In the video clip below from a recent episode of The AI Investor Podcast from 24/7 Wall St., the team breaks down the latest news surrounding OpenAI.

OpenAI, under the leadership of Sam Altman, has been making headlines with ambitious claims and rapid developments. Altman, known for his bold statements—such as requesting massive resources like $7 trillion in investments and 36 fabrication facilities from TSMC—has both critics and admirers. While some view his projections about AI’s future impact as overreaching, OpenAI’s successes are undeniable.

With ChatGPT amassing hundreds of millions of monthly users and the company raising capital at valuations exceeding $100 billion, OpenAI is at the forefront of AI innovation.

Despite this momentum, OpenAI faces internal challenges. Reports of executive departures and previous board issues raise questions about the company’s stability. The transition from a nonprofit to a for-profit entity, scaling from 700 to 1,700 employees, reflects growing pains as the organization shifts from a research-focused institute to a revenue-generating enterprise with over $4 billion in annual revenue. These changes open opportunities for competitors like Anthropic and Meta to gain ground in the AI landscape.

As you’ll hear in the video, David and Eric also discuss OpenAI’s recent release of a new model marks a significant shift in how large language models operate. By utilizing real-time inference, the model aims to reduce hallucinations and improve reasoning, potentially increasing demand for inference capacity and impacting companies like Broadcom and AMD.

As OpenAI continues its complex journey toward artificial general intelligence (AGI), the company’s future—whether it involves going public or remaining privately funded—remains a topic of keen interest in the tech world.

Transcript

[00:00:00] David Hanson: Uh, you mentioned open AI and asking for five gigawatt massive energy. And Sam Alton has been, there’s been a lot of quotes of him going around saying. This is the future. This is everything I need. Seven. What was it? He needed 7 trillion. Uh, go

[00:00:17] Eric Bleeker: around. Yeah. Um, he went to Taiwan semiconductor. I think said I need 7 trillion worth of, I think he wanted like 36 fabrication facilities and, uh, it was an audacious enough, uh, claim.

[00:00:29] Eric Bleeker: They started just calling him podcast bro

[00:00:32] David Hanson: around their offices. So, so from, uh, from two podcast bros to him, um, you know, he’s certainly a character that. I think there’s criticism of him making these very bold claims that, you know, every, every white collar job will be eliminated in six years. I’m not saying he said that exact thing, but he has these very bold claims that some people can be critical of being a little bit too over the top and too promissory of this future of AI.

[00:00:59] David Hanson: But then, you know, OpenAI has been a successful company, both in terms of Adoption of their consumer product. It’s, I think it’s hundreds of millions of, uh, monthly users that are still using chat GPT, and I think they’re raising capital at over a hundred billion dollar valuation. So. They’ve been successful, but then I also saw this week that there’s, you know, more headlines of executive departures from this company.

[00:01:22] David Hanson: So, you know, I observed the company, but as someone you probably just spends a little bit more time in the space, what the heck is going on with this company? You would, you think it’s this rocket ship, but then why does there seem to be so much turmoil? There was a board issue last year. Now there’s executives leaving.

[00:01:38] David Hanson: What are we supposed to make of the situation at open AI? Yeah,

[00:01:41] Eric Bleeker: it’s going to make a great movie someday. That’s for sure. You know, it’s going to blow the social network away. Right. Um, I, I think your point there’s really good high lying Sam Altman and his background. The kinds of claims he’s making, you always need to listen to what he’s saying with a BS detector because he has every incentive in the world to really play up numbers, right?

[00:02:09] Eric Bleeker: Because he’s out there fundraising. I’ve seen critiques from open AI employees that he’s not. In the office day to day or involved because he is essentially constantly on the road, um, on a roadshow kind of promoting AI. Right. And that’s, that’s part of your job. As a CEO, but all the same, he definitely does take it to a very high level.

[00:02:36] Eric Bleeker: Now let’s just unpackage a little bit. There was a very important story about open AI’s, uh, actual products this week. And as you mentioned, they are over 4 billion in annual run, right? So they, they have some momentum, but they released a new model, which is, uh, had been codenamed strawberry and it’s. Uh, four, uh, Oh, one is the name of it.

[00:03:02] Eric Bleeker: And it’s kind of a fundamental change to how these models, these LLMs work, which a big problem that LLMs have is they hallucinate, they give weird responses and it makes them very hard to trust. And, and that’s just a factor of their, what you call imitation learning. And there’s an alternative way of working of, you know, creating these models called reinforcement learning.

[00:03:28] Eric Bleeker: But the problem is, um, you’re, you’re essentially reinforcing, uh, these models for having the right answer, but you know, it requires human involvement and it has a hard time scaling. So their breakthrough with a strawberry model is essentially. They’re using something called real time inference, where they are using inferencing for the model to effectively what you’d call it, think and reason.

[00:03:53] Eric Bleeker: And the more time the model spends inferencing or thinking better answer you get. So I talked about earlier, we need to be really adaptive to changes of what’s going on. I had been, and just. I want to pause here because it’s even, I need to take it to a fundamental level to say with models, there’s, there’s two areas for, you know, all this AI revenue coming in for a company like NVIDIA, there’s training, which are actually creating the models itself.

[00:04:27] Eric Bleeker: And then you have inferencing where you’re kind of running the models. Right. And I’ve been a little more bullish on the training side. Because there has been some data out that I believe Goldman Sachs said there’s such an inferencing glut coming that there’s going to be no possible way to basically use all the inferencing capacity that we’re creating.

[00:04:49] Eric Bleeker: So that was true. Maybe up until two weeks ago, when all of a sudden you might have something where the demand for insert inferencing might suddenly the total addressable mark, as you would call it might be up a hundred fold. So whatever priors you had just had a nuclear bomb dropped on them, potentially, if this is the way that AI models are going to be built going forward, right.

[00:05:16] Eric Bleeker: And, and how we want to separate this podcast, David, right? We. We don’t want to just tell the news. We want to tell people how you are actionable on the news. Right? So when I’m looking at strawberry and trying to understand it and trying to fit into how we invest, all of a sudden different companies get a lot more interesting.

[00:05:40] Eric Bleeker: Um, I think AMD will have some gains in this. They’ve been a little bit stronger in inferencing, but I think the one that. You know, is going to move up a little bit in terms of our conviction list is Broadcom because all of the big hyperscalers, the Microsofts, the Googles of the world, um, this is going to, if this becomes the standard way of operating AI models, it’s, it’s going to mean that they want more inferencing capacity and they’re probably going to want to build that with their custom accelerators and the company that designs it for them is Broadcom.

[00:06:12] Eric Bleeker: Broadcom by and large. So I think there is a very specific number one, if, if this is a fundamental better model, it’s good for AI, it’s good for every AI company. Second, there’s probably going to be some outsized gains to Broadcom. So that’s just an example of we’re watching the news. And something very material happens and we’re going to have an outcome from it.

[00:06:37] Eric Bleeker: It might change how we’re thinking about things two weeks ago. Now, second, what’s happening at open AI. I know it took me forever to get there, but what’s happening is essentially it was started as a nonprofit. It’s scaled in the past year from 700 of the most talented and smart, like research focused people to 1700 and their model.

[00:07:00] Eric Bleeker: I think the recent report is that they are changing out of a nonprofit structure entirely. There’s, there’s contrasting reports, whether or not. Altman is taking a 10 billion stake in the company. I don’t know. There’s, he says he’s not the media says he is. I think it’s, I think it’s largely growing pains based upon that, that it’s a company that started out with very research focused and now it’s becoming kind of a product focused company with 4 billion plus in revenue.

[00:07:29] Eric Bleeker: And you are seeing the growing pains there. So I think open AI is going to, they’re not giving up their leadership at all. You know, they’re still going to be amongst the biggest leaders, but it is definitely presenting an opening for other companies like Anthropic and also, uh, Meta or Facebook to kind of catch up with them.

[00:07:51] David Hanson: In terms of, it sounds like it’s kind of a complicated situation moving from a nonprofit to a for profit company. That’s scaling to the size. Do you have any sense of, or what would be your best prediction of when would we see an open AI as a public company? Is this going to be an example where it’s going to be another, like.

[00:08:08] David Hanson: Eight years where the private market is just so deep that they can continue funding, they have to deal with Microsoft. Or is this like, Hey, in two years, we might have a public open AI, or do you just have no, what would be your best guess of when we get a view? Cause it is such an important company that we don’t have a view into its numbers and the economics of everything.

[00:08:26] David Hanson: Do you have an opinion on that? That’s

[00:08:28] Eric Bleeker: a great title for this podcast. It’s a complicated situation. Um, you know, everything about open AI is just so weird. Another weird thing about them. Like you have this Microsoft investment. They’re kind of partners with Microsoft, but if they reach. Artificial general intelligence, which is kind of the end point that is causing all this investment.

[00:08:50] Eric Bleeker: I think that like ownership is effectively dissolved. So how you even model that into, you know, values is just, I don’t think we’ll see them coming onto the market in the next two years. Um, you know, part of that too is just, well, we, I think I saw a stat yesterday. There’s two IPOs last year and one this year.

[00:09:11] Eric Bleeker: So if the windows open and they’re valued at 500 billion, um, potentially, but you know, so much of AI is going for this long game of reaching this AGI. And there are so many, you know, true believers there. I, I think a lot of their funding would like to see them try and reach this exit point rather than being content with a, only, you know, a 500 billion gain, right there, they are truly looking at the, uh, you know, massive scenario.

Thank you for reading! Have some feedback for us?
Contact the 24/7 Wall St. editorial team.