they're still consuming well under 1% of US electricity. And so there are chips to be moved around. There are new power plants to be brought online. We're actually talking to Doug Bernau from Radiant about nuclear power today. And there are so many different ways to solve the energy bottleneck. But it is very real world. It is very slow. And if there's a timing mismatch, you could see a little bit of a flatline that I think people might be worried about.
So, Jensen himself strongly pushed back against the SaaS apocalypse narrative yesterday, which is interesting because he doesn't necessarily have to. He could be out there saying like, yes, all those SaaS CPU workloads, they're all going to be GPU workloads now. And yeah, you should actually sell all your SaaS because the future is inference. Like every app that you use, it's not going to be code that's running on a CPU. It's going to be inference on a GPU on demand. And so, you know, demand for Nvidia should be even higher. Like, he doesn't really have that many bags with the SaaS companies necessarily. Of course, they'd be upset if he was talking trash, but he's not. He's defending them.
But he buys a lot of software.
Sure. Yeah, of course. He said, 'I think the markets got it wrong on the SaaS apocalypse.' This is Jensen Huang talking to CNBC's Becky Quick, pushing back on the fears that AI agents will cannibalize the enterprise software industry. Instead, he expects a broad swath of software firms to use agentic AI to develop their software and boost efficiency. In what he described as counterintuitive, Huang said that AI agents won't replace these software tools, but will use them instead. Why even waste the tokens building a calculator when you can just download the calculator SaaS or whatever, whatever SaaS you want to grab off. And this is certainly what we've seen with databases. Like certain databases have just skyrocketed in demand because they're the ones that the agents are pulling off the shelf. And so yes, the agent could design a new database and install it, but that's a lot of work. It's a lot of wasted tokens when databases already exist.
So let's tell you about Finn.ai, the number one AI agent for customer service. If you want AI to handle your customer support, go to finn.ai. And let me also tell you about vibe.co, where DC brands, B startups, and AI companies advertise on streaming TV, pick channels, target audiences, and measure sales just like on Meta. And we can move over to the timeline reaction. Nvidia is such a big deal. It's not just on the cover of the business section. It's on the cover of the Wall Street Journal. Didn't get the picture though. Didn't get the picture. The picture is Iran. Iran talking about nuclear enrichment is on hold according to experts. But Nvidia results help soothe AI fears. The surge in profit comes as the AI industry's appetite for chips continues to grow. Data center hardware, the chips and networking equipment that Nvidia sells to AI and cloud computing companies, accounted for 91.4% of the quarter's sales. Terrible news for gamers. Like absolutely the worst numbers you could ever see if you're gamers. Maybe that's why they're maybe they're launching a short attack.
John would be punching a hole. I would be, maybe the gamers have risen up and they are shorting Nvidia to try and get them to go back into gaming.
Computing demand is growing exponentially. Chief executive Jensen Huang said the agentic AI inflection point has arrived. With each passing quarter, the pressure grows on Nvidia, which at a market value of nearly five trillion is the world's largest publicly traded company. Is it not just the world's largest company? Is there a larger privately traded company? I don't think so. It must just be the world's largest company. I don't know. I guess you get into crazy definitions of like, does Russia count as a company since it sort of is all one controlled economy. It is no longer enough for Nvidia to produce good quarterly results. They have to produce perfect quarterly results, said Daniel Newman.
How could they have done better?
Instead of net income of $43 billion in the quarter, they could have put up 50 billion or 60 billion or 100 billion or 10 trillion quadrillion. I want to see a quadrillion dollar quarter. Please, Jensen, I'm not happy. I'm not satisfied. Keep growing. Keep selling these chips.
Gavin Baker gave some thoughts ahead of Nvidia.
