Nvidia is funding an AI boom. Will it trigger a financial crash? | The Economist

Henry, describe to me the kind of remarkable  financial performance of this company. Yes, it's astonishing. So, in the era of AI,  since ChatGPT was released to the public, Nvidia's share price has gone up 14 times.  Uh and that's not just a speculative effect, you know, people punting on the company.  It's justified by underlying revenues, which have also shot up. And we've got a chart of  this. It shows Nvidia's quarterly revenue. If you. took it on an annual basis, it would show doubling  in each of the past three years. And now we're not.

far off a hundred billion dollars in quarterly uh  revenue. And that's just completely shot up from, you know, 2015 when this was much more of a a  a company for uh uh for gamers and enthusiasts. screwing in their graphics cards into their uh  computers. And this has been reflected uh in the. stock price. And then Nvidia has become big enough  that that's reflected in the performance overall. of America's stock market. Nvidia is now 8% of the  S&P 500. If I if I go back to the start of 2023,

it accounts for almost 15 cents of every dollar  that the S&P 500 has returned to investors. So, you know, if you're invested in the US stock  market, uh you're significantly exposed to this. company simply because it is so uh so large now  because the demand for its chips is so great. in. your pieces suite you look at look at its market  share tell us a little about how its market share. is being eroded I think maybe two to three years  back you would say that almost 80% of the AI chips.

that were sold in the market were from Nvidia  now what has happened is roughly around 60% of. the chips are from Nvidia but 40% of the chips  are coming from what is known as custom silicon. these are your um chips that are made by software  companies for their own internal needs and by the. end of uh the decade the the number is going to  be 50/50 so half the chips will come from Nvidia. Half of it will come mostly from these software  companies. I see. Okay. So that that leads us very. naturally I think into this question about the  financial engineering and and how it works. So I I.

think I think we do need though to go into an one  example just to get an idea of how how this works. So pick one of the many examples Shailesh and and  tell us how it works and if I get lost I I'll sort. of ask for help. Let's pick one of the most public  deals which is OpenAI which is a large AI lab. Um it is building an 8-gigawatt data center in  Ohio. Um 8 gigawatts is a very massive data center. uh the size of couple of different uh midsize  nuclear reactors because it's very expensive.

Now OpenAI is a very well-funded company but  it's also losing a lot of cash. So to be able to. uh for OpenAI to be able to afford this what  Nvidia is doing is it is guaranteeing the lease. on the land and power for that data center for 20  years. OpenAI will buy chips from Nvidia and the. fact that Open uh Nvidia is using its balance  sheet to guarantee that particular lease means. that lenders are more likely to lend to  OpenAI for that lease. If for some reason.

um it defaults, OpenAI defaults and cannot make  the payment then OpenAI then Nvidia sorry can. find another tenant for it. So you can right away  see that there are a lot of kind of circularities. here and I didn't mention Nvidia has also  invested $30 billion into open AI itself. Yes, it's interesting because of course why does Nvidia  have to act as a financier here? uh why can't the. markets provide capital in line with the vision of  the company of what is necessary and essentially.

um the way to view it I think is Nvidia taking  the view an even more bullish view than the. market at large about the amount of investment and  infrastructure that is necessary saying investors. aren't going to finance at the scale that we think  is necessary so you know we're massive we've got a. ton of resources and there's enormous revenue  stream we can reinvest we can do it ourselves. and in in in in that sense it's making the  twin bet on itself but also on the economy.

at large and it's able to do both uh because of  the scale of its underlying business shvidia's. defense against it what do people say at the  company and do you agree with them yeah so. the company is very sensitive to the word  circular financing and anytime you use it. they push back really hard uh now my view is  is two things so first and foremost there's no. getting away from the fact that Nvidia invests  in companies uh puts equity investment and then. those companies end up buying chips from there. So  there's no getting away from that fact. However,

I think there are also some real  differences. First and foremost is. um Nvidia is not um loaning the money to these  companies to buy the chips. It's essentially. guaranteeing some supply or some price that is  out there in the future. When Nvidia does this, it sells the chip and recognizes the revenue  right up front where versus these loans coming. these guarantees if ever coming due later. So  that that's that's number one. Number two is. um the fact that Nvidia's largest customers  are really cash flow positive hyperscalers.

and are are quite well cushioned. That is also  important when kind of drawing parallels to. the com boom because that is very different  position from where Nvidia is in. And lastly, I think is this balance sheet which we can I can  just go on and on about because it it holds around. hundred billion in cash and liquid investments  and this year it's going to generate around $200. billion in free cash flow and that's just number  is going to increase. So it's quite well cushioned. to be able to kind of sustain these kind of  guarantees. Um but yeah, I mean there is a whiff.

of that particularly when you invest in a company  and they turn around and buy your chips. I I don't. think there's any way you get around it. I want to  finish up just by asking you all whether we think. Nvidia's practices of this financial engineering  is more likely to make a recession happen. Shal, let's start with you. I don't know if it'll make  it happen, but it it could definitely amplify some. of the effects that happen because of it is quite  possible that some deals that are happening do not. need to happen. They are not that great. I don't  think it'll cause it in itself. Henry, Nvidia is. such a large part of the AI boom and the AI boom  is so important, especially via the wealth effect.

in the US stock market to the US economy that I do  think it's the case that if it's wrong and faces. serious problems, the economy will do worse on  account of that it's macroeconomically significant. owing to the role it is playing. However, it's  still the case that the number one people who lose. out in the event that this happens are Nvidia's  shareholders. They lose more than anybody else. And uh you know in the dot-com bust which is the  historic analog people draw the recession that.

followed that dot-com crash was quite shallow.  It wasn't like the global financial crisis or. something where you truly had this deep systemic  effect of the banking system on the economy. I I. think it's more likely to accelerate the AI uh  boom than it is to cause a financial crash. But. you know it's a sufficiently uncertain environment  that both outcomes definitely possible. and yeah I. think one of the difficulties is that if the if  the demand does not materialize for AI it could. Nvidia's role in the center of this ecosystem  could really accelerate uh the downpull it would.

both be losing demand from its own customers as  it's as it's um the the companies it has invested. in are starting to call due on those guarantees  it's provided so it could be really problematic.