For over three decades, the people most excited about Nvidia’s chips have been gamers. They used its graphics cards to render games in super-high definition. But over the last 15 years Nvidia has slowly established itself as the go-to provider of chips and software to the booming artificial intelligence space. Now it is investors that are paying attention—Nvidia’s market cap has almost tripled this year, briefly soaring above $1trn in late May. 
On this week’s podcast, hosts Alice Fulwood, Tom Lee-Devlin and Mike Bird explore the astonishing rise of Nvidia. The Economist’s Guy Scriven explains how the AI boom made the Californian chip maker into a $1trn company. Stacy Rasgon, an analyst at Bernstein Research, charts the company’s “30-year journey to overnight success” and Pierre Ferragu from New Street Research tells them that Google and Amazon have their sights on the chip market.
Sign up for our new weekly newsletter dissecting the big themes in markets, business and the economy at www.economist.com/moneytalks 
For full access to print, digital and audio editions, subscribe to The Economist at www.economist.com/podcastoffer

 Hosted on Acast. See acast.com/privacy for more information.

GEP helps Fortune 500 and Global 2000 companies navigate a downturn by reducing costs, optimizing cash flow and working capital, and improving their bottom line.

GEP's comprehensive portfolio of procurement and supply chain solutions helps great companies stay on track, ready for the next phase of the business cycle.

GEP, helping the world's best companies do better.

It's 1993, and in a condo in Fremont, California, 3 young engineers are figuring out how to get their new semiconductor business off the ground.

And we would get together And there would be nothing to do.

You get 3 guys and get together, you just talk.

It was like, could you put some donuts in the fridge in the morning for when we come?

I mean, so that would be a big deal for a while.

And so That lasted for a few months, just the 3 of us like that.

After much deliberation over these important issues, the trio got down to business.

They saw that video games were both computationally challenging and in high demand, a rare combination.

As such, they decided to build NVIDIA, a company that would design graphics chips for use in gaming.

The Silicon Valley venture capitalists, however, were not convinced.

And they told me there was no video game market.

People don't start companies to play games.

Graphics cards have underpinned waves of computationally intense technologies, from gaming to self-driving cars and crypto mining.

For 30 years, NVIDIA's chips have shaped what is possible, dominating the market.

In 2006, Jensen Huang, as NVIDIA's CEO, made another bet, launching software that would make his graphics chips programmable for use in artificial intelligence.

And so the reinvention process is very challenging, it's gut-wrenching, takes a lot of courage and it really tests your conviction.

Once again that bet appears to be paying off.

People have been getting really excited finally about artificial intelligence and generative AI and chat GPT and all this stuff.

And NVIDIA probably is the purest way to play that theme.

And so that's been driving a lot of demand, I think, for the shares.

The first version of chat GPT was trained on a supercomputer made up of 10, 000 NVIDIA graphics cards.

NVIDIA, because as I said many times, it's their underlying tech stuff that powers everything that we see in this space.

Nvidia itself is now 1 of the world's top 10 most valuable companies.

Its market capitalisation has far eclipsed that of Intel, once the world's mightiest chip designer and maker.

But for Nvidia and the other companies who are building fortunes on computing infrastructure, the future is far from certain.

What happens if the AI gold rush is just a flash in the pan?

You're listening to Money Talks from The Economist, our weekly podcast on the markets, the economy and the world of business.

In Palm Beach, Florida, I'm Alice Fullwood.

And in today's show, the astonishing rise of NVIDIA.

First, we'll hear how the launch of ChatGPT has fueled demand for NVIDIA's key product.

There's going to be an enormous boom in demand for AI coming over the next few years.

What that means is that we need loads and loads of chips to train the AI models.

Then we examine how and why NVIDIA came to define the AI market.

This is sort of their 30-year journey to overnight success, right?

This is something that they've been laying the groundwork for for years and even decades and it's finally coming to fruition.

And finally, we ask what, if anything, could topple the company from its perch?

Facebook is trying to develop a chip and Microsoft is trying to develop a chip and Amazon is developing a lot of different chips.

The challenge here is that in chip architecture, you don't have the freedom.

thing that I, and I'm sure many of our listeners want to know is Palm Beach.

I may be wrong here, but as far as I know, that's not exactly a major banking hub.

Yeah, it's not a financial center just yet.

Although there are a lot of hedge fund managers who spend a lot of time here as well.

But I'm actually on assignment for our bumper summer double issue, which means that I get to do slightly more frivolous things than I usually do.

Over the next couple of days, I will be doing some sunrise yoga and a screening of Legally Blonde.

We either need to look very seriously at changing our expenses policy, or Tom and I just need to become a lot more ambitious about how we use ours, 1 or the other.

Yeah, I reckon you guys need to up your game.

But returning to more serious matters, we are here today to talk about chips.

Yeah, so we've all been inundated with news about Chats GPT recently and all the other new generative AI products out there.

But the companies that make the hardware that powers those programs are also having a bit of a moment right now.

And we are talking about NVIDIA today because this is probably the firm, apart from OpenAI itself, of course, that has really come into its own as excitement about AI has exploded.

During the gold rush in America, those that really made a fortune were often not the people panning for gold, but those selling the picks and the shovels or people that own railroads or other infrastructure.

And The equivalent for the AI boom is the people selling chips and specifically NVIDIA's chips, which are needed to run the CUDA software that NVIDIA developed and is very popular with AI developers.

Nice way to build a sort of business case if you're not only building the hardware, but building the software that developers use to build those AI tools, which will only work on their hardware.

And the man who's been following this more closely than anyone else is The Economist's US tech editor, Guy Scriven.

Guy, thank you so much for coming on the show.

So how much attention were you paying to Nvidia before its share price went through the roof over the last month?

Well, I was probably paying a bit of attention, but probably not as much as I could have been.

For years and years, it was basically quite a niche player in the chip market.

So if you look back at 2016, its market cap then was still about $10 billion.

For years, it was designing chips for gamers.

And so it got very good at designing chips that were good at rendering computer graphics.

But at some point in the past few decades we realized that chips that are good at rendering computer graphics are also very good at calculating the huge mathematical equations that are needed to perform AI.

And so NVIDIA's chips basically found a new market in AI.

On top of that, the other thing they've done, which was basically very smart, is they've invested very heavily in their software platform.

And they do things like they go to universities and they teach professors how to teach their students about their software platforms.

They give access to their chips and with them the software platform away free to AI researchers.

And by doing this, it's basically developed a large kind of ecosystem of AI experts who all were kind of brought up using Nvidia's CUDA platform.

So CUDA is the name of the piece of software.

So what has happened over the last month or so to spark what is now being called this AI gold rush?

Why is the popularity of ChatGPT benefiting companies like Nvidia?

Excitement started to build around this area with the release of ChatGPT back in November.

So since then, you've had ever more releases of new kind of products infused by AI from really big tech firms, from medium-sized tech firms like Adobe and Salesforce, and from a huge bunch of startups and other kind of corporations as well.

I guess what's become clearer over the last few months is that people are going to try to use AI to improve various types of software offerings.

And there's going to be an enormous boom in demand for AI coming over the next few years.

And what that means is that we need loads and loads of chips to train the AI models and to make use of them.

That's what NVIDIA is kind of really good at making these chips.

And this became clear in NVIDIA's earnings release.

NVIDIA's last earnings, which was kind of a few weeks ago, was really impressive.

They beat analysts' expectations for their kind of revenue in the first quarter of this year.

They also basically told analysts in the current quarter, they expect their revenue to be about $11 billion, which again, beat analysts' expectations.

All of this is basically down to AI, so their AI business is completely booming and they can't ship enough chips.

And that's obviously had an impact on the valuations of all of these companies, right?

So if you take NVIDIA as an example, after its bumper earnings release at the end of May, its share price jumped by 30%.

And if you take a kind of even longer time span, so if you look at NVIDIA share price from the beginning of the year, it's almost tripled in that time.

So we looked at a group of about 30 companies, which do all sorts of things.

So chips, cloud computing companies, companies that help put together AI servers that go into cloud computing data centers, companies that rent data center spaces, even companies that provide their heating elements for data centers as well.

And on a kind of equally weighted basis, this basket of 30 odd companies has increased its share price by about 40% since the launch of ChatGPT in November.

And that compares with about a 13% increase of the NASDAQ composite, which is kind of a tech-heavy index.

So AI isn't just benefiting NVIDIA, it's benefiting a whole broad group of different companies in different industries.

Now, as you suggested, there are a lot of different parts, hardware, software, etc, combined in making AI possible.

We're going to hear a lot of these terms and through the episode things like GPU, CPU.

So could you explain to us sort of a little bit about all of these different chips and to what extent they're needed to make artificial intelligence possible.

So I think there's probably just 2 you really need to know about.

1 is called the CPU and the other is called the GPU.

The CPU is a central processing unit and the GPU is the graphics processing unit.

CPUs are kind of general purpose, and they're good at a variety of different tasks.

Graphics processing units, GPUs, are good at parallel processing, which means that they're basically very good at doing lots of tasks at the same time.

So GPUs were originally used for gaming and were very good at rendering all the kind of pixels and the color of the pixels and the shape of the pixels on a computer screen to make gaming run well.

Being able to run lots of tasks simultaneously is also very useful for AI, particularly for training big AI models, because you have to basically do lots of mathematical equations lots of times over, and also at the same time as well.

And so the kind of boom in AI has hugely increased the kind of importance of GPUs and massively increased their demand.

And the reason that benefited NVIDIA in particular is because it's the beehive when it comes to GPUs, right?

And that's a special focus that has been working on it for years.

So AMD is the closest competitor, but Nvidia really holds the lion's share of the market.

All right, and we're going to hear a bit more about Nvidia in a second, but thank you so much for your time, Guy.

Please stick around, we will come back to you a little bit later.

As we just heard from Guy, NVIDIA is a crucial company for GPUs and is the dominant player in this space.

Their breakthrough has really come as a direct result of the boom in AI.

So to find out more about the company and how it got to where it is today, I spoke to Stacey Rasgon, who is a senior analyst at Bernstein covering the US semiconductor industry.

The market capitalisation of NVIDIA briefly surpassed a trillion dollars recently, and that multiples the market cap of most other chip makers and designers like TSMC, Intel, Samsung, AMD.

It's also bigger than Meta, Tesla and Netflix and so this firm really has just been catapulted into the biggest league of tech companies.

You have to be clear by the way though you know that this company is over 30 years old.

So this is sort of their 30 year journey to overnight success, right?

This is not something that just happened over the last 2 months.

This is something that they've been laying the groundwork for for years and even decades, and it's finally coming to fruition as artificial intelligence and generative AI and accelerated computing and everything finally start to go mainstream.

I think people are finally waking up and recognizing that fact.

And it's sort of set off a scramble now to start to deploy this technology finally after decades of getting ready.

And because NVIDIA is sort of in the driver's seat, the markets finally have woken up to that fact.

What are the choices that NVIDIA made along that 30-year journey to put it in the position to sort of have this success now then?

Yeah, I mean, a lot of it comes down to creating not just the hardware itself, but the ecosystem around it.

And so they have a software platform that they call CUDA, which is used for parallel programming.

And also they built tons of libraries and other capabilities on top of it.

And this is something that they started developing in the early 2000, maybe 2005, 2006, I think.

And they got a lot of flack for it for many, many years.

I mean, this was effectively like a $2 stock on today's share basis for many years as they were investing in this.

It was limiting and impacting profitability and Jensen and the company held fast and they kept pushing forward on this even at the expense of the long-term profitability, right?

But they've created, like I said, not just the hardware, but also the software around it that's created a fairly sizable moat, as well as vastly increased the usability of these technology and the ease of deployment.

They've got a lot of folks that want to try to catch up to them now, like anybody that's trying to catch up has to now try to create their own software ecosystem against an ecosystem that is already entrenched.

That is a very, very difficult thing to do.

And there at times have definitely been choices that they've made about which technologies they think are going to be revolutionary, right?

I covered the crypto boom when it was booming, and it's not so booming anymore.

But obviously, sort of Nvidia chips were used to mine a lot of these currencies.

And it seemed like there was a decision that maybe that wasn't going to be the killer use for these chips, and maybe that they were sort of working towards AI instead.

They did develop some chips specifically for crypto mining, but it was more of an afterthought.

They were trying to limit the crypto miners desire to actually buy their gaming cards.

So it clearly can be used for AI, but it was also used for cryptocurrency mining, in particular, Ethereum.

This was not something that the company worked hard necessarily to enable.

It was just a fact they were useful for mining Ethereum.

And so the company started developing crypto-specific cards.

And the reason was they didn't really want the miners to buy up all the GPUs because it really made the gamers upset.

You couldn't buy a GPU for gaming if you tried.

And so they were trying to limit that in some sense, to try to create products that were useful for the miners and actually were only useful for mining.

And so they could try to bifurcate that marketable.

But it didn't work all that well, unfortunately, just because the demand at the time, at the peak for cryptocurrency mining was so strong, you know, they could have shipped whatever they wanted on the crypto specific stuff and those got bought up, but it wasn't anywhere near enough what the demand was at the time.

And by the way, we don't need to worry about it anymore.

To the extent that people are still mining cryptocurrencies, they're not really buying GPUs to do it anymore.

And so the fortunate thing for Nvidia and even for the gamers that buy this is we never have to worry about a crypto bubble ever again, hopefully.

Can we just talk a little bit about Jensen Huang himself, because all the sort of big tech companies that I've mentioned at the beginning, like Meta and Tesla and Netflix, all those kinds, they have these very visionary founders who have driven those companies to that level of superstardom.

Do you think Jensen Huang is made of the same stuff as them?

It sounds like he's had a pretty clear vision for a long time.

Clearly, he's been thinking far enough ahead, right?

The most visionary leaders, they don't just recognize where they think things are going.

They've actually been shaping the path that the industry's been taking.

They've been creating the markets, right?

So I think that is a characteristic that Jensen certainly has had.

And I'm a sell-side equity analyst, and they have these earnings calls, and you get on and people ask questions.

And I remember especially when they were going through the crypto bust, right?

And people were asking Jensen about like channel inventories.

Do you think Jensen spends 2 seconds thinking about channel inventories?

Like that's not presumably where he's spending his time.

Like he's thinking, you know, 10 plus years out.

I think on that note, we will probably leave it there.

So Stacey, thank you so much for joining the show.

So Tom, Mike, and any listeners who need a quick primer, channel inventories relate to the way that chips are sold.

So they're typically first sold to distributors and when demand ebbs and flows, those distributors can see their inventories backing up a bit.

Which you know, I really liked the sort of last point that Stacey made there, essentially about Jensen Huang, which is that he is this visionary guy.

He made graphic chips for gaming when no 1 else was focused on that.

That strategy turned out to be hugely useful because you could use these cards in all other kinds of things.

And then years before anyone else, he started dedicating his entire firm to AI.

So he's the type of person who's always thinking 10 years ahead about what demands for computing power will be.

And so of course, he isn't thinking about channel inventories.

And because those long-term plans have paid off, NVIDIA is now the fifth largest listed firm in America.

And much of that is thanks to the last few years of growth.

But the foundation is, as Stacey said, 30 years in the making.

So I've got an instinctive soft spot for Yen-Shan Huang, because there's that picture of him that came out a little while ago of him bombing around a Taipei night market.

If you haven't been, it's a place that sells some great food.

And this was happening just as NVIDIA's market cap was hitting $1 trillion.

And I really hope that's what I'd be doing if I made something worth a trillion dollars.

It seems like a good way to spend your time regardless.

I found all this fascinating and listening to it I had 2 sort of immediate thoughts.

1 very similar to Alice in that you've got this element of sort of visionary brilliance or luck depending on how you look at it.

In 2006 I was 15 and I think like most people I only really started thinking about AI and the major consequences of it in the past year or 2.

This guy was on top of it when James Blunt was constantly on the radio.

We're not talking about something he got to slightly earlier than everyone else.

This is an enormously long-term bet for a company like that.

The other thing that makes me think is the emptiness of what would have once been referred to as quarterly capitalism.

This idea that capital markets are really short-termist.

This company IPO'd in the late 1990s, and it didn't crack $10 a share, sustainably in any way, until about 2016.

It'll move around a little bit, but it's enormously valuable.

We usually think of Amazon as an example of a company like that that sort of bursts out of a range after a very long time.

But maybe Nvidia is going to turn out to be an even better case of the sort of proof of capital markets actually turning out to be pretty good vehicles at keeping these very long-term ambitious projects going, contrary to the story that we usually hear.

There is part of me though that's a little wary about just how giddy NVIDIA's valuation has become.

So it's market capitalization has jumped to around 200 times earnings now, which is about 10 times the level of the S&P 500 as a whole and Microsoft, another company that is seen as a likely big winner from the AI boom, is trading at around 35 times.

And if you look at Nvidia's share price over the last few years, it has tended to see-saw.

So during the latest crypto craze, it surged and then it dropped back down again and now it's right back up again and the swings in its valuation are much more pronounced than the swings in its business fundamentals so I don't doubt that Jensen Huang has been incredibly far-sighted and that the company has built quite a robust competitive position here.

But I think there's also some kind of feverish excitement going on based on these kind of giddy expectations for AI that have really yet to be proven out.

On the topic of hype, the big news story in tech this week was the release of Apple's much anticipated augmented reality headset, which is selling for a modest

Obviously, you'll eventually be able to use this to read your edition of The Economist, but until then, you might have to stick to the print and digital editions.

See all Money Talks from The Economist transcripts on Applepodcasts

Money Talks: The trillion-dollar question