A year on from what history may remember as an epoch-making earnings report, the most important company in the world’s still going strong.
That was the overarching message from Nvidia’s Q1 results, released to the usual agitated fanfare on Wednesday afternoon in the US.
Revenue for the period was a $26.04 billion, better than the $24.69 billion the Street expected and up a silly 262% YoY (and 18% from Q4).
Data center revenue was $22.56 billion, up 427% YoY and 23% QoQ. The Street was looking for $21.06 billion there.
“These increases reflect higher shipments of the Hopper GPU,” the company said, referencing the platform that trains large language models, among other critical generative AI applications.
“Strong sequential data center growth was driven by all customer types,” Nvidia added, noting that “large cloud providers” remain big growth drivers as “they deploy and ramp Nvidia AI infrastructure at scale.” Do you know any “large cloud providers”? I do. They accounted for nearly 50% of Nvidia’s data center sales in Q1.
As the figure above shows, the guide was high. Nvidia expects $28 billion of current-quarter revenue, $1.2 billion ahead of the Street. There was a “plus or minus 2%” caveat in there, but that’s just a perfunctory nod to uncertainty.
Plainly, the comps are getting tougher. $28 billion in sales for Q2 would represent “just” 107% top-line growth. Technically, that’s a big deceleration, but somehow “deceleration” feels like a misnomer in this context.
As far as the other key metrics and line items go, gross margin in Q1 was 78.90%, nearly 200bps better than expected. The margin guide for Q2 was a little short, but not by enough to matter.
Non-GAAP EPS of $6.12 was a huge beat. Consensus was $5.65. If you don’t like non-GAAP measures, GAAP EPS was $5.98, which is to say also a big beat. Free cash flow of $14.96 billion beat by $2.7 billion — or something. A lot. Suffice to say it beat by a lot. Nvidia also announced a 10-for-1 stock split and lifted the dividend.
These results will be parsed and parsed over the next 24 hours. Ironically, they’ll probably be run through some type of machine operating on an Nvidia chip. I don’t pretend to have the definitive take, but as noted above, it’d be a stretch to describe the numbers as anything other than strong.
Whether Nvidia did enough to power more near-term upside in the shares is another matter. Nobody’s likely to bid up the stock another 20% based on Wednesday’s report, but the outlook’s likely good enough to protect against a “sell the news” trade.
Jensen Huang was thrilled, as usual. “The next industrial revolution has begun,” he declared, for the umpteenth time. “AI will bring significant productivity gains to nearly every industry and help companies be more cost- and energy-efficient, while expanding revenue opportunities.”
nVidia chips don’t form unions, require health insurance, have babies, or need office space. 1 worker can do the job of 6. What a productivity miracle. Better talk to the HOA about adding barbedwire around the gated community.
A network of cameras, controlled by an AI network, will look better.
This morning as I was driving my son to school, I was briefly stuck behind some sort of agricultural implement piloted by an Amish farmer being pulled down the road by a 4 horsepower engine (which is to say, 4 horses). Then I came home and read this article.
I don’t actually have a point, but the contrast is so striking that I felt like sharing.
There are two kinds of life cycles arising from technology. One is the machine. In this case, the soup d’jour is Nvidia’s GPU (and similar items from competitors, and there are competitors). As neat as these chips are, once you’ve bought them, you don’t need to buy them again and again. Rather you install them use them. The other cycle has to do with the use of these machines. Applications will multiply faster than the chips will and their cycles will engulf the chip cycle. The chips will evolve, requiring constant innovation by the few firms capable of making them. The applications will have thousands of applications with even more innovators involved. Investors need to be careful which cycles in which to invest. Intel invented the integrated computer chip. The applications have exploded but Intel is not really reaping the gains enjoyed by many of its customers. Likewise, the computers that use Intel and AMD’s chips, have untold value-added applications but the applications are often more valuable than the machines themselves. What happened at Nvidia happened in the last few months. That’s today. Tell me what I should have put my money in five years from now. We wouldn’t have the Internet without Cisco, Broadcom, and others in their technology niche. While these companies are still viable, they are largely stocks one moves along from, as is Intel. (I know from painful experience.) Very soon (today, in fact) we will also have to consider the value and cost of the intellectual property that is required to create AI. There will be war in this area. For those interested in this issue I recommend some nice models developed by the late Igor Ansoff in his research and writing in his book “Implanting Business Strategy,” for example. Technology life cycles have been complex producers of both happiness and heartache over centuries.