Bloomberg Audio Studios, podcasts, radio news.
Perhaps no company embodies the promise of AI more than in Vidia.
Very quick story of in Nvidia is that it used to make video game chips that gamers really liked. It turned out those things we're good for AI, and now it's one of the biggest companies in the world.
Joshua Breustein is a technology editor at Bloomberg BusinessWeek. He's covered in Videa's stunning rise and says the company is now on the cusp of its next chapter.
And video doesn't a funny place because, on the one hand, it is in a great position. It's very good at making these chips that everyone wants.
Microsoft, Meta, Amazon, pretty much all the big tech companies have collectively announced hundreds of billions of dollars of investment in AI, and much of that investment will go straight to buying in Vidia's product. In some ways, it's a dream position for a company, but being at the top also means there's a law way to fall. In Vidia got a glimpse of what that could look like in late January, when the Chinese AI startup deep Sea released
its R one chatbot. Deep Seek's developers say their model uses a fraction of the resources of American competitors. That news rattled assumptions about how much computing power and how many in Nvidia chips the AI revolution will actually need.
The main event yesterday one single name in video in video getting absolutely hammered yesterday session down by seventeen percent, a record five hundred and ninety three billion dollars taken off the market cap.
It was the largest route in market history, and Joshua says the risk of something like that happening again is something in Vidia's founder, Jensen Huang takes seriously.
Huang looks at the history of semiconductors and knows that when you see a big transition in tech, the infrastructure companies do really well at the beginning of it, and then their products become commoditized and they crash.
I'm David Gera, and this is the big take from Bloomberg News today. On the show, in Vidia's scrambled to future proof it's technology and hold on to its top spot. Today, the semiconductor manufacturer in Vidia is one of the largest companies in the world by market cap. Bloomberg's Joshua Brustein says its story starts more than thirty years ago with its CEO, Jensen Huang.
He just really is known within Nvidia in the wider industry as kind of a force of nature. He's a very hard charging boss. He is very enthusiastic about the technical aspects of running a semiconductor company.
To put it simply, he's a nerd.
But not in the way that I think a lot of people think of tech nerds, where it's like a lot of talk about our computer is gonna become sentient, and you know, will they lead to the end of human existence. It's more kind of like how many transistors can we fit on this chip? And is there a new architecture we can use to do it even better? So he's not doing the science fiction thing. He's doing the engineering thing.
Joshua says. Doing the engineering thing has been at the core of Nvidia's business since the beginning.
So the founders thought that they saw a niche for making a specific kind of chip that would be really good at rendering graphics, primarily for video games. The idea was that there's general use chips, but if we focus on this one use, we can find a good niche and maybe we can you know, make a couple of tens of millions of dollars.
And it was in pursuit of that goal that in Vidia develop was a key breakthrough in chip technology.
The basic way to put this is that Intel makes chips that are general use chips that are very powerful and that do things sequentially. They do a task by doing one part of the task, then the next part of the task. In Vidia's chips take a task, they break it up into a bunch of little parts, and they do it all at once. There's a forthcoming book about in Vidia where they describe it as instead of having a truck that drives around the city making deliveries,
they buy a bunch of motorcycles. Each motorcycle carries one package, and they go to everyone's house all at once. So that turns out to be a very good way to do modern AI, which is based on these things called neural networks, which pass information through a whole bunch of nodes in parallel.
But the wider use and mass market appeal of Nvidia was still years away, and Joshua says the company went through some rocky times, like.
A lot of companies in the semiconductor industry. It went through a series of existential crises. It would release a chip, there'll be something wrong with it. It almost went out of business several times early in its existence. This is part of company lore. Jensen likes to say that over always thirty days from going out of business.
In Video sought to widen its customer base. The goal, Joshua says, was to take its video game graphics cards, which were getting faster and faster, and sell them to non video game clients.
The problem was doing this required making the chips more expensive, so it started doing things that made its core product worse in the hopes that someone would find some use for it.
And eventually someone did. In Nvidia built its own computer language called Pudha, which allowed developers to maximize the potential of their GPU chips, and in Vidia's coding language and in Vidia's chips became the standard hardware for early pioneers of what's called deep learning. That's a branch of computer science that relies on those neural networks modeled after the
neurons in the human brain. Early on, Joshua says, the commercial applications of deep learning weren't yet clear, but Nvidia saw an opportunity.
It just happened that in Vidia saw the seeds of its product being good for this use and really pursued it from a very early point and was very effective at building this technology.
Then OpenAI's chat GPT took the world by storm, and so suddenly every player looking to invest in AI was in need of Nvidia's chips.
In Nvidia now finds itself having the thing, maybe not that customers want the most, but it has a thing that Meta wants the most, Google wants the most, Microsoft wants the most, and that puts it in an incredibly powerful position.
For the last few years, the tech Titan seems certain that the AI future would be powered by massive data centers filled with Nvidia chips, and the Frenzy showed just how quickly the fortunes of semiconductor manufacturers can change. At the start of twenty twenty, Intel's market cap was roughly
twice the size of Nvidia. Today, in Vidia is around thirty times more valuable, and Joshua says the precariousness of being on the cutting edge of tech is something that weighs heavily on Nvidia's CEO, especially after the deep Seek route in late January.
There's some jitteriness now, not only within video but throughout the economy that like, are we reaching the end of this stage of the AI boom? I mean, people who are pessimized it will tell you we're reaching the point where the bubble pops. Other people will say it's time to move on from building the foundation to actually creating
something more than that. And because expectations are so high, if this isn't a really really big success, even if it's just like a pretty good success, then in Vidia could be in kind of a tricky spot. And that's sort of the water that it's navigating right now.
Joshua Sayshuog is now scrambling to find new customers and markets for Nvidia's technology.
His plan for doing that is to do more than just make the hardware that everyone builds their tech on top of, but to actually be involved in building a little bit more of that tech, both which brings that tech along faster, increasing the demand for the company, and which makes it harder to get rid of Nvidia and go with something else.
So what is the company planning for its future? That is, after the rate, Even though in Vidia is the world's most valuable chip maker. It's CEO Jensen Huang is always selling as he was back in January Welcome to CUS at the industry's biggest trade show in his trademark leather jacket. Do you like my jacket? Blimber's Joshua Brustein says one of the markets Swang is trying to develop is what he refers to as physical AI, and.
That's basically just instead of chatbots that operate on the Internet, it's stuff that operates in the real world.
With this physical AI, there are many downstream things that we could do as a result.
We could do synthetic so you have robotics factories that would be automated self driving cars. Obviously, that's been a big thing for Nvidia for about a decade, and if they can get all of those things to happen sooner, then those are enormous markets that they can.
Unlock, Joshua says. Another big strategy is developing new software technology in Vidia could sell not only to their existing customers, but tech that also could be used to unlock AI applications in some of the world's biggest industries, like healthcare.
That would be not only hospital systems but also pharmaceutical companies doing drug discovery, maybe robotic surgery, but also just kind of the administration of hospitals.
Joshua Siswang is pursuing this with a lot of urgency.
So what he's doing now is really trying to almost will into existence all these AI applications that you'll hear people say are coming soon, are coming down the road. He wants those things to come in six months, not in seven years.
It's not just will, it's also money, Joshua says. In Vidia is investing in all kinds of companies that are trying to realize new commercial applications for AI.
Not so much in the hopes that they'll invest in a startup that's now worth a million dollars that will be worth a billion dollars, although they wouldn't mind that great, but that they want to make sure that they are having as many irons in the fire of future in Vidia customers, Like they just want a startup that is now small to be an enormous in video customer, and so they invest in and work with lots of startups in many different fields to try to form those relationships
and to have them like working on in video hardware from the very beginning.
You mentioned that in Vidia makes its own software. It's not just making chips, but it's serving dips.
There are kind of other.
Things augmenting the kind of core of its business model. What is the appetite so much as we know it for that other stuff. Are we at a point at which tech companies AI companies really just want the chips they don't really care about that stuff. Or is the other stuff that in video makes appealing to them? Do they need it? Do they want it? Is that catching on?
I think that that question is like a core dynamic to watch within video right now. There's a metaphor that computer scientists use a lot, which is like a stack of technologies. You start at the bottom, that's the chips, all the way at the top is like whatever software the user's using, and in videos, building upwards on the stack and talking to a lot of startups who are in video customers, the enthusiasm starts very high at the
bottom and diminishes as they go up. So they really want the chips, and like the software that makes the chips do the basic functions they're really into. When it gets to like the NVIDIAI models, they'll kind of politely acknowledge that in video makes those but like they kind of wouldn't build that stuff, So it's hard to tell where that line is. And in video will say, well, we're happy as long as they're you know, like we
want them to use whatever is useful to them. But they're also aggressively trying to sell you the whole computer. So it's interesting to watch how that plays out.
I think in VideA is a proxy for a lot of people for AI generally. I think for investors, it's a way to get into this this sector, this part of the tech sector. What is its success or failure navigating say these next twelve months going to say about the industry more broadly, it's in other words, how good is it as a proxy for where AI is and where it's heading.
So I think that on the one end, if you saw like a real retrenchment in the construction of data centers, a real lack of demand in video chips, then that would be really bad for in video. It'd be bad for video share price and all those investors who are hoping to get in AI, and it would indicate that maybe AI has been overhyped over the last couple of years. On the other end, you could see full speed ahead,
all of these things are getting built. In video is still doing really well, and there would still be the question of when we get past the building out phase, what happens next? And I think even in video will tell you like, that's not going to We're not going
to know that in twelve months. It will tell you that it will be everywhere, but the world will change more slowly than maybe the next year, and so you'll have to just continually watch, like is AI progressing, and if it is in video will likely be in a central role, and if it's not, then you know, we may have to reassess some things.
Is there a moment? Are there going to be moments in the near future when we're likely to learn more about what he has planned?
Yeah? One is coming up actually next week and Vidia has an annual conference in which it sort of takes over downtown San Jose. Almost a thousand other tech companies come and present how they interact with Nvidia. Jensen will give a big keynote in which he will lay out the vision. I'm sure we'll hear a lot about his pet projects at that point, and given the potential moment
of inflection. We're in for AI. Generally, it will be an interesting time to kind of take in video's temperature and see how the company responds to what could be a tricky moment.
He'll don the leather jacket and yes, always done to the leather jacket.
If he has on a different jacket, then we'll know something is big and changed.
This is the Big Take from Bloomberg News. I'm David Gera. This episode is produced by Alex Tye and David Fox. It was edited by Tracy Samuelson, Mark Million, and our senior editor, Elizabeth Ponso. It was fact checked by our editorial team and mixed and sound designed by Alex Sigura. Our senior producer is Naomi Shaven. Our executive producer is Nicole Beamster. Board Sage Bauman is Bloomberg's head of Podcasts. If you liked this episode, make sure to subscribe and
review The Big Take wherever you listen to podcasts. It helps people find the show. Thanks for listening. We'll be back next week.