AI’s next big test: the public market

0:00Hey, everyone. This is Aria. Just wanted to let you guys know that we recorded this episode before the SpaceX IPO, as there's some discussion of SpaceX in the show. Enjoy. Reid, it is great to be here. And for listeners, a quick heads up that today actually marks my last Reid Riffs until the fall. I am taking the summer off Reid Riffs, but Reid Riffs isn't going anywhere. It's just getting some new company. You're in excellent hands. And so I'll just say to people, get excited. We have some incredible guests coming this summer on Reid Riffs. All right, so jumping right in. Over the past month, it feels like the entire AI industry has essentially moved towards going public all at once.

0:39Alphabet pulled off the largest equity offering in corporate history,$85 billion, with Warren Buffett's Berkshire Hathaway taking$10 billion of it. We saw that SpaceX, of course, priced its IPO. They're targeting a$75 billion raise at a$1.77 trillion valuation. And part of that pitch was actually AI data centers in orbit being a core part of their business. We saw that Anthropic officially filed with the SEC on June 1st, and OpenAI followed shortly thereafter on the 8th. And so this combined 2026 AI pipeline could exceed the total proceeds from every U.S. public listing since 2022. So this is potentially massive.

1:24So my question for you is, when these frontier AI labs answer to public markets, does that change what they build, how fast they build it, and even what they'll be willing to say no to? It will, of course, be one of the marks that once the public infrastructure, public investments are all part of this, it'll change the dynamics around AI a lot. Your questions, of course, are one of the things, which is, you know, the incentive is if you're going to be, you know, kind of trillion-dollar companies, you have to be kind of reflecting the kind of margin structure, revenue structure, growth structure that, you know, kind of call trillion-dollar companies.

2:08And by the way, the volatility of your swings to up and down will be much higher because back when we were talking about the significance of billion-dollar companies or$100 billion companies, and you go, well, today, hyperscaler company number, letter X – actually, now that we have SpaceX coming out, letter Z – you know, gained, you know, 10 billion or lost 10 billion today in the public markets, now it's going to be even, you know, much more significant. And so that will, of course, drive a focus on business, on, you know, kind of revenue, on operating margin. That'll be important. Like, you only get these kind of numbers, you know, by being global companies and what plays out to global and the kind of hostility of the current administration to globalism will be a problem here because actually, in fact, as you do trade wars and tariffs and all the rest, this will be a natural place where all of this will play out.

3:22And, you know, there will be questions around, you know, kind of like, okay, so how much of this is the AI companies, you know, kind of providing AI infrastructure, like, you know, capabilities across whole industries? And how much of it is it going in industries? And this is all the way back to in the year 2000. Disney listed Microsoft as one of its top competitors because it's like, oh, God, there's one big tech company back then. And I'd say a little bit of the prediction that I made a number of years ago, which was they said, oh, we're worried about kind of antitrust of the top tech companies.

4:07And I was like, look, if we were five top tech companies heading to three, I would be concerned. I think we're five to seven heading to 10 to 15. And I think that's what is clearly being shown. NVIDIA has been added. OpenAI will be added. Anthropic will be added. SpaceX will be added. And so more of these large tech companies competing with each other is the zone which creates space for entrepreneurship, creates the questions of balancing out where you don't have kind of – because the key companies are competing with each other, you don't have the kind of antitrust considerations that is the base of it.

4:53Like what enables entrepreneurship? What enables new industry? What enables kind of cheap pricing for consumers? What enables competition for providing services across industries? And I think that will all become part of these being businesses. Now, I think there's a subtle point here that I think is particularly important that I think no one's really commented on, which is part of the benefit of these companies becoming public is it means the public participates in the economic benefit. Pension funds buy in or as part of it, 401k funds do this. And the public participation in the equity side of it then becomes one of the countervailing things about like, well, what is the success of the AI industry doing for the broader society, for kind of everybody in society?

5:51And the answer is, well, everyone starts participating in the economic growth and benefit of these industries, which then becomes part of the narrative and direction on, well, what are the AI companies doing for society in addition to providing, as you know, I say, look, actually medical assistance and legal assistance and educational tutors and better information systems for everybody and all the rest. What are they doing? Well, actually, in fact, as they become part of the infrastructure, they're also providing a distributed economic basis to all of the different kind of infrastructure that does public market investing.

6:33And I think that will be an important also corollary to what's going on. Reid, I think that's so important because when you look at the Google IPO, everyone was like, oh, my God, the Google IPO made so many employees, millionaires, and made investors so much money. But if you actually look at the arc of history, most of the wealth created by Google actually happened post-IPO. And so to your point, it went to people's 401ks and pension funds and all these things. And people are worried that as companies stay private longer, it's actually just the people who have access to the private markets who are getting rich instead.

7:07So but a sort of question on the flip side of that, a lot of people are talking about SpaceX and their S1 claims a$26 trillion total addressable market in AI. And they sort of need that to justify their sky high pricing on relatively low revenue. Even Ben Thompson called the numbers obviously absurd. Do you think the SpaceX capital story becomes its own kind of risk? And perhaps for people who have money in index funds or 401ks or things like that for everyday Americans? Well, if you look at it and you say, hey, I look at the numbers of open AI, I look at the numbers of Anthropic, and I look at the numbers of SpaceX, how should you allocate money across it just as a comparison to be part of the AI future?

7:56And, you know, SpaceX self-describes that its XAI effort is essentially a three-time failure catastrophe, that they're essentially selling their data centers, you know, the data center compute which they built to Anthropic and to Google because they're incapable of doing anything else. So it's just like kind of data center, you know, kind of provisioning. And you'd have to look at like how does it compare to other data centers? And, you know, my glance at it looks like it's like, you know, seventh or eighth in capability and competence in that list. So it's uninteresting. The profitable part of it, Starlink, which is amazing, but it's a tiny set of the whole thing.

8:48And so it's kind of trying to position itself as an AI company. Like if I had$100 to invest across those three, I'd probably put one into SpaceX and 99 in whatever balance you want into open AI and Anthropic in terms of where I would be putting money. because you kind of look at the growth of revenue, the operating margin, the capabilities of, you know, kind of what's playing into, you know, kind of like what is the AI growth versus a, hey, we're selling an old data center, you know, to, you know, availability of compute to Anthropic and Google. you know like that's how I would allocate money across this just speaking as looking at the financials and you know like it's you know obviously there's the whole magic of Elon that's going to cause a bunch of people to to to put in money but like you know ultimately when you get to these kind of numbers it's not a question of of do you believe in the magician it's a question of where is the AI future going?

10:03And so I think it's a, you know, the AI IPOs are, you know, kind of open AI and anthropic, not really SpaceX. Well, the show is an investment advice, but for those Americans who do have index funds and just do passive investing, I'm not going to lie, that makes me nervous that SpaceX is going to come crashing down, but we will see. So good to know where you would allocate your funds. All right, moving to the topic that always has to be at hand when we're talking about AI, which is jobs. So we've all seen that recently, you know, Meta, Amazon, Oracle, they're cutting tens of thousands of developer jobs.

10:43And Stanford's HAI, where you are on the advisory board, they found that developer employment for workers aged 22 to 25 has fallen nearly 20 % since 2024. But then on the other side of it, it feels like small businesses who have sort of never been able to have in-house technical capability are gaining those jobs for the first time. And Fortune wrote that their argument was that AI is enabling custom-built software that's super close to the business. So these people understand the business and they can use AI to become these amazing developers. And so it's fit to a single company's workflows and data.

11:20And people were also saying that actually the reason for sort of job loss or job cuts or not hiring for 22 to 25-year-olds, it had more to do actually from work from home than it had to do to AI. But regardless, we're clearly going to see like changes in the labor market. Is it going to be less developers? Is it going to be more developers? Is it because of AI? Is it because of work from home? It's sort of hard to predict right now. Can you walk me through what you think is going to happen over the next five years, specifically for developers in this market? Well, so the first is, you know, very clearly within a small number of years, what the software engineering job is going to be is managing and working with a bunch of coding agents.

12:07And this is a lens into a lot of other kind of knowledge and production work in terms of how it's going to operate. Now, you know, there's some people who then think there won't be such a thing as software engineers and like AI will be all the software engineers. And that, you know, like there was very fashionable a year or two ago to say, you know, like software engineers will all be out of work and so forth. And, you know, part of what I said then, and I, you know, triply say now is like, no, actually, in fact, there is essentially infinite demand for software engineering. And even with the software engineer tools, what it does is massively change what the job is of software engineering.

12:47So if you said my capability was that I actually knew the language direction of Python and I could do the semantics of typing it out and making it work and that was my unique ability, that ability is completely taken by AI in terms of how it operates. And by the way, that's today. It's not distributed across the whole thing. The future is already here. It's just unevenly distributed. And so that job doesn't exist in the future. But the mindset of how you do software engineering, how you build things, all the rest, that does exist. And by the way, that's a little bit like Jevin's paradox, which is there's infinite demand for it at a certain price point.

13:32Now, I do think kind of like what is the economics of software jobs and does the software engineering job now look paid like a lot of the people who are paid at Accenture and so forth and kind of accounting and so forth? There's a whole swath of that. But yeah, that's still a reasonable job. It's just no longer the – it's paid somewhat less than it was paid before in all of these sort of things. But there's still infinite demand for it. And what's more, the people who are hiring this isn't just like the Accentures and the other folks. It's also like small businesses and every other non-tech industry doing this kind of saying, hey, I can now hire some software engineers who are managing the, you know, kind of the AI development for what I needed for software engineering for my business.

14:23And it isn't just like front-end apps. It's like workflow management process. It's kind of collaboration. It's analysis of how our business is going. It's like all of this kind of thing, just like, for example, the amplification of, hey, I want to do financial modeling with Excel macros. Then those Excel macros now is like managing AI agents in doing all of that. And that becomes the thing versus the knowing the language of the Excel macros as a way of doing it and getting it constructed. And so I think that's what that is now meaning. And what the job transformation is, is, oh, I went and learned how to do Python, and now I'm no longer, like, they're not hiring a whole bunch of Python code monkey, you know, jockeys.

15:13what you should be doing is here I'm now diving into how I use AI to do this. I'm now much more productive. I now have a whole bunch of capabilities that I didn't have before. And it's that mindset of how I am essentially operating in terms of it. And that's the nature of it. And this kind of fits into the prediction of how is your job going to be taken? It's with another human using AI and you should be that human. And I think this is where the coding improvement has now kind of rung the bell first. Because by the way, coding is kind of most essential. And most developers are kind of like, they like the kind of mode of thinking.

15:58And I've seen so many people being refreshed by, oh my God, I've got these new superpowers with clawed code, with Codex, with CoPilot. And these are the things that I'm getting into doing. And that's the thing to kind of make happen. And I think that's the place where we are. And by the way, when you get back to young people, part of the young people is, this is your chance to say, I am AI native. You've got this insanity of, oh, I'm booing people at an AI commencement. And it's like, no, this is your opportunity. Your opportunity is, I am AI native. I should be helping. Like, you need to have AI fresh, AI open-minded, AI engaged people in your work.

16:50I can be that person. This is literally the disruptive opportunity. And it's like, look, if you're like, hey, I'm refusing to use AI in anything, then you're like saying, okay, no, no, I'm essentially going to go ride horses and be the horse taxicab kind of person. And it's like, that's up to you. You can decide to do that, but this is where the world is going. Right. This is where we're going. You got to get on board. And one consequence of this, if more companies have more software developers and more people are writing code, and when I say more people are writing code, more people are organizing and orchestrating AI to write code, we might have the same consequence that we're having in content all over the internet.

17:34We've seen e-books explode because people are using ChatGPT and Claude. We've seen AI music explode. Content is exploding. But it also means that code will explode. People are positing that most software will be written post-2026 and it'll be written by AI. Does that change things meaningfully in a world where all software is new and most of it is written by AI? So it'll change a lot. And yes, I think already today, you know, I think Jack Clark said, you know, 80 % plus at Anthropic is written by, you know, Cloud Code. I think the same similar thing is at OpenAI with Codex. I think we're seeing that grow in all tech companies at a pretty fast rate, and eventually it'll spread beyond.

18:27And the speed and intensity at which it writes code will mean that very quickly the highest percentage of all operating code will be AI-driven. And this enters into a whole bunch of new questions around like, okay, what does cybersecurity and safety look like? And it's one of the questions obviously surrounds Mythos and other things like it. What is, you know, kind of enterprise provisioning? What does it mean in terms of the auditability of code? What does it mean in terms of, you know, like, okay, so, you know, how much of that code is your own, you know, kind of IP, you know, in terms of doing this, there's a whole bunch of questions, which, you know, it's part of partially being on the board of Microsoft that, you know, They started addressing last year much earlier than everyone else because they have the best telemetry and knowledge about how to provide to large enterprises.

19:23And I think it's one of the places where Satya and Microsoft people are being very smart. And by having that telemetry moving early on all those things, obviously at Greylock, we see a bunch of that kind of thing in terms of investing. But what happens when there is now, call it 1 ,000x amount of software? And what does that all mean in terms of how the world operates, how industry operates, how society operates? That's literally where there's just tons of work going on. So let's move from software to music. That's another place where we're seeing just an explosion of AI-generated songs, albums, lyrics.

20:06And you are someone who have experimented with AI music. And we saw just last week, Suno, which is an AI music company, they raised$400 million, valued at$5.4 billion. And that's double the valuation from seven months ago, more than double what it was seven months ago. And so we are seeing people generating 7 million tracks a day of AI music. And Suno has been the top app in the App Store's music category in multiple countries. We also saw that Warner Music Group settled its portion of a joint lawsuit filed by all three major labels against Suno. And I feel like we saw similar things with Spotify and Napster and sort of the evolution of online music.

20:47It was first the record labels were suing them, and then they figured out what the partnership was. and how we can all work together. And so Suno is really excited about this AI future. Some people are less excited. And I feel like music, though, is getting less attention than some of the creative models, the image models, the video models. Why do you think it's getting less attention, maybe in the press, but that it's making it such a compelling frontier for everyday people who are creating this AI music on Suno every day? So I think, you know, part of it, like one of the things that's weird about music is it has its own kind of you know intellectual property regime and copyright and everything else and it's you know kind of one of the things that kind of enshrines the you know recording industry you know legal association right um of all this and it's part of the reason why like there's relatively less innovation uh innovation and invention and you know congrats to suno and all its progress you know we use suno and and some of the stuff we do you know made an alternative christmas holiday album because it was kind of as an instance of this and you know are working on other projects as you know um and i think part so i think that's one of the reasons it's been it's been um you know kind of slow it's also one of the things where you know kind of in the early days you know open ai was playing with it and then kind of went okay it doesn't seem to be on mission to how do we create kind of AGI for the benefit of humanity.

22:20It's a cool thing, but it's not absolutely essential. You have all these complications when it gets to the IP and copyright kind of material. Why engage in that battle? So they had some of the earliest music generation thing, which is basically kind of put on the shelf and then move past. and then other people have kind of developed it out and it's part of the reason why in order to do this. And I think it's really important because if you kind of look at, and it's part of the reason why we started doing this as kind of a creative project, so you kind of say, hey, we've got kind of writing as expression, you've got image generation as expression, you've got video as expression.

23:02Well, music as human expression is also really important. Now, obviously, this brings in all kinds of questions around, well, if you said, hey, I would like to have a new Beatles song, then you're like, okay, well, the Beatles, you know, kind of the IP is owned by the kind of estate and there's all kinds of things that are kind of treasured and the kind of the build of it and there's reasons why, you know, we shouldn't have AI simply go and it's here's a new Beatles song, you know, kind of as a way of doing it, that should be not, you know, kind of a

23:44created thing and as an artifact. And how do you navigate that? On the other hand, there's nothing that stops like a human band saying, hey, we're going to do Beatles homage. We're going to create a sound that's our own sound that sounds like the Beatles as kind of way we're doing it. And as long as we don't say it's a Beatles song, that's okay. And by the way, that band might have listened to all of the Beatles music, might have played the Beatles music in various ways and venues and so forth as ways of kind of going, hey, we're getting familiar and knowledgeable about this. That's all okay. So you go, okay, AI training regimes, AI listening regimes to doing this seems to be fine in terms of doing that, in terms that it just needs to, you need to have a distinction between these two things.

24:31And I think it then creates a wide range of, you know, in our podcast possibility for creativity and creation for doing all this, which I think is really important. And it's part of the reason why it's like, okay, let's go create, you know, kind of, you know, kind of music that is in, that is unique and new, like, you know, new Christmas music that has never been created before that helps, you know, kind of amplify human creativity, human expression, you know, that kind of thing I think is super important. It doesn't take away from anything from how artists, you know, make their own, you know, kind of brand and career and money for music, which I think is really important.

25:20And I, by the way, I think AI music can amplify artists in various ways because it's like, oh, it becomes a new tool by which I can create new music for the thing I'm doing and all the rest for how this operates. It should be co-opted and owned by the various musicians of all types. But then it also enables people who are not like proficient at singing and piano and so forth. And so as you know, we created this, you know, kind of Pi Day animated video that's using ReadAI to sing. Sorry, it might be fun to include a little bit of the ReadAI singing voice here just so people can get a sense of it.

26:01Every pain eventually, that's what infinity can do. Every number you could ever need is in a number about circles and new. But no one's ever created a holiday video celebrating March 14th and Pi Day and the mathematical concepts and all the rest of it. It expands all that. And I think that's what we want for humanity in addition to things for the artist and things for the industry and all the rest. And I think that's what's amazing about this. So I think part of the reason why this is important for society, important for humanity, important for humanity, and important for industry is all of these things that I've just said.

26:58All right. Well, we will definitely put the link to your Pi Day song in the show notes so everyone can see for themselves what they think about some of our AI music creations. Reid, thank you so much. Pleasure. Last but not least, today, Stripe and Frontier announced$1 billion in new carbon removal commitments backed by Stripe, Google, Shopify, Salesforce, H &M Group, and Anthropic. When Frontier started, there were barely a dozen companies in this space. Now there are hundreds. Removals are doubling every year, and over 350 other corporate buyers have followed their lead. Next week on Possible, Reid and I sit down with Nan Ransehoff, head of climate at Stripe and founder of Frontier, so the timing couldn't be better.

27:44So subscribe and stay tuned. Possible is produced by Pallet Media. It's hosted by Ari Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Tanasi Delos, Katie Sanders, Spencer Strasmoor, Imo Zhu, Amon Suri, Lexi Kevin, Danny Garrison, Trent Barbosa, and Tafadzwa Nima Rundwe. Special thanks to Surya Yalamanchili, Sayida Sepieva, Ian Alice, Greg Beato, Parth Patil, and Ben Rallis.