00:00NVIDIA's AI chips are no longer just pieces of hardware. Wall Street is starting to treat the
00:06compute power they enable as a financial asset, opening a new chapter in the AI boom. Joining
00:11us now to break it all down is Nathan Bomey, business reporter at Axios and author of the
00:16Axios Closer newsletter. Thank you so much for being here, Nathan. Glad to be here. Yeah, I mean,
00:21we have so much to get through here. This is such an interesting and exciting topic with everything
00:26going on these days with AI. I mean, NVIDIA has helped bring more than $500 billion in financing
00:31to AI infrastructure with major Wall Street firms, including Goldman Sachs and BlackRock.
00:36Why is this such a significant development for the AI industry? Well, multiple reasons. I think one
00:42thing it shows is that there's so much capital that's necessary to run these data centers, to
00:50build the data centers, that NVIDIA can't sit on the sidelines anymore if it wants to keep
00:54this big and running. This is a deal of such incredible proportion that it just shows you
01:02that this is not the kind of thing that a company like OpenAI can finance on its own. It needs
01:08the help of a very well-capitalized company like NVIDIA to be able to do so. In some ways,
01:15this just shows that the massive revenue potential is worth it for NVIDIA. Now, on the other hand,
01:21critics would say that it shows this circular financing threat within the AI economy is continuing
01:27to build in the sense that you've got the companies like NVIDIA that are supplying chips to a company
01:34like OpenAI are becoming closer, more entangled, and that could build risks in the long run.
01:40Yeah, I mean, both of those sides to the story are so important here. As we look to what's going
01:44to
01:45happen next, NVIDIA CEO Jensen Huang has described AI Compute as a potential new asset class even.
01:50So what exactly does it really mean to treat NVIDIA chips as financial assets?
01:55Well, I think they certainly are assets. Now, I think one of the questions is,
01:59can they make money off of these assets in two different ways? So for starters, they're selling
02:05the chips, of course, but then they're also sort of now capitalizing on the actual sort of leasing
02:12of those chips in a way by effectively providing lending capital to an OpenAI. Not directly,
02:21they're actually going to be guaranteeing the financing on this data set, a very complex
02:25arrangement. But the point is, they're standing behind this, and this ultimately means by extending
02:31their own capital, they could actually make some money there as well. So they're kind of making money
02:35in the deal in both ways. That's why the chips become an asset. Now, the issue here is these are
02:41going
02:41to be very quickly depreciating assets in the sense that these are not going to be here for the long
02:46term. You're not investing in a house, you know, that you expect to be here in 50 years. These are
02:51chips that will probably be replaced in a matter of a couple of years. And so that makes them very
02:56quickly depreciating. So, you know, maybe there is some reuse capability after a couple of years.
03:01But the point is, this is going to have to be infrastructure that gets upgraded very quickly.
03:04Yeah, I mean, speaking of that depreciation, again, so important to remember that here.
03:09Why are investors now betting that NVIDIA's chips can retain their value long enough to support any
03:14long-term financing? Well, I think that there's just so much hope for massive amounts of revenue
03:20in the AI economy. Look at what's happening with Anthropic, for example, and the incredible revenue
03:26report that came out this week there, which just illustrates, I think, to a lot of investors,
03:31just the mammoth amount of potential. And so the chips are going to underpin that revenue
03:37potential. Obviously, you can't have an AI model without the chips to develop, to make it possible.
03:43And so, you know, we've certainly seen data centers throughout the country popping up. And that's,
03:48of course, causing a backlash throughout the country as well, which is a risk to these data
03:52center projects and a risk also to NVIDIA's strategy of trying to make money off of them.
03:56Yeah, I mean, those chips, obviously, so essential for all of this to be able to happen,
04:01especially as we move forward and it just ramps up more and more. How does this financing model
04:05really work in practice? What role do Wall Street firms play in helping companies acquire
04:10and deploy NVIDIA's chips here? Well, NVIDIA made a deal with a slew of real blue-chip Wall Street
04:16firms like Goldman, for example, or BlackRock, just essentially stepping up to put all this capital
04:23in one pool together. I think what it shows you, though, is that the need for this capital is so
04:28huge
04:29that one company can't do it alone, even a company like Goldman with as much money as they have.
04:35And so this is the NVIDIA, of course, being the, you know, $5 trillion company that it is,
04:40has incredibly deep resources and can ultimately participate in this as well. When you're talking
04:45about the kind of profit margins, they have something like 70, 75 percent, unbelievably profitable
04:51companies. So they've got lots of money lying around. So they feel like, OK, we if we work together
04:56with these banks, we can put enough money into the same bin together. That said, the guarantee
05:01that NVIDIA is providing on this OpenAI data center in Ohio is not as large as some people thought it
05:06would be, which could be an indicator that they are being a little bit reactive to the concerns about
05:12the circular financing. It was notable to me that in his ex-post, Jensen Wong actually asked and answered
05:19a question about whether this is circular financing. He said, no, like kind of in a terse way that this,
05:24and that sort of suggests to me that the concerns are starting to reach the C-suite there.
05:30Yeah, it really seems like, again, there are so many pieces at play here, especially with that data
05:34center that you mentioned in Ohio. And again, this idea of this kind of circular notion of all of this,
05:40what does it really mean for NVIDIA itself? I mean, does this end up helping the company expand demand,
05:45reduce financing friction, shift some of the risk elsewhere? Where do we think this is going to go?
05:50Well, I do think that it, I mean, if, as long as OpenAI maintains a very viable business and revenue
05:57model, there really shouldn't be much of an issue in terms of paying back this lease, paying off the
06:03lease, which would then make sure that NVIDIA doesn't have to step in and provide the guarantee
06:07that they have offered. So as long as that maintains the case, that should be okay. But it's
06:13a highly competitive industry. And there's no guarantee that the current leaders will be the
06:17leaders forever. We see the AI companies introducing new models almost every other week at this point,
06:23because the competition is so intense. And when we talk about recursive improvement in the sense that
06:29the AI models are improving themselves at this point, that that can move so fast that who knows
06:34who could be the leader in a year or two from now. And so that's the risk. And I think
06:37for NVIDIA,
06:38you know, if they lose a massive chip supplier like an OpenAI, that could put dramatic effect
06:44on the revenue. I don't think anyone's expecting that to happen. But the point is
06:48that there are risks here. And there's no guarantee that NVIDIA stays on top.
06:54Yeah, I mean, there's such an important point to keep that competition in mind, especially as
06:58everything kind of ramps up and more and more companies are kind of getting in on the action.
07:02Do you think that we could eventually see NVIDIA GPU-backed loans or other securitized products
07:07traded more broadly on Wall Street, kind of similar to other asset-backed financing here?
07:12It seems like a possibility. The idea of securitizing the assets here, I could see that
07:19having appeal on Wall Street sort of as the next evolution of the AI boom. But, you know,
07:25when you talk about securitizing products like that, that starts to give me bad memories of the
07:29financial crisis and what happened with mortgage securities there. So I think the investors have
07:34to approach this with caution to ensure that they don't get caught up in the bubble there.
07:38Yeah, I mean, as we look to the future, as we're talking about all of these different
07:42things that have happened in the past, and we now use those to kind of inform what do we think
07:45is
07:45going to happen in the future, do you think that we're watching the creation of a genuinely new
07:50asset class? Or is this just a creative way of financing an extraordinarily expensive AI build-out?
07:56It could be a new financial asset class. But I think if it is,
08:00it's going to be one with high risk because of how competitive the space is. Again, you know,
08:06we almost see on a semi-weekly basis, the leaderboard on who's ahead in AI seems to change.
08:13It's almost like the race to make the playoffs in a major sport where every week the power rankings
08:18come up. It's a different team. It's a different AI team at any point. And the question is, could
08:24someone fall further and further behind over time? And all of a sudden, the assets that are tied to their
08:29market revenue model become less valuable? You know, then you're talking about trying to invest
08:34in an asset class with high volatility. So it could be one, but it could be a high risk.
08:39Well, we'll definitely have to wait and see what happens with all of that. But we really appreciate
08:42you being here and breaking it all down for us today. Nathan Bomey, business reporter on Axios,
08:47author of the Axios Closer newsletter. Thank you so much.
08:50Thank you for having me.
Comments