SpaceX, Elon Musk’s rocket, satellite and A.I. company, is about to go public at a record-breaking $1.77 trillion. This summer, Anthropic and Open A.I. will follow suit, also with sky-high valuations. Are they worth it? The Opinion writer David Wallace-Wells and the contributing writer Natasha Sarin, an economist and law professor, tackle that question and discuss what these I.P.O.s mean for the American economy in the near future and beyond.
(The New York Times has sued OpenAI and Microsoft claiming copyright infringement. The companies have denied those claims.)
Below is a transcript of an episode of “The Opinions.” We recommend listening to it in its original form for the full effect. You can do so using the player above or on the NYTimes app, Apple, Spotify, Amazon Music, YouTube, iHeartRadio or wherever you get your podcasts.
The transcript has been lightly edited for length and clarity.
David Wallace-Wells: I’m David Wallace-Wells, a writer for Times Opinion and a columnist for The Times Magazine.
It really wasn’t very long ago that chatbots, L.L.M.s and big A.I. first got the attention of the public. ChatGPT launched in late 2022. An awful lot has happened since then, but even so, every few months we have another burst of commentary about whether this is all a big bubble, whether the leading A.I. labs are raising too much money and spending too much money, given how much they’re earning, and leading the whole sector, and maybe the whole economy with it, towards a crash. But we’re about to enter a new phase, because two of these companies are preparing for absolutely mammoth I.P.O.s.
SpaceX, too, which is both an A.I. company and a satellite company, is about to go public for a total value of $1.77 trillion. So, what are these I.P.O.s telling us about the risks of a bubble, about the state of an American economy so highly leveraged on A.I., and about where we might be heading in the future? With me is Natasha Sarin, an Opinion contributor and an economist and law professor at Yale University, where she also runs the Yale Budget Lab. Welcome, Natasha.
Natasha Sarin: Thanks so much for having me.
Wallace-Wells: So, let’s start with a really naïve question: Why are these companies doing I.P.O.s right now?
Sarin: So, if you think about these companies — and SpaceX, OpenAI and Anthropic are all essentially trying to go public within just a few months of each other in the second half of this year. And they’re at the scale, as you’re describing, David, which is kind of unheard of. It’s hard to understand what to make of these multitrillion-dollar valuations. But one way I’ve been thinking about them is, if you take the three of them together, the sort of expected market cap of these three I.P.O.s is going to be something like over $3 trillion.
And if you look at, essentially, all of the technology I.P.O.s in the internet boom — so, from 1995 to about 2000 — and you combine their value entirely, all of them, and this is inflation-adjusted, SpaceX alone is almost as large as that. And the three of them together are significantly larger than that.
So, these are huge I.P.O.s, and part of what is going to happen as a result of that — and part of what the motivation is, if you’re thinking about why these companies are deciding to go public now — is about access to capital. It is about being able to sell stakes in this company to a broader pool of investors, and being able to have the valuations attached with that, and the public market valuations attached with that. And I think that’s really a significant moment — not just for these particular companies, but for all that portends with respect to artificial intelligence more generally.
Wallace-Wells: So, just drilling down on that for a second. I mean, you’re saying that the motivations here are kind of classic I.P.O. motivations. You know, raise money for the companies, liquidate cash for the investors, make some amount of social compact with the public by getting them some slice of the pie. But is it the case that Anthropic is unable to raise money now in the private markets? Like, why turn to Wall Street?
Sarin: Yeah. It’s such a good question, and in part, we know the answer to that is no, right? And so, if you look at something I’ve written extensively about over the course of the last year or so, it really has been the growth in private markets, and particularly in private credit markets, which involves a lot of, traditionally, private equity firms that made equity investments. Things like Apollo, things like Blackstone, actually making loans to a lot of these same companies that are powering their growth or powering the data center infrastructure build-out.
And so, private markets are keen and eager, and stand ready to invest very substantially in these firms. But, to your point, that’s not the only motivation for why these companies are deciding to go public at this moment. There are lots of motivations for why they’re deciding to go public. One of them has to do with the ability of their own investors in the company to realize the benefits of these multitrillion-dollar valuations that they’re essentially going to be forcing on the market. If you take SpaceX, which is coming this week, and you take the valuation that they’re assigning to themselves, which is this record $1.7 trillion that they’re going to go out with, that’s significantly higher than a lot of what analysts are assigning to it, right?
Wallace-Wells: And they’re actually trying to relax some of the conditions to buy, right? To allow smaller investors in, which tells you something about the kind of buyer that they’re hoping to sell to.
Sarin: And there’s so many pieces of this that, I think, are so important for how the market functions, and for actually regular people and their portfolios, that are really important for your listeners to grapple with. One is that, part of what is happening, just because of the scale of these I.P.O.s that we started by talking about, is they’re going to instantaneously be a really significant part of household portfolios through things like their retirement accounts. And the reason for that is, because as these giant I.P.O.s are happening, the stock market indices are actually relaxing their own rules with respect to how long it takes in order for companies to get onto the index.
And so, the idea that 15 days after SpaceX I.P.O.s, it’s essentially going to be part of all of our retirement portfolios; and all of our index investing is actually really consequential for households. And in general, these I.P.O.s and the concentration that they’re going to represent — even before you had these I.P.O.s, by the way, something like 60 percent of stock market growth last year was about just a few technology companies — that trend is going to accelerate in a world with these mammoth I.P.O.s coming in the second half of this year.
And so, that too has really consequential effects for household portfolios, because, in some sense, it means that we’re all massively exposed to the idea that there might eventually be — and if what history tells us is true is, in fact, true, that when you have these types of technological changes, even ones that are hugely beneficial, and bring a lot of welfare, and a lot of economic growth — things like the internet, things like the railroad — they come with a bubble that eventually pops, and households are going to be massively exposed to that. And more exposed now that these companies are going public than they were a few weeks ago when all of them were private.
Wallace-Wells: So, one thing you’re talking about is that we’re all, collectively, going to be pulled into this investment cycle. And I wanted to pull back, a little bit, from the question of the I.P.O. and talk about the political economy of A.I. at the moment, generally; because, at the same time that we’re having the prospect of all of us getting enlisted in this profit machine, we’re also seeing huge amounts of public backlash, particularly around data centers. There’s broad unease about the future of an A.I.-powered economy. We see Bernie Sanders proposing the American government taking a 50 percent ownership stake in these labs. We see Donald Trump making similar noises, and the A.I. labs themselves saying, we’re open to this; kind of like, let’s talk about it.
And on one level, this is, for me, quite strange to be happening at the time that these companies are going public. We’re simultaneously ramping up for this huge distribution of ownership to the public as we’re contemplating the federal government coming in. These companies, which in certain ways are flush with cash, are nevertheless trying to bring more of the public on board to their project, presumably to protect themselves, to stabilize, or buffer themselves against public backlash and to make their proposition about their own role, their own large role in the future of the economy seem more palatable to more Americans. How big a part of this story do you think that is?
Sarin: Yeah. You can totally understand why there is a fair deal of nervousness about these companies, and, more generally, about what artificial intelligence is going to mean for all of our futures. Michelle Goldberg has a piece on how, if you look at commencement addresses over the course of the last few weeks now, you’re hearing a lot of backlash from students who are graduating out into an economy where youth unemployment is starting to tilt up. Again, that’s not really about artificial intelligence. That’s kind of what happens at the end of a long credit cycle where there’s been a lot of money flowing into the economy and a long sustained period of growth.
Wallace-Wells: It is striking that young people do seem to be skeptical of the A.I. future, which is not what you would anticipate.
Sarin: Totally. If you look at where artificial intelligence is having a very clear impact already, the life of a college student on any campus in this country is a great place to look for where you are actually seeing, day-to-day, the educational experience of students being impacted in a way that, frankly, as someone who teaches at a university, I feel I’m watching it in real time. Like, how do you deal with the fact that this obviously is going to bear on the educational experience of those students and ultimately, sometime down the road, the labor market outcomes of those students? If you take what Dario Amodei or Sam Altman has said — Dario, in particular, has been of the view, over the course of the last few years, that we’re going to displace a very significant share of white-collar workers as a result of this technology. And so, again, it doesn’t really feel clear what path the future holds.
And as a result of that uncertainty, if you think about it from a policy perspective, it’s also very difficult to think about how we should actually envision regulating, ex ante, these types of technologies, these types of models, and ultimately these types of firms. And how do we ensure that the revolutionary potential that they have is harnessed for good and harnessed for progress, and harnessed for economic growth, as opposed to some of the real risks that we know that the same technology in fact poses.
And so, going back to, is this part of why these companies are going public, or is this part of the rationale for why the discipline of public markets — both access from an investment perspective, but also, relative to private alternatives, the transparency that comes with some set of reporting, that comes with public valuations, that comes with tradable shares — maybe that is, in fact, part of the story from the perspective of these firms. And, in that sense, it’s sort of a more benevolent story than the one you’re telling that is really about like, once SpaceX goes public, Elon Musk’s on his way to being a trillionaire.
Wallace-Wells: Well, I think, at the moment, a lot of Americans look at the A.I. companies and do see an especially vivid illustration of the plutocratic structure of our society, right? They see these five companies; they’re run by these five visible people. They’re all worth an unbelievable amount of money. And to the extent that we are imagining futures being dictated by the companies themselves, that can be quite scary.
And, to some degree, going public and having government stakes in the companies both address that problem to a certain extent. It would mean that the country, as a whole, is invested in the success of these labs and may benefit to some degree — although at what scale is an open question — from the success of the company. But there are other ways in which some of these approaches — public offerings and/or government investment — don’t change the dynamic. Which is to say — maybe, most notably — if this is a bubble then it’s the public that is left holding the bag.
It’s not like once you go public, you’re on this glide path to huge future profitability. It may actually be more likely that from the point of initial public offering, the companies lose money at least for a period of time. So, what do you think about that prospect? The prospect of a genuine correction, a bubble popping?
Sarin: You know, part of what makes me somewhat nervous — and should make everyone nervous — is that it’s not like you and I are alone in our view that, oh, we might be on the verge of a bubble, a bubble might be on the horizon. Last summer, Sam Altman was asked some version of, “Is this an A.I. bubble?” And he said: “Are we in a phase where investors as a whole are overexcited about A.I.? My opinion is yes.”
And another thing that should make us somewhat nervous is: If we look at history, if we look at every large technological innovation that has changed the way that humans work and the way that we all live — most recently the internet, but if we go back to railroads, whatever moment you want to look to — there is a very predictable, in some sense, cycle that you see, in terms of what happens to the economy at those moments of technological change.
Everyone sees the emergence of this new technology and gets really excited about it and its potential for massive change. Investors see that, too, and money rushes into this new technological prospect. And it rushes in productive ways, but it also rushes in ways that ultimately don’t end up being that productive. So, this is, if you think of examples during the internet bubble, like the growth of everything, every company that had .com attached to it. That doesn’t take away from the fact that the internet actually did change all of our lives.
But ultimately, what happens is that the bubble bursts and a bunch of debris is left behind, and that isn’t just about a couple of companies that ultimately fail. It is about what that means from the perspective of the broader economy that we all inhabit — in that, often, those corrections come with deep economic downturns and have the consequence of having large-scale unemployment, having an economy that isn’t growing quickly, having the need for the government to step in as a potential backstop.
And so, from my perspective, the question isn’t are we in a bubble or will the bubble burst? The question is: When?
Wallace-Wells: Yeah, one thing that I think about in this moment, when thinking about the I.P.O.s and what justifies these massive, massive valuations, is: These are five companies, and three of them are going public. In the public imagination, they do dominate the A.I. landscape. But of course, they are only providing one set of products, which is to say access to their L.L.M.s; and they’re providing it in different ways at different price points, at different tiers. But it seems to me like the massive boom story that they’re trying to tell is one that’s a little bit of a holdover from an earlier era of A.I. thinking, in which the companies and the people who are designing the products often talked about artificial general intelligence, artificial superintelligence, and they said that these products are improving so much that at some point they’re going to be able to improve themselves recursively without human interference.
And at that point, there’s going to be a kind of a takeoff in which the products themselves, the companies that made them — and to some extent the economy as a whole — would be rendered almost unrecognizable to people living on the other side of it. Some people call this “the singularity.”
But I wonder how much that still feels true today. And what I mean by that is, I was just looking at some data today, that just over the course of this calendar year, 2026, the amount of use of Chinese open-source A.I. models has tripled, while the use of the American A.I. products has basically flatlined. We see a lot of companies — Uber was maybe the most high-profile one — saying, “We’re actually winding down our employees’ use of A.I. because it was too expensive, given what we were getting out of it.”
And so, if we think about a future in which there’s going to be a superintelligent Borg running the whole economy, then yes, racing to be the biggest, best monopolistic A.I. company is hugely important and it does justify these absolutely gargantuan valuations if you believe that, for instance, Anthropic will be the one to win.
But if you’re thinking about a world in which, yes, A.I. is everywhere, yes, everyone is using it, but it’s not totally clear how many people think it’s super important to pay a huge premium to buy the absolute best-in-class model. And how many more people are likely to think, “I can use this open-source product from China that’s 80 percent as good as Anthropic’s first-rate model and pay only 5 percent of the price.” That’s a very different world.
The A.I. companies used to talk about building a moat — what they could do to secure their advantage. And they thought that getting to something like A.G.I. or ASI faster was the main way to do that. In a world in which that’s at least not imminently on the horizon, and we have all of this low-price competition from below, isn’t it the case that these companies are at some real risk of expecting much, much higher returns than they are likely to get in the medium term?
Sarin: Yes, 100 percent. And I will say something that has given me a fair bit of nervousness around A.I. and the ultimate possible profitability of these companies. ChatGPT was, as you were pointing out, launched in the fall of 2022, which feels like yesterday, but was less than four years ago, you know? But I guess it’s all relative —
Wallace-Wells: It’s both at once. It’s like a whole different era and the same.
Sarin: If you think about that moment, it feels like we’ve gone through many chapters. The case against A.I. was coming from outsiders to the technology: Doomers or short sellers who were betting against it, or Luddites who just couldn’t possibly think about the sort of transformational potential that existed. And the new skeptics are coming from inside the boom in some sense, because, as you’re describing, it’s like Uber capping A.I. usage in three months or four months over the course of this year.
Or you have a bunch of these companies, by the way, like GitHub, moving Microsoft Copilot to usage-based billing because of how costly it is, in order to deploy the technology in ways that, as they were starting out, they didn’t fully appreciate. And if you look at a bunch of these consulting firms, they have started to do surveys of companies asking them about their own A.I. usage, because the optimistic case of the world hinges on the idea that this technology’s going to be so revolutionary so quickly that we’re going to get all this productivity growth. In fact, we’re going to displace a lot of labor. And they’re essentially finding that the technology is working, and we all experience it. It’s working in new and better ways over time. But that value proposition hasn’t yet arisen for the firms themselves that are trying to deploy the technology.
Again, that’s not to say that that productivity growth isn’t ultimately going to come on the horizon, but it is to say that over the short and medium term, I think companies and the economy writ large are still in figuring-out mode with respect to what exactly it means to deploy A.I. in its most optimistic, most growth potential, most productivity potential way.
And flip side, for a while we were all talking about, and we were hearing a lot about, the idea of singularity or A.G.I. as this gold star that was coming right on the horizon. And now you have people — I’m using Sam Altman because he’s spoken publicly about this recently in ways that have gotten a fair bit of attention, but he’s not the only one saying this — where they’re talking about A.I. and describing it, even internally themselves, as not really all that useful of a term; and kind of describing it as not some sort of magical switch that’s going to flip on at some moment in the short horizon, but instead as the idea that these models are over time going to continue to get better and more useful and more transformational. But that’s not something that’s going to happen instantaneously.
Wallace-Wells: But even the way that you’re talking about these questions is illuminating to me, because you’re talking about, on the one hand, the big A.I. companies, and then the firms that are using them. And when you’re talking about productivity, you’re focusing on the firms that are using them. But these are two separate questions, right? If OpenAI and Anthropic are going to justify trillion-dollar valuations, or even larger valuations, they’re going to have to make a lot of money, too. Even if tons of people are making money on A.I., it has to be in these companies to justify the value.
And when I hear Sam Altman talking about the possibility that, in the future, A.I. will be like a utility in the same way that we pay for our electricity, I think to myself: The electric utilities are not worth a trillion dollars. This is a technology which absolutely has huge transformative potential, but to me, the question is: How much of that is captured by these companies?
Sarin: It feels like both an unanswered question, and an inherently, frankly, unanswerable question. But also, it should make you even more nervous about this bubble conversation that we were having because — and Ray Dalio said a version of this last week — if you’re thinking about it from the perspective of these firms, you have to spend a ton of money and justify these valuations, not just because you’re worried about, like, is this a good way to deploy resources, but because you’re worried about losing market share.
If you’re of a view that the way this all shakes is that there’s going to be one, two, maybe three large players that are able to capture the market, you have to try to be one of them. And that results in, frankly, the incentive structure to spend a lot, and to look like you are doing a lot, in ways that might ultimately not be tied to fundamentals with respect to investment opportunities and what is profit maximizing from the perspective of the firm.
So, you should be worried about that. But there’s another piece of this, which is that the companies themselves are asking public investors to pay prices at valuations that assume that A.I. is going to reshape the economy; and to pay those prices at the same time as these companies themselves haven’t figured out how to stop losing money; and at the same time, as these companies themselves haven’t figured out how they are going to be the ones left standing at the moment when A.I. ultimately is a developed technology with a developed set of market players that we all have grown with and understand. And I think that is something that is just so striking about this moment.
Wallace-Wells: So, this has been a relatively skeptical conversation about the I.P.O. cycle, at least. And I wanted to close, because of that, by asking you to tell us, what is a version of the story that we could be telling two years from now, four years from now, in which that skepticism looked naïve? In which, actually, there was no bubble, these companies did earn these valuations and more, and we were looking back and thinking, “Why were Natasha and David so skeptical? We should have known that all of this was happening.” What would be required for that to unfold?
Sarin: I should specify my skepticism, in that — and I think this is your view, too, and I’m curious if it is. I am actually not skeptical of A.I.’s transformative capacity, in part because I, like you, have been living with it over the course of the last few years and have seen how much it has changed my own life and my own work. I happened to be traveling last week, and used Gemini to try and figure out a walking route to allow me to see all of the sights of Madrid in an afternoon despite the fact that the pope was visiting, and boy, was Gemini incredibly good at doing that. And so, I think it’s great. I think it is really so phenomenally impactful.
Wallace-Wells: But in the context of this conversation, you don’t need world-class A.I. to do that for you, right?
You need a pretty good A.I. to do that, especially because you’re asking a kind of generic set of recommendations. It doesn’t require that much customization or personalization. It’s basically aggregating and presenting to you, in natural language, the same kind of result you might have gotten a few years ago from a search engine, right? And that’s really useful. But the question is: How much are you going to pay a month for that value?
Sarin: For that capacity, yeah. And how much extra are you going to be willing to pay for the best version of that?
Wallace-Wells: Or, what number of people are willing to pay that?
Sarin: Totally. And so, again, this is not skepticism then about the technology, and it’s not even skepticism about the technology’s ultimate impact on productivity, where I think partly we’re being a little unfair to A.I., in that we’re in early innings. If you look at the internet and its impact on productivity writ large, there was a famous saying by the economist Robert Solow, who said that “you can see the computer age everywhere but in the productivity statistics.” And so, I think that’s probably a version of what you’re likely to see here, which is that it’s going to take some time in order to ultimately have the productivity growth from A.I. unleashed. And the story of the optimistic versus the pessimistic case is going to depend on what the horizon is for that type of productivity growth. And it is going to depend on what type of market share is ultimately controlled by these few very large, currently leading A.I. labs.
Wallace-Wells: Yeah, I often think about that Robert Solow quote. There’s a related one that Paul Krugman gave in 1998. He said, “By 2005 or so, it will become clear that the internet’s impact on the economy has been no greater than the fax machine’s.” Obviously, the internet has transformed American life, it’s transformed the American economy; but it’s also, in total, given us a boost of maybe half a percentage point of G.D.P. growth a year.
Sarin: Which is a lot.
Wallace-Wells: It’s a lot. It makes a huge difference in human well-being, especially over long horizons. But compared to the stories that we, as a public, and in particular the leaders of these A.I. companies have been telling us for years, it seems really paltry. And I think there’s something quite weird about the way that we’ve conceptualized this transformation in our lives. We’ve basically told ourselves that we’re either heading on a path towards superabundance in which labor is over — and that may be disruptive, but it’s going to be completely a different world in a relatively short order of time. Or we tell ourselves that it’s all nonsense. These people are selling us fish oil and it’s all completely worthless — a scam, self-dealing, etc. The likeliest outcome is in the middle, in which the world is transformed. But is the world going to be transformed in a way that justifies the growth stories that we’ve been told?
I’m not sure, and this episode in particular — the I.P.O. episode — is striking to me, because we used to think, as you were suggesting earlier, the market has this disciplining function. But these days, the people who are selling the stuff into the market, the people who are proposing buying that, all of the analysts, they seem to be buying the incredibly dramatic story of growth much more simplistically than I would’ve liked, and not applying the same level of skepticism that I would’ve expected from market analysts.
And to the extent that we expect that those valuations will be roughly met by the market, it means that the public is accepting those stories. And whether or not we end up, five years from now or 10 years from now, living in a world transformed by A.I., there’s still this big question of whether the growth in the economy and the growth in the profit rates of these particular companies will justify the story that we’ve been told at anything like a level that earns back to the investor.
And if that doesn’t happen, if we’ve gone into a phase in which the public on the market level takes a large ownership stake —
Sarin: And our retirement accounts, right? Which are automatically going to take a large ownership stake.
Wallace-Wells: If we end up as leveraged on these companies as these market valuations suggest, that’s a lot hanging on the success of these five companies.
Sarin: Sam Altman said, “When bubbles happen, smart people get overexcited about a kernel of truth.” And so, here’s the kernel of truth: This stuff is transformational. It is changing the way we work, the way we live. But he also said that when bubbles happen, “someone is going to lose a phenomenal amount of money.”
And part of what gives me a little bit of pause about the I.P.O.s and the valuations, is some of what we’ve been describing. I think I have a bit of nervousness that comes from the unknown and relatively novel and relatively vibes-based approach that it feels like these valuations are falling prey to. I think that we should all have a bit of pause, because it does feel like if you just take a set of fundamentals, we’re not really priced relative to what the outcomes that feel like they’re reflected.
Wallace-Wells: Yeah, I mean, one thing that I think about there is, take the example of Tesla. It’s not like fundamentals are driving that share price in general, right? I mean, there are a lot of companies that are able to sustain market interest over long periods of time without actually justifying it on the fundamentals. And so, we may be in a future in which these propositions don’t come to pass, the companies don’t gain monopolistic positions, are not earning huge profits, and yet, in the market, they’re treated as the new kings of the economy, and we just have to sort of —
Sarin: And that could sustain for quite some time, right? So, we’re not telling people to go short SpaceX this week, because, in fact, who is to say when, how, if the market will correct itself?
Wallace-Wells: Who’s to say how deranged the American investor is? Natasha Sarin, thank you so much for the conversation.
Sarin: Thanks so much for having me.
Thoughts? Email us at [email protected].
This episode of “The Opinions” was produced by Jillian Weinberger. It was edited by Kaari Pitkin. Mixing by Carole Sabouraud. Original music by Sonia Herrero, Pat McCusker, and Carole Sabouraud. Fact-checking by Mary Marge Locker and Kate Sinclair. Audience strategy by Shannon Busta and Kristina Samulewski. The deputy director of Opinion Shows is Alison Bruzek. The director of Opinion Shows is Annie-Rose Strasser.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post The A.I. Bubble Is Coming for Your Retirement Account appeared first on New York Times.
