Could OpenAI Go Bankrupt While AI Takes Over the World?

Video Transcript

So, there was reporting that Open AI, quote unquote, has a 20 billion revenue gap.

Um, and it mainly arose from an attempt to produce a direct comparison with Anthropic, which is booking revenue from AWS and Google Cloud sales according to Reuters.

Um, because Anthropic's quarterly revenue topped Open AI's for the first time in Q2, Open AI is now saying it will reach 70 billion by year end and is raising 30 billion ahead of an IPO in 2027.

Axios is saying in public markets investors will prefer actual revenue rather than financial constructs that are far more appropriate for early stage startups.

Obviously, the market didn't like this, Oracle down, Nvidia down, Coreweave down, chipmakers overall down.

Um, AI pivoting miners were sliding and

the question is, is it accounting or is it demand? But I want to use this as a vector to look at a bigger problem, Chris, that you're raising here, which is

where are the economic gains from AI going to show up period full stop? And I think this is similar,

if you've studied history of financial markets, to some of the problems that we had with attribution in the tech bubble.

Right? Because you've really got to take three different questions apart and people like to compact them into a single ball and you should not.

And those questions are, is AI net valuable to the economy?

If so, to whom is it valuable, and then on what timeline?

And I don't think your answer has to be open AI as the winner to believe that it's net valuable and on some timeline that it will win because

as you said, it's driving significant demand for chips.

And if you want an alternative model for AI, Meta just continues to ship, ship, ship, ship, ship with muse.

And by the way, to anybody not looking at the news,

Gemini Zargon 4 allegedly is doing very well on benchmarks. I haven't seen it publicly yet.

But Muse one just hit today and is basically touching the capabilities or even exceeding them of Open AI's latest models and like Claude.

And so if that's true, I have a lot of questions about where the economic winners are. Is frontier modeling essentially a game where you're just constantly burning money and never have a moat, and so it's actually other stuff that will accrue the value.

So, Chris, I want to throw the ball back to you.

You've thought about this a lot as a venture investor. You're now at Franklin Templeton. If we're seeing a world in which chip's demand is insatiable,

the frontier model companies continue to incinerate money and while revenue is growing, the money incineration is not slowing down.

And you've got competitors in the water doing call it orthogonal things like the Chinese open weight labs or meta, how do you make sense of this whole picture?

You wait for regulation.

And what regulation does is it um forces you into a probably like a duopoly or a duopoly plus. That's the natural state of of mature markets.

Um, if you get that regulation, which you're which you're hoping for, maybe it can take care of some of your liability issues, maybe not. But what what it will do is it will create a material moat which will allow you to

when you have regulation, it forces um everyone to have the same fixed costs.

And what it does is it squeezes margin um and forces scale.

And usually not that many folks can compete. So that's why you see them so focused on this because that is the end game is where you have that regulatory mode um where there's there's a couple folks that are scaled. Think about the different industries, you know, Pepsi, Coke, you know, you you just see it happen over and again particularly with regulated industries.

Now, maybe it's not two, maybe it's like three or four, whatever, but like that that keeps out the competition. That's probably the end game that they're seeking.

Um, and and look, there's a lot of folks that are very focused on on regulation.

Um, I go back to what we talked about previously with hacking. There are laws. New technology doesn't change the law. Like you still got to follow the law. But for some reason, every time we throw a new technology out, we're like, oh no, the law doesn't apply. It doesn't matter. You can go hack and like rip people off and

So I I think,

you know, maybe those laws adapt over time, maybe the regulations adapt over time, but until that point, like we have laws. So, yeah, maybe maybe that's that's the end game here is we get that regulation, we get that um, and and then there's just the the compression of fees and scaling out.

For me, I think part of what I'm looking at here is it may be the case that the winners in this space are the people provisioning the hardware

and the compute capacity. Like this is an inversion of some of the previous things we've seen, but it may be that the physical atoms become the driver of this, right?

One of the things, you know, with plenty of investors yelling at him about it that Zuckerberg has been focused on is just compute, compute, compute, compute, and he's been focused on that for years.

And now they have a moat. And that moat is not regulation, it's physical capacity.

So 100%.

Yeah, I mean there's compute's not going away anytime soon, my man. Um and anything maybe they you see it getting changed when we have the quantum systems online and there's a lot of investment going in there as well.

So

Yeah, it's creating this framework where you need to have bet not just on something on the frontier, but the right thing on the frontier because I see a totally plausible future where five years forward,

Open AI is bankrupt, Anthropic is bankrupt and AI is taking over the world.

Picks and shovels, my man. Picks and shovels.

Exactly. It it is a totally possible outcome that Nvidia remains a winner.

you know, meta is a winner, Google is a winner, right, data services becomes the thing that wins and the other part that's really interesting to me about this, um, is we're starting to see more model tuning and like specialized models.

So to pull on two threads in the Muse drop for Muse Wonder today, they had a specifically tuned like legal model, right? That they're starting to use that is like increasing specialization within these frontier models.

And Brian Armstrong was tweeting about tuning an instance of Qwen, which is Alibaba's open weight model where they were able to host that themselves, train it on coinbase data and greatly improve fraud detection.

So it may also be that it's not so much

call it moving the entire frontier because that takes more and more and more parameters and ever more complex models, but instead starting to specialize the models for specific tasks and making them more agile.

And this kind of goes back to what we saw with the development of the internet in general, right? Like things start to fragment, things start to specialize. It's not the all singing all dancing everything that wins.

I see in that last set of comments you're able to channel Ron and uh say something about Open AI like and you know, at a deference to him, but uh look, I I I'm not as down on Open AI as as you guys are apparently.

Well, my point is, look, do I see worlds in which they succeed? Yes. I think what's understated is I also see plenty of worlds in which they fail.

Go back and look at the 90s at all the big tech names and how many made it to today.

Right?

Yeah, but not a comment specific to Open AI. It's just it's hard to predict the distribution of future outcomes.

Yeah, but Value does accrue to applications and those applications that win, win. And so, it's still early. We'll see what happens.

That true, but a lot of that's distribution.

All right, so on that note.

Austin that they have distribution.

Do they?

How how does Open AI not have distribution? How many users do they have?

I mean, they have an app and they have tens of millions of users. I agree with that level, but that is a disruptible level of users. They are not like Apple which has billions of users locked in.

That's fair. Yeah, and yeah, 100%. and they were quick to start disrupting Google when they came out. I remember all of a sudden I started using chat GPT instead of Google and I was like, wait a second. Um, you're right. this stuff can move quickly in either direction. Um and what are we seeing from a value perspective distribution.

You you distribution rules. Like as time marches forward, I think more value is being accrued to distribution well more and more every single day. Like if you got distribution, it's valuable.

If you like this segment, please like, subscribe, and tune in every Monday at 4:30 p.m. Eastern Time. I'm Austin Campbell, the host of Bips and Bips, along with my friends Rom Alia and Chris Perkins and our slate of exceptional guests.

Every week we're going to discuss macro, crypto, and the collision of worlds, covering topics that move markets and shape the financial landscape.