What’s the deal with OpenAI's deals?

0:00In 1898, John Nuveen had a vision to underwrite municipal bonds that would propel American infrastructure into the 20th century. More than 125 years later, that vision lives on. Today, Nuveen has grown into a leading global asset manager with expertise spanning public and private markets, still investing with a future in mind. Nuveen. Invest like the future is watching. Visit nuveen.com slash future to learn more. Investing involves risk. Principal loss is possible. Hi, everyone. It's Michela. Before we start today's show, I wanted to let you know that I'll be in London this December for the FT's Global Banking Summit.

0:39The event will feature CEOs, top regulators, and policy leaders as they discuss everything from geopolitical risk to exploring the future of AI in finance. And I'll be hosting a few panels myself. The summit runs from the 2nd to the 4th of December. If you want to register, you can find more information in our show notes, including a discount code for 20 % off. See you there. First question. If each of you had to describe OpenAI's new era in a word or two, how would you do it? Spend, spend, spend. OpenAI Unchained.

1:26It's been a deal crazy past few months for the artificial intelligence company OpenAI. OpenAI has signed a$300 billion deal with Oracle. NVIDIA saying it's going to invest$100 billion in OpenAI. The business recently signed a whopping$1.4 trillion worth of deals to secure more computing power. That means deals to get access to things like more chips, memory, and processing power to further its AI ambitions. OpenAI making a deal with AMD on data centers. Back of the day, Broadcom shares surging after inking a deal with OpenAI. It's become this tangled web of deals that tie the largest listed chip makers in the world, their fortunes into OpenAI's orbits, and we've never seen anything quite like it.

2:15And then to top it all off in October, OpenAI entered a new era, flipping its company structure inside out. This restructuring takes OpenAI away from its origins as a non-profit and closer to a conventional for-profit company. These computing power deals and the restructuring have set OpenAI up for its next era. But that has analysts, investors and journalists asking a lot of questions. It's over a trillion dollars worth of commitments, dwarfing OpenAI's revenues and raising questions about how the company will fund them.

2:53I'm Michaela Tendera from the Financial Times. Today on Behind the Money, I'm asking our OpenAI experts at the FT some of those big questions to get a better picture of what the company's future will look like.

3:26I'm here with our West Coast financial editor, Tabby Kinder, and venture capital correspondent, George Hammond. They've both been covering OpenAI for a few years now at the FT. Welcome to the show. Hi. Thank you. So let's start with OpenAI's decision to restructure the company, creating a for-profit entity called the OpenAI Group. George, how did we get here? OpenAI became the preeminent AI company by launching ChatGPT. And that product, with hundreds of millions of users, is incredibly intensive on computing power. So the company needs billions and billions of dollars to run it and train it. That has put them on this tightrope where they need to keep raising money from some of the biggest investors in the world, their big tech partners and venture capitalists.

4:17And that is at odds with their origins as a nonprofit. Right. We've talked about this on the show before. It was originally set up as a nonprofit research lab, which is, you know, pretty different from your typical Silicon Valley startup. Yeah, its mission when it was founded was to enable the creation of artificial general intelligence, all powerful AI for the benefit of humanity. It was always really strange that OpenAI has been a non-profit company for the last few years because we've seen this small startup with big ambitions to create an artificially intelligent chatbot become this hugely valued superstar company in Silicon Valley with billions of dollars coming in from investors.

5:10Now, the challenge with OpenAI's nonprofit structure was that it made it more complicated to raise money from investors. That's because those investors weren't receiving traditional equity in the business in exchange for the funds they were putting in. OpenAI raised a ton of money last year and this year on the condition that they would be able to transform into a for-profit company with those investors taking a traditional equity stake in the company, which they've not been able to do in its current iteration as a non-profit. If they didn't get the restructure done, then a lot of that capital would convert to debt.

5:47So the restructure has been this really closely watched process over the last year or so. So how does this restructure change things for the company? The restructure splits OpenAI into two parts. So there is the OpenAI Foundation, which is the original nonprofit, and that controls something called the OpenAI Group. The OpenAI Group does something very, very important, which enables investors to hold equity in OpenAI for the first time. Prior to this new subsidiary being created, investors were just entitled to a future share of profits. So that's the crucial separation that has happened, and that's what it enables.

6:30ultimately that unlocks a initial public offering for OpenAI. And it theoretically clears the way for fundraising to be much, much easier. Right. But just to point out, I mean, clearly they haven't necessarily had problems raising money. OpenAI is valued at$500 billion and it's the world's most highly valued startup. So what role does the restructure play in all that? As you say, OpenAI has raised more than$60 billion from investors to date. So clearly no issues raising cash. But investors are putting money into the company on the basis of a promise of future returns. And part of that promise is that AI becomes this hugely lucrative technology.

7:13But the other part is that they can cash out. And part of the investment into OpenAI has been on the assumption that at some point that will translate into equity. And that's what the restructure unlocks. Now, alongside this restructuring that happened, OpenAI was, as we've established, signing over a trillion dollars worth of deals for computing power. Can you tell me more about those? So OpenAI in the last few months has signed a flurry of just massive infrastructure deals with big chip makers and chip designers like NVIDIA, AMD and Broadcom and cloud computing providers like Oracle and CoreWeave to effectively provide them infrastructure to run AI over the next five to 10 years.

7:56How are these being funded? Is it cash, debt, equity? What's the deal with these deals? So far, we actually don't know how OpenAI intends to fund$1.4 trillion of deals. All of them have very limited financial terms attached, so there are lots of details that have yet to come out. Now, what are they planning to use all this computing power for? The phase of AI that we're in at the moment is this big boom in creating the infrastructure that is needed to power AI models. And that includes chips, data centers, acquiring the power, the land. And OpenAI's aim here is to get its hands on the chips and the data centers that it believes it will need to power bigger, more sophisticated AI models that it can run all of its products on.

8:50I mean, it's hard to kind of explain really how ambitious and aggressive they are being in trying to get so much new computing capacity online to power their AI models. And effectively, what they're doing through these deals is sucking up all of the power that will come online over the next decade for themselves. That's a lot to even comprehend hearing that. But as these deals were being announced, could you describe what the reaction on Wall Street has been? Wall Street responded really favorably to these announcements where big chip companies are doing deals with OpenAI. I mean, there was an immediate surge in the stock prices of Oracle, AMD, NVIDIA.

9:37Oracle, for example, it came out in the summer and said it had done a deal with OpenAI that would bring in$300 billion of revenues over the next five years. and their share price pretty much immediately jumped, you know, by over$200 billion in response to that news. So you both have been reporting on some of the concerns around these deals and the way that they've come together. And one of the main concerns that you've written about is that these deals are circular in nature. What does that mean exactly? Some of the deals have been structured to give OpenAI a financial incentive for buying chips.

10:16So, for example, NVIDIA will invest$100 billion of equity incrementally into OpenAI in return for OpenAI buying and deploying 10 gigawatts of chips, which in today's prices is about a$500 billion commitment to deploy those chips. There are similarities in the AMD deal that OpenAI struck last month. here opening IA buying AMD chips and in return they will get AMD stock. Now there is a world in which they could sell that stock to then buy chips from the same company, creating an obvious circularity. Yeah. What are the worries that come from this? I think the central concern with these deals is that they tie some of the biggest companies in the world to a heavily loss-making startup and they're a bet on that startup's ability to finance the deals.

11:11Tabby mentioned Oracle's share price going up by more than 200 billion on the back of that deal. AMD and others have seen similar rises. So if OpenAI can't make good on its commitments, then you would expect those share price rises to become share price falls pretty quickly. Yeah, the big concern is whether this is creating an artificial market for these chips that wouldn't have been there otherwise. You know, if you give me money and I use that money to buy something from you, is that a real market or is that some kind of artificially inflated, you know, manufactured market that makes demand look far higher than it would otherwise have been?

11:52So that's where some of the criticism of these deals has come from. Right. Now, I mean, is this kind of deal making even legal? Has this been deployed elsewhere? Is OpenAI the first company to do this? It's completely legal. We've seen circular deals before. In the late 90s and early 2000s, this sort of deal was generally called vendor financing, and it tended to rely more on loans than it did equity investment, whereas now what we're seeing is companies like NVIDIA taking a stake in OpenAI in exchange for them buying a certain number of chips from them rather than NVIDIA lending money to OpenAI.

12:37But what is definitely new is the scale with which OpenAI is going out to market here. I mean, this is$1.4 trillion of deals. So the share prices and the future revenues of some very big companies have become highly dependent on OpenAI, this relatively young startup, being able to follow through on massive financial commitments in a technology that is still relatively new and that has a long way to go before it starts to change the world in all the ways that people like Sam Altman predict it will, and also to make real money. So how does OpenAI plan to make money then and finance these deals? As you mentioned, it's loss-making right now.

13:28And it's worth pointing out that, George, this is something we talked about on the show just about two years ago. And we're still talking about it today. So what is the company's plan to turn a profit? OpenAI is building a lot of plates. They have CheckGPT, which is their central product, and they have subscribers to that. It's roughly 5 % of the 800 million people who use it. They're paying around$20 a month. So that's one big way that they are planning to make money. They are also looking to serve governments and businesses with more bespoke tools. They're exploring shopping tools. They're launching AI agents.

14:08They have a video creation service called Sora. and there are more plans beyond that. They've partnered with Johnny Ive, the former Apple designer, to work on a hardware device. They have a number of different ideas, some of which may take off, some which may not, to boost their revenues. And that is all well and good and this is part of a five-year plan to become profitable, which they think could see them pay off most of these commitments, but they really need to execute all of these product lines at close to perfection in order to get to that 1.4 trillion target. Yeah, that's a lot of pressure.

14:43And a lot has to go right in order to take the revenues they're making right now to over a trillion. Has the company or CEO Sam Altman addressed the concerns people are pointing out and the huge mismatch there between their current revenues and their spending? Well, he actually went on a podcast recently where this question came up. How can the company with 13 billion in revenues make 1.4 trillion of spend commitments? you know and and and you've heard the criticism we're doing well more revenue than that second of all brad if you want to sell your shares i'll find you a buyer i just enough brad gertz know who is the interviewer in that clip is one of the big investors in opening eye he runs a fund called altimeter and he was asking sam what felt like a very simple and obvious question and i think sam's defensive reaction there speaks to the fact that he doesn't entirely know how this is going to happen and he's asking investors to to trust him and come along for the ride we do plan for revenue to grow steeply revenue is growing steeply we are taking a forward bet that it's going to continue to grow and that not only will chat to be t keep growing but we will be able to become one of the important ai clouds that our consumer device business will be a significant and important thing that AI that can automate science will create huge value.

16:05Sam is saying we've done really well in the past and we're still doing really well and you just have to trust us Brad and I think that is very typical of Silicon Valley startup thinking but it's one thing when you're a billion dollar company saying that you might grow 10 times in the next year it's a very different thing when you're a 500 billion dollar company. Silicon Valley founders love this kind of you know grandiose, trust us, if you are a cynic, you just don't believe in the mission type way of speaking about things. But what we're seeing with OpenAI here now is, you know, very real money being committed to the balance sheets of very real publicly listed companies that make up a huge proportion of the stock market, the US economy.

16:53And that interview, as George said, you know, It came off as defensive, as Sam Altman not really having an answer. And it's that sort of thing that make people worried that we are in an AI bubble, that there is a lot of investors chasing opportunities, chasing hype, chasing vibes, but without the underlying foundational finances that we would need to see, to have faith that all of this is going to kind of come through as Sam Altman describes it.

17:29The future of fixed income investing will continue to blur the lines between public and private markets. Nuveen's credit platform was built to navigate this shifting landscape, channeling capital toward the businesses, infrastructure and energy solutions that are shaping tomorrow's economy, all while aiming to deliver the resilience and returns that portfolios seek today. Nuveen. Invest like the future is watching. Visit nuveen.com slash future to learn more. Investing involves risk. Principal loss is possible. Now, speaking of public markets, OpenAI has signaled that the company eventually wants to have an initial public offering.

18:08With Sam Altman calling that the most likely path for the company when it restructured, how close or far away do you see that being? And do you guys think that's a realistic path for them, as Altman says it is? I think an IPO is definitely a realistic path for OpenAI. There are a lot of public markets investors who would love to have a piece of the action with OpenAI. They have a few impediments to that at the moment. The first and most obvious is that they are still massively, massively loss making and anticipate being so over the next few years as they plow more money into research and development and into running their tools.

18:46So that's one issue. The other is that they have this formidable momentum and they've grown incredibly fast in the last few years. And that is appealing to public markets investors, but they may also want to see some consistency because momentum can come from new signups, people wanting to try this tool, but not necessarily sticking with it. And much more compelling to public markets investors is a sense of revenue that is really nailed down. So if they can figure both of those things out in the near future, then And yes, I suspect there would be a huge appetite to invest in OpenAI and an IPO.

19:21But they're big things to sort out ahead of that. Yeah, there is definitely huge appetite for OpenAI to IPO. I mean, people are clamouring to get their hands on the returns on investment that OpenAI could make investors. Whether OpenAI will ever get itself into a position on financial stability, governance, risk, to actually list on public markets, I think I'm more of a cynic. Now, what about OpenAI's rivals? I'm thinking about companies like Anthropic or XAI. How does what these companies are doing compare to OpenAI? The thing that unites all of these AI startups is the massive rush for computing power.

20:06So all of them, OpenAI, Anthropic, XAI, they are all desperately trying to get their hands on as much chips and data centers as possible. And I mean, these companies are using lots of different ways of trying to get the funding to buy chips, to build data centers. But yeah, this rush for computers is kind of the big theme that we're seeing play out. And it's why all of these startups are rushing to raise and spend as much money as possible. Now, with the amount of money swirling around AI right now, there has been a lot of commotion around a possible bubble forming, as you mentioned earlier, Tabby.

20:49Is OpenAI putting itself in the center of that bubble right now? Yeah, I think there's no doubt that OpenAI has become very much conceptually and realistically at the centre of an AI bubble. I think there is systemic risk here, or at least the potential for systemic risk, if we see all of the ambitions that OpenAI has laid out come to bear, purely just on a kind of power generation basis, the scale of the power that they are trying to build. It's making them a hugely powerful and important player just in the US economy in how GDP growth looks over the next few years. And a lot of big companies now have a lot of skin in the game on OpenAI's success.

21:37So is OpenAI setting itself up to be in a situation where it's too big to fail? How big of a risk is that realistically? I think where OpenAI differs from Anthropik and XAI is in the scope of its ambition. It is really trying to do everything all at once for consumer business, for enterprise, for government, building the data centers itself, as well as the products. So in that respect, it's trying to become the company that is synonymous with AI in the US. And that makes it very, very systemically important. So it's at the center now of a nexus, which includes the biggest companies in the world, the US government, consumers, other enterprises.

22:21And to that degree, it has become essential to the system or as essential to this system as any startup can be. So if OpenAI can't fulfill these deals, what does that look like? I think an investment in OpenAI is a bet that AI is going to change the world and create massive value in doing so. And I think there's two ways that could not come to bear. And one is that OpenAI fails on its own terms. And in that case, it's not the end of the world for the companies who've built all this capacity because others will step in to fill it. And those companies, Anthropik, XAI, we've mentioned, and then Meta, Google, some of the biggest companies in the world.

22:59So in that respect, it would be not so problematic. If the thesis that AI is going to change the world turns out to be untrue on its own merits, then it's a much bigger problem. And you have this build out of huge amounts of infrastructure, primarily data centers, which become redundant. And there was an analogy to the dot-com boom where a huge amount of fiber optic cable was laid and not used. And we've only just grown into that capacity 25 years later. So that's one risk, but that requires not just opening eye to fail, but really the whole sector to not live up to its promise. Well, and speaking of risks, are there any places where you're already seeing this excitement for AI turn into risk or potential risk rippling out into the broader public markets?

23:49Yeah, I mean, it's been interesting to see how quickly the conversation about where all of this money is coming from to fund this huge infrastructure build out. It started off very much being funded by big tech companies. They've got really healthy balance sheets, loads of cash on the balance sheet, loads of money to invest in these sorts of projects. And that's what they've been doing for the last few years. But very quickly, as the scale of these projects has got so large, their balance sheets aren't really ripe for spending this amount of capital. So very quickly, we've seen the conversation move to, well, a ton of this money is going to have to come from debt markets, public debt and private debt.

24:33So in the last few months, I mean, Google, Amazon, Oracle have all raised, you know, multi-billion dollar bonds, which is a huge pivot in terms of the financial strategy of these big tech companies. And I think raises a lot of different risks over the next few years if we do see demand, you know, peter out and if the projections that the companies are making about demand for these services, demand for these data centers just isn't right. So as reporters covering this story, I'm curious, you know, what are you watching for in the coming, I guess, let's say, year from OpenAI? I think for me, it's interesting to see how quickly the sentiment is shifting.

25:16And when it is shifting, the swings are big. We've gone very quickly in the last few weeks from barely anyone saying a bad word about AI and the future of demand and the potential of the technology to there suddenly being some very real bearish views out there in the market. We've seen a big sell-off in both tech stocks and bonds. So, yeah, I'll definitely be watching the debt markets to see how they're responding to the massive, massive spending and borrowing that we're seeing. And George, what about you? What will you be watching for? The perennial question with OpenAI is whether they can keep pace with their own ambitions and take investors and now potentially the public market and the public itself along with them.

26:04And that question is only getting more acute as their capital needs increase, their set of products they're going after expands, you know, as their ambition gets bigger and bigger. Can they tough up investors for more and more money? Can they start pulling some of this stuff off? So those are some of the moving parts that I'm watching. Tabby, George, thanks for coming on the show. Thanks so much. Thanks for having us. Behind the Money is hosted by me, Michela Tendera. This episode was produced by me and Safiya Ethmed. Fact-checking by Simon Greaves. Sound design and mixing by Sam Giavinko. Original music is by Hannes Brown.

26:47Topher Forges is the FT's acting co-head of audio. Thanks for listening. See you next week.

27:21returns that portfolios seek today. Nuveen. Invest like the future is watching. Visit nuveen.com slash future to learn more. Investing involves risk. Principal loss is possible.