Anduril Hits $100B, Saronic Hits $9B 💥 Inside the Defense Tech Boom w/ Dave McClure and Aman Verjee

But oil is over 100 bucks. Google is spending a lot on capex. SpaceX is still a little bit underwater. Big news this week is I guess everybody talking about China openweight models, open source models. Big letter that everybody signed on to Nvidia etc. But Anthropic and OpenAI did not sign. Anthropic's IPO coming up in October maybe sometime in Q4 and speculations about revenue growth. Is it growing or slowing? Couple of Chinese IPOs. CXMT list in Shanghai. Moonshot listing in Hong Kong. Data bricks raising some secondary at 188 billion. Let's start with a little macro news here. President Trump is angry that the ceasefire is not holding and threatening more strikes, but maybe some heads inside the administration like, 'Hey, do we actually have enough rockets and should we be continuing this?' Probably not helping our price of oil. Not going to be great if oil is four bucks a gallon or more towards November.

Yeah, I thought we had a bit of a window there for a while where it seems like there was at least an agreement to pause the blockades in the streets of Hormuz while they were negotiating the broader issue. But remember, we were looking at those poly market charts and thinking, man, these sides are so far apart on some of these key issues that I don't know how exactly get to an agreement. I know nobody wants to fight to continue, but in terms of the substance of what they're negotiating, they were very far apart on just the ownership of nuclear weapons, and it didn't seem like there was any imminent agreement. The price of oil did go up 100 on Thursday, but it was back below on Friday, Saturday, and now we're filming on Sunday. So, every day is a little bit of a new thing.

Google had their earnings call and although the numbers were great, the stock went down because they're spending $200 billion at least on capex. Yeah, I thought they had a great quarter. I think they are maybe a little bit on the back foot in terms of the guidance and how they're setting expectations, but let's just go through the quarter that they had. I mean, if you look at their stock price, it looks like they traded off by about eight or nine percent in the four days or so since they announced their earnings. You got to think they must have really missed on something bad. They're about 21-22 times price earnings ratio right now. So, a little bit more expensive to have, but doesn't seem—no, that's right in line with S&P 500. That doesn't seem like an expensive multiple. If you look at the highlights on their earnings, here's the summary. They had topline growing at 24% year-over-year. Their Q2 revenue within search is up 17%. This is their franchise that it really funds the business. I think is the most under attack from AI, but they don't see any downside if anything is accelerating. Their cloud business is 82% through the roof, hundred million dollar revenue run rate right now. Their—if you look at the metrics on Gemini and YouTube, some of those ads are going to be AI powered just monthly users and tokens per minute. They don't disclose revenues around Gemini but all those numbers are extremely positive and if you look at just take out the Google Cloud business for a second, this is what their services revenue looks like. It's up 15% year-over-year just the total topline subscriptions search profit, you know, going up to 40. Business looks really healthy. There's no damage being done to their franchise by AI and then the cloud business has been ripping along. This is their both their top line and their bottom line for the cloud business. The last couple of quarters, you can see they've just been piling on more and more revenue, 24 billion last quarter from 20 the prior quarter before that, 18 before that, 15 before that, and they're now running at 30-40% margins here. There was nothing wrong in the print and they should have stopped talking about 20 minutes into the earnings call because once they got to the spend then I think it all kind of went—yeah, this is their free cash flow chart. It's the first quarter since they went public let me put it that way that they had negative free cash flow and if you don't know accounting just as a reminder you take your revenues you take out your costs and expenses that gives you operating cash flow and then you deduct from that capital expenditures and that gives you a free cash flow measure and then you add tech non-cash items and expenses like stock comp and stuff like that. So you put it all together and they had their first ever negative quarter.

No, I think that's a different part of the—that's investing cash flows that should be separate from what they'll call free cash flow. So this—this would if they can finance this with either you know new bond sales which they did or new equity that would be on a separate graph. So this is just take your—

Yes. Despite that. So they're having to fund this off balance sheet with new money coming in. And then I think that's part of it. It's like well they're cash flow negative. So a lot of value-oriented investors who've been moving into this name and they just joined the Dow Jones Industrial Average 30. So maybe there's some new investors who got a little bit freaked out by hey I don't understand this red. I don't understand what the red means.

And then look at the guidance. This is the expenditures on a quarterly basis. They've said we're going to—we spent 45 billion last quarter on capex. That was 35 billion the quarter before and 27 quarter before that. So a massive increase in capex in a short amount of time. They had guided to 180-190 billion in capex on the last call and now they're upping that and they're saying we're going to spend even more in 2027. So that's kind of a negative. You'd have a business that a couple years ago was spending 30 billion in capex a year, went up to 50 in 2024, up to 90 last year. Now we're talking 200, maybe on its way to 300. Like, who knows? They missed their guidance a little bit. And I don't think people know, you know, quite what to do about it. And you look at how the analysts treated it. I listened to CNBC and figured, hey, what is everyone saying about it? 10 analysts lowered their price targets. Three of them increased it. And all the people who were lowering their price targets whether it was Oppenheimer whether it was DA Davidson Piper Sandler they were all saying capital program hey we're concerned about you know margins and margin compression the higher spending outlook a couple of people referenced the headcount so if you look at the headcount numbers which they do provide on their earnings and their earnings releases they have gone from 187,000 employees to 198,000 employees doesn't sound like they're gaining efficiency here none of this is that surprising to me to be honest if you'd said before the start of the year topline great accelerating you know growth making money this team since the beginning of time has been one of the most capital efficient spenders of all time in terms of corporate efficiency huge track record over 20 years very capital disciplined they're now betting behind something that's working the only negative I'd say would be well if you told people one number and then surprised them with a new number maybe that was you know maybe that's a little bit of a messed up but I would trust this team with their capital efficiency and the fact that their revenues are growing and accelerating seems like there's something here that's really working.

And I think even if you show revenue growing, which clearly Google Cloud has, and it's got 80% growth and it's a hundred billion dollar business, I think the fact that they guided to a lower number and are now surprising with a bigger number is just freaking people out. They're not trusting you right now with the kinds of those kinds of big spent numbers. But I think Google is probably demonstrating real traction, real results. As long as they can do that, you know, I think they'll be fine in the long run.

So, the reason I think we've seen some loss in the stock price is they broke issue. So, they issued—the issue price on the IPO was $135 a share. Their first tick was 150. They went as high as 225 a share but those were all very right you know small temporary movements and then once you go below issue price there's a lot of people who sell and that's just an artifact of public markets is people go into the price target and if you go below the price target it just triggers selling there's a lot of short interest in the stock right now and then Tesla also has had a very soft couple of weeks so they're down to 1.2 trillion market cap and to the extent that Tesla goes down there's some sympathy trade with SpaceX because a lot of people think Tesla and SpaceX may be one company someday myself included.

So I think there's just you know negatives like that. I think on the earnings call we'll probably see the best case would be reaffirm the rest of the year revenue guidance you know put out some solid numbers or a deal with another hyperscaler they've got deals with Google Anthropic Reflection. And so what does that look like for AI business? Fundamentals are unchanged, but the sentiment has gone up and down. So I think if you're them, you just deliver on the results. You know, a couple more deals would help solidify that AI business, the stock price will take care of itself.

On August 6, right? That's the first 20%. If the price goes back above the IPO price, it could be another 10% lock up. Doesn't seem likely where the price is right now, but there'll be a 20% partial release. That'll be another factor to consider in the second half of August.

I think it is notable that this is I think Jensen's first time on Twitter. Maybe it's Chris's first post in a while where he's—

I give him credit for putting himself out there on this issue. And as you said, the ones who didn't sign the letter were OpenAI and Anthropic. They have a vested interest in keeping a lid on any kind of open source or openweight model. So, I expect them to be on the other side of this. And then Google, I think, is actually a big beneficiary of open source and openweight, but they're also the biggest shareholder of Anthropic. So, maybe they're trying to keep their—they don't really want to take one side or the other. I think what is happening is Anthropic is pushing hard to ban open source models because they want government protection for their business and some of it is safety related and we talked a bit about last week about whether if Chinese models have been engaged in something called distillation. You take someone else's model, you fire it up, you ask it a question, you use that to inform and train your own models. This happens all the time. We did this at eBay when we were trying to build a search engine. We took Google's search engine. We looked for, you know, what's listing on eBay, how it works, how it doesn't work, and try to use it to improve our models. If Chinese models are doing that, that seems like that's just part for the course. You know, I think we would all benefit if we had a more open-source set of competitors. It would be cheaper for a lot of their customers. It would mean that Anthropic and OpenAI would have to compete harder for what they have. So, it's beneficial for the ecosystem just as it was in the 1990s. I hope the Trump administration is thinking about this the right way and sort of solving for the big picture here.

I don't know. If you believe what I believe which is we need a strong healthy open-source ecosystem to maximize innovation. You would think there would be and there's a business in there. You would think that a lot of American companies would be in the game. So far, the ones we've been talking about are all Chinese, whether it's DeepSeek or, you know, Kimmy or Moonshot. And I suspect a big part of the reason is if you're an American, it's the same reason why Americans don't play soccer very well on the international stage. Like if you're an American athlete, you don't really want to play soccer, right? You play football or basketball or baseball. I guess if you're an American AI engineer, you're like, 'Ah, I want to go work for OpenAI or Anthropic or Google. I can get paid better, you know, maybe ElevenLabs, maybe someone like that.'

Meta was in the horse race early with Llama, but that Llama has been running like a llama against thoroughbred racehorses. They don't seem to be—so I feel like but I feel like I'm hoping that there will be an American heavyweight player. You know, openweight and open source are very important. I keep saying open source and open weight interchangeably. In reality, open source is something more specific. It is transparency across the entire supply chain, including access to the training code and the blueprints and the data curation technologies and licensing frameworks. Open weight is really just the numerical parameters. A lot of these companies that we like—Alibaba's Qwen or Moonshot, Kimmy—they have zero visibility into the exact training data or pre-training pipeline. So, it's really you just get a black box, but it is open weight. I think it would be beneficial if American companies got into the game, but so far it doesn't seem to be like there's a lot of them there. So, we'll have to see how that goes.

I think they're trying to get ahead of not just rival OpenAI, but also DeepSeek, which is announcing that they're probably going to do an IPO as well. Aman, what do you think is happening with the situation for Anthropic revenues?

I would think that $74 billion number is not that far off and maybe a month off, but I think by October, November, they'll be close to that. I would say actually one other thing really helped their IPO this week. I would take my odds up considerably as of what happened this week. This class action settlement with thousands of authors and publishers was a $1.5 billion dollar payment. They basically admitted that they had been using information—I won't say pirated books but maybe I would say without getting the permission from all these copyright holders they took information to train their models and then didn't compensate them and the class action lawsuit that came back actually had two very important holdings. One was yeah the way you got about the data you went about getting the data was illegal like you should have done a better job of securing those permissions but the other part of the ruling was you can use data in order to train your models as long as you don't lift someone's exact words and use it and pass it off as your own and that's part of fair use and that was a little bit controversial going in but the judge really sidelined the IPO if they had gone the other way and said hey this is not fair use now have to go back and get permission from all these different copyright holders that could have really sidelined the IPO. So I think Anthropic kind of dodged a bullet here and I think this clears the way for a much smoother IPO. This would have been like a top three question on the road show before this ruling like hey what's going on with your copyright issues and now I think the focus is on business results you know Chinese openweight models or openweight models in general coming to get you and is there enough demand to justify and this takes one potentially significant issue off the table. So it's probably a good week for Anthropic.