解放日、关税、美国对中国开源、OpenAI、$CRWV、TikTok | BG2

A riddle for you before we move on. What do Salesforce, Netflix, Square, Amazon, Palo Alto Networks, Facebook, Snap, Proofpoint, NetSuite, and CoreWeave have in common?

Guess.

No idea.

They all broke issue. Oh, wow.

Bill, great to see you.

Good to be seen. You have to be pretty stoked coming off those wins last weekend in San Francisco. I’m still repping the Gator hat. I’d say we eked by. For those who watched, that was an incredible final few minutes.

Explain it. Take us through the final few minutes.

To be honest, I was afraid. I didn’t think there was a chance at that point because they were behind by so much heading into the end of the game, and they basically scored 4 three-pointers against 0 from the other side. They intentionally fouled, and the other team missed free throws.

I think someone said the Yahoo game predictor had it at something like 98% Texas Tech with very little time left on the clock, and they somehow eked it out. The bullish people will say, “You live through something like that, and now you have confidence to deal with anything.” But the oddsmakers have put Duke in front of Florida at this point, and that started in the opposite place.

One game at a time. It was super exciting. I was out in San Francisco at the Chase Center and got to hang out with the coach a bit, along with some of the players’ parents. It was a good trip, and now that I’m in Austin, it’s right down the street here to San Antonio. A lot of friends are coming in, and we’ll see what happens.

It’s going to be a heck of a weekend. Maybe if a certain couple of teams end up in the championship game, I’ll sneak in there on Monday night. I guess as long as we’re repping and rolling, I have 2 degrees, 1 from Florida and 1 from Texas. The Texas ladies are playing in the Final Four in Tampa, so there’s a lot of good stuff happening.

1. Liberation Day & Tariffs

Speaking of a lot of good stuff happening, I think there are some people in the world who think a lot of not-so-great stuff is happening based upon the market reaction to the president’s announcement of these tariffs. Why don’t we kick it off by talking about Liberation Day?

Yeah. We’re recording this right after Trump’s presentation. I think you and I both watched it, and then we jumped on the pod. You’ve been talking about this. It’s obviously been choreographed that this was coming, and you’ve been saying for a very long time that Trump and his team are very serious about this, rather than the argument that it’s just a means to an end and a way to get a negotiation started.

You have made the point that you believe they believe this is actually where we need to take the economy. As such, you’ve been conservative and worried about where this would go. You gave a presentation last week on how to think about this. What did you talk about there? I think it was at the J.P. Morgan Tech conference.

It was pretty clear to me and you—we were talking about it in early February. This was doctrinal. There was a philosophical belief around trade that they wanted to create a more fair and level playing field. The real debate has been how big, and there are a couple of different camps.

J.P. Morgan had this great event in Montana last week with 100 tech CEOs. They had Howard Lutnick and likely Elon Musk, David Sacks, and Doug Burgum all talking about various aspects of this. They asked me to do a little bit of a presentation on decoding the Trump economic agenda. Really, it boils down to this, Bill.

At the top, I think all the CEOs in the room are pretty excited about the golden age people have been talking about: a pro-growth administration, pro-business, pro-investment, lower taxes, less regulation, and pro-M&A. We’ve seen this M&A flywheel starting to kick up, along with this AI supercycle. But everybody’s been pretty terrified about these tariffs.

The real question going into today, Liberation Day, was whether tariffs were going to land closer to the $600 billion-to-$1 trillion tariff level that Peter Navarro and Howard Lutnick had been talking about, or maybe at a little lower end of the spectrum, which we heard a little bit more from Scott Bessent and Kevin Hassett. I think everybody was holding their breath.

We got the answer today, and I was framing that at the J.P. Morgan conference. I showed this slide you’ll get a kick out of. I ended the presentation with 2 planes coming in for a landing. They both actually land, believe it or not. But it’s like the glide path that we land here with tariffs and budget cuts matters a lot.

You’re right: we just listened to the president talk, and he came in on the larger end of this. There’s no other way to slice it. There was a headline that hit right after the market closed from The Wall Street Journal, I believe, saying that it was 10% across the board, and the markets jumped up around 2.5%.

Then, as the presentation started unfolding, the tariffs started coming in. People saw this chart that they presented on reciprocal tariffs, where they said tariffs on China were going to 54%. Can you believe that? That’s 34% on top of the 20% that already exists.

The futures—the S&P futures, the Qs, and the Nasdaq futures—started sinking. They had a 600-basis-point fall between where they initially jumped and where they ended up. The market is not liking this at all. Depending on what index you were looking at, we were already down 8% to 15% on the year. Whatever comes in tomorrow will be on top of that.

That’s the chart. That’s the initial reaction out of the market. We can break them down a little bit if you want, but that was the initial reaction.

One of the pieces that I witnessed, and I think everybody else did as well—and maybe you have more data on it—is that when they said “reciprocal tariff,” they kind of redefined what a tariff is for the reciprocal calculation. They are including other things. I don’t know what else goes in there.

Maybe you know exactly what fills this in, but basically, the Trump administration is calculating an effective tariff, if you will, for each and every other country, which may not be the explicit tariff. Correct?

Correct. They call them non-tariff trade barriers. This is everything from what they call currency manipulation to things like judicial actions that restrict free and fair trade of our products into their countries.

By the way, we know there are non-tariff trade barriers, so it’s not totally surprising. But if you and I were to do the math on these, you can make those numbers whatever you want to make them. If I had to go through this, the tariffs largely break down into, let’s call it, 3 or 4 big buckets.

There are the auto tariffs, which are largely on Mexico, Canada, and Germany. The auto tariffs were imposed at 25%. In fact, he had 20 members of the UAW in the front row at the event. He invited, I think, the president of the UAW up on stage to make some comments.

He said, “These people used to be Democrats. They’re Republicans. Now we’re the only ones who fight for them.” By the way, he said, “We won the state of Michigan because of this. This is what I campaigned on. These are the promises we made, and we’re delivering on the promises.”

It is striking to me that, just politically, this is what Democrats were running on 20 years ago. It just shows how much the political parties have changed. This is a big tariff differential as it relates to autos.

Then he had the reciprocal tariffs, which are the ones that I outlined here. Bill, remember: Trump is the negotiator in chief. This is the starting point. All these tariffs go into place. We’ll put these charts up on April 9, country by country.

We’re going to hear all these ad hoc negotiations going on, some of which I’m sure—like Vietnam—he’ll declare victory on even before we get to April 9 because they’ve already capitulated on a bunch of tariffs. He’s also declared that there are $6 trillion of new investments that people have committed to in the United States.

He mentioned Nvidia, Apple, TSMC, SoftBank, and OpenAI in his remarks. In fact, I particularly noted when he talked about SoftBank and OpenAI, he said, “Great companies.” For the people who are watching the battle between OpenAI and xAI, that was notable.

Then he said, “We’re going to have a minimum tariff of 10% on all countries.” Even if you’re not on this list, we’re going to have a minimum tariff of 10% on all countries. Of course, China is in this bucket on its own, right? That’s going to be a huge negotiation on its own.

There are a lot of things that go into that negotiation, everything from the Panama Canal to the TikTok stuff. Set that aside for a second. There’s no way that lands, I think, at 54%. Those are the headline tariffs.

When we do the math and add all of these up, we can ask, “What does this come up with?” The headline is that we were at $77 billion last year, and we end up at about $750 billion.

Remember Peter Navarro, the hawk—the person who had been saying we were going to land big tariffs. He was estimating $600 billion, so this definitely landed on the larger side. But then they came out right after that, and there was a footnote about things that were exempted from the tariffs.

Exemptions, which included pharmaceuticals and, notably for you and me, semiconductors. Wow. Right. So Taiwan has a 32% tariff, but semiconductors are exempted. And so we're going through the value of those exemptions right now. My hunch is that this is going to land right around $600 billion.

But Bill, I have to ask you: here we are, you and I, trying to make our way through this and make sense out of it. I don't think there are many CEOs we know who support this or like this. In fact, I think a lot of congressional Republicans don't like this. You've made an eloquent defense of the benefits of free and largely fair trade. When you start hearing things like this—okay, this category got an exemption, or this category got exempted—just give me your reaction. As somebody who totally understands the benefits of free trade, when you see the Republican Party doing this, how does it make you feel?

Well, yeah. Look, at a high level, I'm a believer in open markets, free trade, and comparative advantage. That's been studied for a very long time. There are very solid mathematical arguments for why, if you put up trade walls between multiple countries, you're going to hurt the efficiency of both of them in the long run. At least from a theoretical perspective, I'm a believer in that.

2. Coreweave IPO & AI Demand

I think there's another issue for the markets and for the CEOs, which has to do with both the slow pace at which they could realistically respond to this and the amount of ambiguity that's been out there. So let's say what the administration wants to encourage is for you to relocate a factory or production that you have in Thailand and put it on American shores. That's not a quick process. And if you start that process today, it might take 3 years before you'd have the type of volume that would be capable of bringing that back, and probably at a higher cost.

I mean, one of the things that I've said over and over again is I don't think our labor is globally competitive, nor do I think it wants to be. And so, even bringing it back, you're going to have a higher cost of production because we're going to have a higher cost of labor.

That being said, with this ambiguity, there were a lot of people, even going right into this announcement, who didn't know what percentage of it was real versus bluster. And I think, at this point, just reading the papers—not making my own assessment—the administration has a reputation that maybe some of this is for negotiation. Maybe some of it's not real.

And so you're left not knowing what the policy is going to be 3 months from now, 6 months from now, or 12 months from now, which makes it very hard to allocate capex in any meaningful way whatsoever.

I think that's so well stated. You and I have said markets and business abhor uncertainty. They can deal with almost anything, but it has to have predictability so they can build a forecast, so they can look at an investment and say, “Is that NPV-positive?”

I was literally texting with some big CEOs during the president's announcement, and they were asking questions: “Do you see us exempted? Are we in there?” This is just amazing to me, right, that you have this level of uncertainty. I was with those 100 CEOs in Montana last week, and I would say, almost to a one, they said things are slowing down in February and March because nobody knows what to do.

And remember, the Fed had just come out last week and taken down its forecast for GDP growth, taken up its forecast for inflation, and taken up its forecast for unemployment by the end of the year. So most major economists are increasing their probability of a recession and slowing the rate of growth. And the question today was: Was Liberation Day a clearing event? Does everybody have clarity now?

I think your point is a great one. Even though they may have gotten an exemption or they may not have an exemption, the question is, can I count on this? How long can I count on this? Can I really plan a year out based upon this, or is this going to change yet again over the course of the year?

And by the way, there are cascading effects. If you're unsure about things like this, you're not going to hire a bunch of people, for example. You're probably going to pause hiring because you don't know how much earnings you're going to have. So those kinds of things can proliferate downward and affect unemployment and eventually affect consumer spending, just by freezing nearly everything in the economy.

Let me tell you 2 other cascading effects. I heard from one of the CEOs last week that 4 contracts with them had actually been canceled. They had 3 European contracts canceled and 1 Asian contract canceled because the countries were so upset with America that they're going to do deals instead with European companies or whatever the case may be.

And I sent you the tweet where China, Korea, and Japan were actually going to collaborate in a response to the U.S. tariffs. Somebody said, “We haven't seen the Koreans, the Japanese, and the Chinese combine forces on anything since the Mongols caused them to get together.” So it is causing a lot of strange bedfellows.

And I sent you the tweet where the Europeans—President Macron, the president of France—they're all going to China, and they want to talk about closer trade negotiations or relationships with China. You warned us about this. I think it said the meeting was in Vietnam, and obviously I got that from The Economist, which had predicted that that would happen. But, yes, there's no reason that that wouldn't happen.

And there's a lot of ramifications of this. I don't think anything came out of it that's going to be good for the markets or good for stocks, but that's more your world than mine. I've always shunned macro analysis. Let me just maybe opine on that for a second. Where do I think we go from here?

After hours, with the NASDAQ down peak-to-trough post-Trump at almost 18%, that's down a lot. A lot of names in the NASDAQ are down 40% or 50%, so we're starting to get some of that fear into the market.

Somebody asked, and I said, at the very root, I do believe the president wants to do deals. He believes in fairer trade. I think we're going to land the plane here closer to $300 billion or $400 billion in tariffs, not $600 billion or $700 billion, and certainly not $1 trillion in tariffs, even though it feels today like it was bigger than that.

And one of the things I think that's going to kind of force the president's hand is that he talked at the press conference—he had a bunch of senators there and House members. The senators and House members are hearing from their constituent CEOs that they don't like these tariffs. And remember, most important to the president, he wants to get this reconciliation package passed, which he calls a big, beautiful bill.

He wants to get this thing passed, which puts in place no taxes on tips and the permanence of the tax cuts, which he passed in his first administration. He can't afford to lose a single Republican vote. And so I think that that also is going to guide him a little bit more to the center. And that's what we'll see—whether the market believes that. It certainly didn't believe it after hours today.

We're going to get a little bit more positive on the companies we like the best because we think some of this fear is now getting priced in. What will you be looking for in the next 30 to 60 days as this plays out?

Yeah, I think we're still in the fog of war, certainly, but I will be looking at whether these exemptions on things like semiconductors and pharmaceuticals hold. Are we seeing the country-by-country renegotiation on some of these things? And probably most importantly, Bill, it's really about China.

China is the second-largest economic power in the world. It scares me how big the tariffs are that we are suggesting are going to go in place on China. And I think it's imminent that he and Xi are going to have to talk and get a big trade deal done. So those are the things I'm going to be watching for.

I don't think I see any clearing event here for at least another 30 to 60 days. But remember, the best opportunities to buy something are when people are a little fearful. So you may have to take a bit of a leap of faith on this one if you want to purchase at the best prices.

Makes sense.

3. US Open Source vs China Open Source

You know, another thing that we heard a lot about this week, Bill, speaking of China, is some developments in Chinese open source and some developments on the U.S. open-source front, particularly with respect to these frontier models.

You have a lot of understanding, I think, about the history in China around open source, as well as the history in the United States around open source. Help us unpack, if you will, what you think is strategically going on in China with respect to these open-source models.

I've seen some people tweet that maybe DeepSeek was forced into open source by Xi. Do you think that's going on, or is there something else going on here? And by the way, all this culminated in the past week with OpenAI moving toward—or talking again about—open-weight models, which is, I think, a really important data point. How did we get from where we were to where we are?

So, yeah, I read that same tweet, and I think it was remarkably misplaced.

China has been supportive of open source for well over a decade now. If you look at most of the major open-source products and look at the maintainer page and who the sponsors are for these, like Linux, you’ll see many of the major Chinese companies have been there and supporting it for a while. Why? They’ve been accused of stealing tech IP for years, and so when something like open source comes along, this looks like the best thing possible. There’s no one that can accuse us of IP theft because there is no IP ownership in an open-source world.

Having dealt with those accusations for probably 40 or 50 years, I think everyone in China—the government and entrepreneurs at large—view open source as a very positive movement for their country relative to the West. So they’ve been in on it for a long time. They’re very adept at it, and they’re very big believers in it.

When we talked about the interview with the DeepSeek founder, I would say his entire emotional mindset was tied to open source. He believes in it and wants to support it. So that’s an important backdrop.

I don’t think China got there in some calculated way, or that it was some recent move. I think they embraced open source over a decade ago because it made a ton of sense for them in a world that had pointed a finger at them from an IP standpoint.

So let me just double-click on that. Basically, what you’re saying is they may have been fearful that they would be cut off from other types of software products in the United States. There might have been export controls or other things put on them, so they said, “I may as well support Linux because I may not be able to get Windows,” right?

But I think that discussion happened a long time ago. It wasn’t recent. But I think it’s important because that lays the foundation, right? If that’s a foundational belief among Chinese entrepreneurs and Chinese companies, then it’s understandable that this new generation of entrepreneurs might also see the advantages of open source.

Another thing that I think people have to remember is that within the past decade, and maybe 15 years, many US companies have learned to use open source as a defensive tool rather than just an offensive tool.

Say more.

These are the biggest companies out there. If they get in a position where they feel like they’re behind the 8-ball—so they’re not in a leadership position—they will embrace open source as an attempt to level the playing field.

A great example is Kubernetes. Amazon took this huge lead with AWS in the hosted-server business. Everyone was afraid of that. Google had a piece of technology called Kubernetes. It was orchestration that would allow you to move a workload, if that became a standard, from one large server vendor to another, right?

It basically created ease of distribution so you could run on multiple clouds. They went to the Linux Foundation, recruited IBM and a whole bunch of other people, got everyone behind it, and it gained so much momentum that Amazon had to embrace Kubernetes. So it worked, and we don’t have a monopolist in that cloud business right now, perhaps because of that deft move made by Google.

Google did it with Android against Apple—big, notable—and Meta did it with Llama here, right? They came to the table. They weren’t first to the table in the AI space, but they were disruptive with open source.

One other thing I would point out about that type of move, in addition to saying it’s defensive, is that I think it’s great for consumers. If you study economics in business school, there’s this notion of pure competition. Where do you have the most fluid competition, which leads to the lowest prices for a consumer? Certainly, open source does that versus proprietary code. It’s just hypercompetitive. That’s why it’s disruptive, and that’s why people use it in this way.

So that’s a huge backdrop to where we are today. I believe DeepSeek has been remarkably successful in the enterprise. It’s hosted by AWS, it’s hosted by Google, and it’s being used around the world. I’ve heard from Hugging Face that it’s been forked over 1,500 times on their site. So it’s prolific.

I’m beginning to hear concerns that DC may take action to limit the use of DeepSeek. And you’re saying Washington may intervene because there are people perhaps lobbying against it, or other concerns that Washington may have about open-source Chinese models being used by Americans?

I think it’s safe to say both those things are happening. There are people who are really concerned about Chinese technology getting underneath our products. Whether or not this particular code could be bad, they just might have that default. And then I think there are people who are lobbying because it would benefit their company.

Either way, if that gets traction, you have a window in the US right now where someone might move to go left of DeepSeek in terms of openness on one of their models, either in an offensive or defensive move. I think it’s a short window. It’ll be very interesting from my point of view to see whether Google does that, or Meta. I think Meta has an announcement coming up in a few weeks for its next model.

I don’t think Anthropic would do it. They’ve been so anti-open source that it would be very out of character for them. It’ll be interesting to see what happens on this front, and that leads us up to OpenAI making an announcement, which I’ll let you describe.

Well, yeah. We’ve talked here, and Sam has been dropping breadcrumbs on Twitter, right, that they wanted to launch an open-source model. They’ve been GPU-constrained and bandwidth-constrained, but he got the announcement out this week, which I was thrilled with, where he said, “We’re excited to release a powerful new open-weight language model with reasoning in the coming months, and we want to talk to developers.”

So here he’s inviting all these developers to participate and give feedback. He said, “We want to make it a very, very good model.” They’re planning to release an open-weight language model, one of their first since GPT-2. And he says they’ve been thinking about this for a long time.

The interesting thing is, somebody in the replies asked, “Are you going to make people buy licenses if they get a lot of users, like Meta is doing with Llama?” He kind of takes a jab at Meta and says no. Yes, he said no. He says no, which indicates maybe we’re going to out-open Llama.

That’s OpenAI on the one hand. On the other—by the way, just hold your thoughts. In case people don’t know, openness is a continuum. It’s not black or white, and that’s true of all the open-source technologies in the open-source model world. Some of the players, most notably Meta with Llama, have a usage restriction against the free use of the model at 700 million users, and that’s what you were referring to. At least in a tweet, Sam suggested they won’t have that in theirs. So, back to you.

I think it’s notable because, remember, at the end of this month we have LlamaCon, which is the developer event for Llama. It’s a big deal for Meta. The launch of Llama 4 has been rumored for a long time, and in fact, I think there are just a lot of people who are surprised they haven’t released it. But it seems to everybody they’re going to have to release it ahead of LlamaCon.

What I’m hearing, Bill, is that it’ll be a 400-billion-parameter model. It’ll be a mixture-of-experts model using 50, 60, or 70 experts. It’s going to have a huge context window, like a 10-million context window, and it’s going to launch this month.

I think it’s terrific that OpenAI has now fully committed. Everybody on the team—Brockman, Kevin, et cetera—were all retweeting this, fully committed, and maybe even suggesting that this is going to be a more capable model and even more open. I think it’s good that we have competition in the US for an open-source model.

When it comes to the administration and what Washington, DC, is going to do, my best sense is they do not want to see a Huawei-DeepSeek Belt and Road, either with chips or with open-source models. They do not want to see the world run on DeepSeek on Huawei 910 chips.

This gets back to the AI Diffusion Rule and how we’re going to restrict these things. I think they would love to see the world continue to run on US compute and US silicon, and they would love to see Llama and perhaps OpenAI’s open-source model around the world. They know it has a lot of distribution. So I think this was a really positive step forward on that.

You bring up an important point, which is that if something does happen to limit DeepSeek usage, whoever is going to jump in to try and lead the open-source movement in the West, if they want to be a global player, doesn’t just have to get left of everyone else in the West. They still have to compete with DeepSeek on a global basis.

I do think it’ll be interesting to watch. I’m very curious how this plays out. I’ve already asked the people at Hugging Face if maybe they will create a continuum so we can rank all these different models from an openness perspective, because there are so many different facets by which you could be open or not.

Since you’re involved at OpenAI, you can correct me if I’m wrong, but you’ve been saying for, I would say, a couple of quarters now that the real opportunity for OpenAI is on the product side versus the model side, which hints at being more of a consumer product than, say, the enterprise API business. And if they think that also—and I’m not involved, so this is conjecture on my part—being more open with your model is a really deft move because it will put more pressure on other players to try and keep up, and it will allow your model to have more pervasive usage globally.

And you talked about running out of compute. The minute you put that model out where other people can download it, they’re doing that on their servers, not on yours. I just think it’s a very clever move, for the same reason Google would have supported Kubernetes. You’re kind of wiping out the business opportunity for other models to play on the API side if you make yours open, which helps protect the competitive flank and, once again, is great for consumers.

Yeah, I think it makes sense on a bunch of fronts. On the first front, I think they want developers to develop on their platform and build applications for OpenAI, and so this brings them into the ecosystem. Number 2, I think they fundamentally have a view that they want to build products and applications that move humanity forward with AI, and this is another way to do it at scale.

Sam has said publicly—I’ve heard him say it several times—that he thinks models are commoditizing. They’re in the game. They will continue to push the frontier, so he thinks they’ll have the best models, but that general intelligence—that average-level intelligence—we already see is going to be widely distributed. The battleground, Bill, is really going to be fought around products and services.

I wouldn’t say that they view themselves as exclusively a consumer company, but clearly ChatGPT is a major thrust, a major focus of the business. It’s the market-leading consumer application, probably with 80% to 90% market share. I think network effects are kicking in and other things.

But I also think their enterprise business is, if not the biggest, one of the biggest, and also growing at the fastest rate. Remember the consumerization of the enterprise. One of the fascinating things that’s happening in the enterprise is that these are all users of ChatGPT.

When the CEO shows up in the boardroom and somebody says, “Yeah, we’re looking at bringing AI into the company, and we’re looking at ChatGPT Enterprise,” it’s an easy yes. It reminds me of when every CEO said, “Hey, get me an iPhone in the enterprise.” They were on BlackBerry, and they all switched to iPhones because they loved using them at home.

I think the natural thing for them to do is not to say enterprise is going to be winner-take-all. But I think these guys do have their eyes squarely set on building probably 2 different types of enterprise. There’s probably a product that people buy user licenses for, for doing white-collar research-type work, where I think what you said will be very relevant, like having the UI they’re used to. Then there’s the separate side, which is models that underlie the types of business processes that you’re building.

4. OpenAI Fundraise

Well, one of the things you pinged me on this week was the investment round around OpenAI, and I’m happy to share what I can share.

But, you know, what was announced?

Yeah. Well, they announced the long-rumored investment that was led by SoftBank, which many people described in the headline as a $40 billion investment round. If you read the breakdown of it, it comes in a couple of tranches: the first tranche being closer to $10 billion, and the second tranche being closer to $30 billion. It’s an extraordinary amount of money. It’s bigger than any IPO—bigger than, I think, the largest IPOs, maybe 1 or 2, that have ever been done.

I’ve often described these as private IPOs. Altimeter participated along with several others who were reported, and the valuation was like $260 billion pre-money, which would make it, if all the money were to come in, a $300 billion post-money valuation. It certainly got a lot of attention this week, and you asked me the question, I think, just around valuation: How did we think about valuation?

The first thing I would say is, market leaders never look cheap. When I invested in Google in 2005, when I invested in Meta when the IPO broke, and we looked at those latest-stage private rounds, I certainly remember the Microsoft round in Meta at $15 billion that was roundly criticized as being incredibly expensive. None of these things—you’re certainly not going to buy a market leader on the cheap.

But if you really look at this, I think they’ve said publicly they expect their revenues this year to be around $13 billion. To do $13 billion in revenue probably means you have to exit the year closer to $15 billion to $18 billion in run-rate revenue. So, as I look at this on a forward run-rate for this year, you’re paying something like 20 times revenue for the business.

Now, we also had a couple of other announcements this week. There’s the Anthropic funding round, and there’s talk that they’re doing $1 billion to $2 billion in revenue at a $60 billion valuation. So that, to me, looks like something like 50 times revenue. Again, you’ve got OpenAI at 20 times, Anthropic at 50 times, and then we had the merger of X and xAI, correct, which are rumored to have around—let’s call it—$3 billion in revenue. The combined market cap there is like $125 billion, so that looks like closer to 80 times revenue.

The market leader here, which usually trades at a premium, not at a discount, to me, again, we can argue about the sustainability and whether somebody could disrupt them, and whether 20 times is a good valuation in this environment. Aren’t they spending a lot of money on compute? Is it really high-value revenue? But apples to apples relative to their peers, it certainly appears to me like 20 versus 50 versus 80. It’s hard to say that this would be more expensive on a multiple basis than Anthropic or xAI.

Yeah. I also read that there are still contingencies on whether the full conversion from a nonprofit to a for-profit happens. So I think, if that’s true, there’s still some stuff to play out.

But one thing I would say, when I witnessed this from afar—and, once again, you’re involved and I’m not, so correct me if I’m wrong—is that, having lived through the Uber-Lyft situation, which oddly had Masayoshi Son coming to the table also, our world has just evolved into one where so many people believe in the power law, so many people believe in network effects, and that these markets are winner-take-all, that you end up with these massive capital-raising rounds.

It’s not lost on me that both with Stargate and with this one, the headline number is bigger than the piecemeal number when you start to unpack it, which, for better or worse, from my position, smells of being promotional. That headline number does get repeated in the press, so it does work in that way.

I come back to this: I suspect the company’s sending a message to Anthropic and anyone else in the game that we’re here for the long haul. They probably didn’t anticipate all the moves that Elon’s made with X and Twitter, and obviously that is another deep-pocketed player. But boy, if you’re on the Anthropic investment side, I’d be scared. I’ve lived through this before. It is a sport of kings, if you will.

And then lastly, one thing that naturally falls out of that is that unit economics get postponed. You have to believe in the power law and the network effect. In addition to that headline number, I think they’ve said publicly they expect to lose $5 billion to $7 billion this year. With an employee count of, I think, 6,000 to 7,000, that’s not going to run you more than $1 billion or $2 billion.

So there’s some number between the revenue number you talked about and subtracting $2 billion for expenses, and the rest is your operating cost of keeping this AI machine going. I ran some loose numbers, and I come up around $15 or $20 a year for a non-paying user, just to run the servers on their behalf. Eventually, you might get to advertising. Eventually, you may convert more of them to paid. These are things we’ve seen play out over time. They played out with ad models, but once again, if you’re going to try and give chase to them, you’ve got to be prepared to underwrite that cost yourself, right?

Yeah, listen, I think the analogy is a fair one, Bill, and obviously Masayoshi Son was involved in the Uber-Lyft battle, so it’s an easy one, particularly with his involvement here, to say—you’ve referred to it before as weapons of economic destruction—all of this capital.

But I would remind you there was a moment in time in 2020 when the headlines were that Uber would never be profitable. It was a failed business model. It would never make money. And here’s a business that’s going to do $6 billion in free cash flow this year. No doubt, right? And so the winner does take all. The winner does take most.

As a shareholder, I speak with the leadership of the company all the time about unit economics. Obviously, if I’m investing in the business, I feel confident in their leadership around unit economics. One of the things that I think is really important here is just what’s happening in the business, right? Sam tweeted this week that they added 1 million ChatGPT users in an hour.

In an hour, they crossed 20 million subscribers—paying subscribers—for ChatGPT. They crossed 500 million weekly active users of ChatGPT. In fact, they were going so gangbusters, they were throttling all their demand.

In fact, I don't know if you saw David Sax's tweet where he said, “America's leading AI companies are all reporting that demand is off the charts, so much so that they're being forced to impose rate limits.” And he said, “Fortunately, we have massive new infrastructure projects coming online.” Which gets me to the point of why they're raising so much money, right?

You and I are talking about taking the pod down to Abilene, Texas, to see Stargate, and to Denton, Texas, to see the CoreWeave facility that they're standing up for OpenAI. The fact of the matter is, I think that they need to bring on a massive amount of compute just to support the demand they currently have.

I can tell you, when you look at the product pipeline for OpenAI, there are 2 or 3 models they already have completed on the shelf. There's a lot of agent stuff that they want to do that's on the shelf. I think there's a lot of stuff they want to do around pricing, but they can't do these things today and open-source with their current level of GPU demand.

Sam went online and said, “If anybody has a cluster of 100,000 GPUs, send me a DM.” You may say it's promotional and hyperbole, but the round was already raised. You're not doing that.

It could be both.

That's right. I actually think, in this case, it's true. I know they were pulling a lot of things offline just to support the demand.

The irony is, where was this demand coming from, right? The demand came because we didn't mention Gemini 2.5, which happened to release in this last week. Part of the reason we didn't mention it is because, literally on the day that they launched it, OpenAI launched this upgrade to image generation, where people are making all these anime photos of themselves. That literally blew up demand for a billion anime photos a day, from the United States all the way to India, and they can't support it.

Some people may say, “Oh, well, this is an example of how dumb AI is. People are using it to make anime photos.” But I would point them to Chris Dixon's blog that he wrote some time ago, where he says, “Listen, the next big thing will first appear as a toy,” right? There are a lot of things that we do for entertainment. We know that OpenAI and ChatGPT are being used for a lot of deep research.

The fact of the matter is, at least as to this one—and I'm not going to get into the other valuations for the other models—but I'd say, at least as to this one, I was an early investor in Google. I was an early investor in Meta. I saw what those early consumer products looked like, what those demand curves looked like, and what that cohort retention looked like.

I would just say to you that what I see out of ChatGPT reminds me a lot of those winner-take-all consumer applications. They're not infallible. It's not that they can't be assaulted. Grok has been a great model launched by Elon, but I think they really do have a big moat, and I think the network effects are kicking in.

I think that not only are they an order of magnitude bigger, but they're also growing a lot faster. I think that the consumer business here will ultimately be valuable, but to your point, the unit economics can be terrible along the way. It's up to the company to launch things like advertising, paid offerings, different pricing tiers, et cetera, that bring all those things together.

Let me ask you a question. Since you went down that avenue, I think they've announced—is it 20 million paid users and 500 million total users? So you have about a 4% conversion rate. How do you think about paid versus advertising and that conversion rate? How do you think about the business model with those facts on the table?

Yeah. Honestly, I think that right now, we're probably throttling ChatGPT. When you bring on more compute, all those numbers would be higher, right? If you just have more compute.

Secondly, I think most ChatGPT users are using a model that's like a year old, because they haven't been able to upgrade the models. I don't think they can support the things they want to do in the upgraded models from a GPU perspective. My suspicion is that when they're able to do that, they're going to have a lot more flexibility around things like pricing tiers.

Sam has said he doesn't particularly like advertising, but at some point they will obviously have something that they think is beneficial to consumers that will be around that. If you look at Operator, if I say to Operator, “Book me the Four Seasons Hotel in New York next Tuesday,” and it does that for me—which I think they're getting a lot closer to—you and I have had this back-and-forth on that, but let's stipulate that even you believe at some point they're going to get there.

If we're driving that kind of value for users, either the user will pay or the end merchant will pay. I think there are all sorts of business models that will evolve around that. My hunch is you're going to see a mixture of advertising, a lot of different pricing tiers, and models.

I don't think we're going to have this long menu of model choices that forces the consumer to understand the difference between o4-mini, o3, o1, and all these different models. I think you're just going to have a smart model, ChatGPT-5 or ChatGPT-6, which may be coming sooner rather than later, that's just going to make those choices for you.

There are a lot of ways. By the way, we've talked about this in the past, but I've often felt that one of the reasons Google is so susceptible to disruption is how they've maximized the revenue per visitor. I personally don't think there's any way, when the world you're talking about—that agent world—evolves, that a partner in a hotel is going to pay a fee anywhere close to the fee that's paid to Google by someone who's marketing a service.

I always say to use LTV math versus transactional math. I just don't think there's any way you can get there. I think it's a huge disruptive advantage for OpenAI.

Well, let's click on that for a second. You know, we have our friend Glenn Fogel, who runs Booking.com. He's an incredible CEO, and they've built an incredible business. Historically, they've been one of the largest advertisers on Google. I think it's been reported that travel—hotels specifically—is one of Google's largest advertising categories in the world, and Booking is one of its largest global advertisers.

So you sell a $100 hotel room and Booking.com takes $20, right? Let's call it $18 to $20. Then you pay a portion of that to Google, maybe half of it, maybe more than half of it.

Well, actually, to be fair, in many circumstances they'll be using what I call LTV math, and they'll pay more than 100%.

Oh, they'll pay $50 instead of $20, and then they'll say, “Well, if the customer comes back twice in a year, we get to break even in the first year, and we're going to hold them forever.” This is why that won't work in the agent world: no one's going to think that way. If you're a white-label service that's underneath the agent model, the most you can share is $10 of the $20, right, or whatever.

So there's no doubt there's competition coming to that. Google's traded down from $200 to $150 and change, and they see that disruption coming their way. The irony here is we still have this antitrust investigation with Google. I always get a laugh out of the fact that, finally, in 2025, the first time we've actually had competition for Google—it's very clear that competition is coming to all those categories—we get around to talking about breaking up their search monopoly.

I mean, it's ridiculous. I don't think that's our biggest problem. But I do think we're going to see business-model evolution around these different categories.

So let's just say, Bill, that it settles out at $10, right? The hotels are clearly willing to give $20. Let's say it settles out at $10. Well, that's all upside for OpenAI, right? But it's replacing $50 for Google. That's my only point.

Yeah, no, very good point. By the way, another thing played out in this space, a little out of order from the competition, but there's this acronym people are starting to use: MCP. It's a way for you to represent your service—if you were Booking.com—to a model, so that it's not simply scraping your website, which is certainly not the ideal way to do this thing.

That standard, I believe, got started at Anthropic, but OpenAI has agreed to support it. Another factor that plays out is that whoever's most aggressive with the open standards might be able to take a lead in defining these things, which could be advantageous to them.

I’ve got to tell you, I meant to say this earlier when we were talking about the open-source stuff, but I’ve got to believe the anxiety is high at Google. I’ve got to believe the anxiety is high at Meta. We've seen some high-level executive shifts and departures, also at Apple, in terms of who's in charge of these things.

I see those moves and I think that must represent anxiety. There’s a lot at stake. It'll be really interesting to see how open people are, how willing they are to be open with their models, and how aggressive they are, because I think it's a really critical window.

I'm talking about how the next 3 to 6 months could dictate who's standing on top of the hill 5 or 10 years from now.

The tectonic plates, as I've said, as we've said for now 2 years, this is the first time they've shifted in this magnitude in 20 years, right? This is a 20-year event. For 20 years, the search paradigm ruled everything in consumer internet, and Google stood at the top of that mountain.

It took something—an AI-level shift. It took a ChatGPT moment and them getting to the scale of maybe 1 billion monthlies and 500 million weeklies to even lead to this conversation.

But things happen very slowly, Bill, as we know, and then they happen very fast. I think that's your point.

Yeah. Before we run out of time, we had an IPO, which we haven't had very many of. Let's talk IPOs—both CoreWeave, but after that, let's just talk about IPOs in general.

It's great. We're shareholders in CoreWeave. We've been shareholders since a couple of rounds ago, and we were one of the largest buyers in the IPO. We're happy we did.

I have to say, on Friday, I was pretty damn nervous, Bill. It broke the offering price and went down to about $37 a share. I think today they had some announcements of this deal with Google, where they're going to provide NVIDIA Grace Blackwells through CoreWeave. So Google is going to be buying a bunch of NVIDIA chips through CoreWeave.

One of the big criticisms of this company was that they were too dependent upon Microsoft, but now they've diversified. They have Microsoft, Meta, Google, OpenAI, NVIDIA, Cohere, and Mistral. I think they've really emerged as the leading AI cloud, and the stock in the last couple of days, despite the fact that they took it public last Friday, has done well.

I mean, talk about taking a company public into a Category 5 hurricane. We had Liberation Day staring us in the face, and they had to fly into that. As you pointed out, it wasn't the least controversial of IPOs, but I have to give credit to Mike, likely Intrator, and his team.

I'm here in Silicon Valley. They started this company 5 years ago, and it's worth over $30 billion today. It has played a really important role in standing up OpenAI and a lot of the leading AI labs, and I just think that's a good thing for all of us. But it's also fair to ask the questions that you've asked around the durability of CoreWeave and the revenue.

Yeah, and look, you're absolutely right. It's so funny: we sit around and complain about the IPO market not being open, and for the entirety of 2024, the markets were up 30%. The sunshine was out, and no one was going.

So here, someone finally gets the guts to go, and the markets have, of course, turned the other way. That's why I like anyone that tells you, when you're ready to go public, that you have to wait for the markets to be in a particular place. I would tell them to shut the fuck up and take your company out. You can't control that thing. That's an external factor.

I also believe a lot of people, because of my stance on direct listings, said, "Was this a good IPO, a bad IPO?" As you said, there's stuff moving all over the place. You have a leader in their field, unquestionably, with huge revenue growth. Their business model isn't fully unpacked because the capex is invested so far ahead of the product, so you can't look at the income statement and say, "Oh, the unit economics are perfect." There are questions about what the appropriate depreciation schedule is and all these things.

But I am glad that they got out, and I'm glad that it's done fine. They've had basically 2 customer announcements since they went out, which shows how fast this AI world moves. I think that's why the stock went from $40 to $36 and then way back up above that now.

Let's hope that it brings more IPOs to the table. This company needed capital because it's a capital-intensive business, and those are the ones that tend to come to the markets eventually, no matter what.

We have this offsetting reality: the Stripes of the world, Databricks, and others are choosing to delay being public and have massive access to private capital. Maybe that's a discussion for another day, but there are a few others—Klarna's in the pipeline. We've talked about Cerebras being in the pipeline, and I'm always hopeful. I'm always wanting there to be more companies that are willing to move into the public markets.

The offset of what we talked about at the beginning is going to be there, and so we'll see how those things interact with a choppy market.

Yeah. Somebody else I should mention is Morgan Stanley. They took a lot of heat last Friday on this deal. Why are you bringing it now? The stock went below $40. CNBC was roundly critical of the company and of Morgan Stanley, and the stock's at $60 or whatever it ended today at.

Again, we feel good as shareholders. I feel good for the people involved in the company. Obviously, there are still questions that remain about the business. You mentioned depreciation.

The one thing I'd tell you about the depreciation argument as it relates to this company is that a lot of people pushed on a statement by Jensen at GTC that Hoppers may not have value because Grace Blackwell is so much better. He was a little more aggressive than that.

Say it.

He called himself the chief revenue destroyer and basically made statements that, if you interpreted them accurately, would imply that maybe you should have a 2-year depreciation, not a 6-year. He made it sound that way. It's very abrupt, and he took a lot of heat. He probably wishes he hadn't said it, but he said it.

My view on this is, listen, we've got to square this circle. We have the Sacks tweet. We know that inference demand is off the charts. Everybody is demonstrating their need for more GPUs to run inference. Everything in the world is becoming inference. We've talked about that at length.

My view is this: when you talk about 2 years for GPUs, cutting-edge GPUs are going to be used for cutting-edge training for the frontier models in that first 2-year period. But all these things are going to continue to get used for inference.

The right way to think about CoreWeave—and I think the consensus margins for this business are about 25% EBIT over the course of the next couple of years—is how they get there. Think about their unit economics, Bill. They have their capex and their opex. They've got to get a data center, pay for all their operating expenses, and then buy the servers.

The way this works, I think, is they sell a 4-year deal to Microsoft, or a 4- or 5-year deal to OpenAI, or a 4-year deal to Google, or whatever. They expect to pay back all the capex, opex, and GPUs in 3 years. The 4th year, which is a 4-year guaranteed contract, is your profit margin, right? Anything you earn past the 4 years is all gravy on top.

The consensus earnings are not giving them any credit for anything after the end of those contract periods. What I'll tell you is that we've done a lot of research on this, and there's still a lot of A100s in use in the world today. In fact, Jensen has talked at length about that. That's a 2020 product, so we're in the 5th year, and A100s are still out there being used by almost everyone that bought A100s.

If you look at it, I think Jensen at GTC said last year that OpenAI had just retired the V100s. That was a 2017 GPU, so that's like a 7-year life cycle that they were using those for. I think we have a lot of comfort that, at a minimum, people are going to be using these things for 4 years—a couple of years for training and a couple of years for inference.

I've yet to hear of anybody throwing away any GPU because it doesn't have value. Remember the way CUDA works: the software that runs these GPUs constantly gets upgraded.

It's like my Tesla, right? I had an old Tesla Model S that was about 7 years old, but it felt like a new car because my software got updated all the time. It wasn't as good as the new model I bought in December with full FSD and everything else, but it didn't feel like a really old car because the software was constantly updating.

Frankly, it still got me to the places I needed to get to. I kind of think of it the same way for these GPUs. The GPUs are getting better every year, even though the hardware remains the same.

I'm not nearly as worried about that depreciation schedule. It seems to be a big hit on the company, and lots of people are talking about it, but they're out the door. Kudos to them on this big new deal today.

But look, the pushback on that is obviously that it's not a 0 or 1. You make it sound like it's binary: you either throw it away, or it's super valuable. What inevitably happens is the earning power of that product drops over time.

There is, I think, a reasonable question: should it be more of an accelerated depreciation schedule? The idea with depreciation is to reflect the useful life, so you'd want it to mirror the earnability of the asset over time. 6 years straight probably isn't the best fit for that.

Their peers, the incumbents in this world, were at 4 years ago and pushed it to 6, which wasn't the answer. The answer may lie somewhere in between. But we'll see. We'll see what happens over time.

And, like I said, I don't think they need it to be more than 4 in order to achieve the margins that they have. But they're also, to your point, a highly levered business. They've got to delever the business, so there are a lot of things in play here with CoreWeave.

That's why, again, if you look at the multiples it's trading at—I don't know what they are today—but the multiples that it came public at were not overly taxing from our perspective. But there's a lot of headwind for all these AI companies. I mean, you have Nvidia trading at 19 times fully taxed earnings, and so there's a lot of skepticism in the world, notwithstanding all the stuff we hear about demand—a lot of skepticism in the world about AI demand.

A riddle for you before we move on.

Yes.

What do Salesforce, Netflix, Square, Amazon, Palo Alto Networks, Facebook, Snap, Proofpoint, NetSuite, and CoreWeave have in common?

No idea.

They all broke issue.

Oh, wow.

And so, wow, when the talking heads on CNBC and others are critical of a company because it trades below its IPO price, it's just such a wrong way to look at things. I think one of the reasons those high-quality companies get priced to perfection is the founders are stronger-minded and have more leverage and negotiate more on this agreed-upon price, which would also go away with a direct listing. But boy, what a silly way to think about quality. Whether or not you give away more in the pop, that's what a lot of people think.

I don't think so. Well, I kind of thought that maybe you thought this was a perfect IPO because it ended Day 1 at precisely $40, which was the engineered offer. Let's be realistic. There are also some peculiar terms in this company that you may have played a part in. The Series C has a put right at like $38.75. No doubt in my mind people wanted to make sure it priced above that, which may have played a factor here. Who knows?

5. Rumors on TikTok deal

But let's move on. Let's talk about one more thing before we go. So much news in 1 week. The TikTok thing—there's new information as we speak. Tell me what you know.

Well, listen, there's a lot of rumors swirling. Not surprisingly, this deal is set to expire or needs to be extended by April 5 under the terms of the first congressional extension that was made by Trump. They've made it very clear that there are a lot of buyers for the TikTok asset, Bill, and that the president wants to put together a deal. Of course, we have all these tariffs going on with China, so I'm sure this will end up as part of a big trade negotiation as it pertains to China.

As you know, just for everybody, we're shareholders. I've been a shareholder in this company since 2015, one of the earliest venture-capital rounds in ByteDance, the parent company that owns TikTok. For the last 2 years, I've agreed largely with Elon Musk and David Sacks and others that we should engage with China. We shouldn't just shut down TikTok. We should make TikTok abide by the rules and regulations that we have in this country. And that's what this whole legislative unwind was about: the sale or spinout of TikTok U.S.

Here's what I'm hearing. I'm hearing that there will be a new company stood up. I'm not privy to any information; I'm not party to these negotiations, but I'm hearing—let's call it TikTok U.S.—that TikTok U.S. will be partly owned by ByteDance, but I think they have to keep that ownership threshold under 20%. So let's call it 19.5% owned by ByteDance.

It will be owned partially by the existing shareholders. Remember, the shareholders in ByteDance—60% of those are U.S. investors like Altimeter—so we'll get our shares in ByteDance or in TikTok U.S. And then 50% of it, or thereabouts, will be new investors. Think folks like—some of the rumors I've seen—Amazon, likely Andreessen Horowitz, and Oracle, et cetera. These are investors who are not currently in the cap table of ByteDance. Altimeter and Coatue are currently in the cap table of ByteDance, so we're not going to be part of the new-investor syndicate, or at least that's my understanding.

So imagine they stand that up. And then the question is: Where would that money go, Brad?

The money would go into this newco, right? The newco would be capitalized with this new money. It'd be new, fresh capital for the newco; it wouldn't go to ByteDance. That's my understanding: It would go into the newco. The newco would get a license to the algorithm, and it would be up to the newco to audit the algorithm and the data, because remember, that's the whole point here, Bill: We want to have some control over the algorithm and the data.

So it makes sense that Oracle would be involved in that, because remember, TikTok runs on the Oracle Cloud in Texas.

I think a logical question is, okay, what's the big “so what” here? I'm hearing that the valuation for TikTok U.S. could be pretty low, which I would expect, right? Because remember, Trump has said maybe we'll put this in the U.S. sovereign wealth fund. So he's negotiating the deal. I expect that he wants to get a pretty damn good deal.

You didn't mention what percentage was for that, but is that part of the cap table too?

No, no idea.

Yeah. One particular question: If you go back 6 months, maybe 3 or 6 months, there was a lot of discussion that—I would suggest—the parent company, ByteDance, had no interest in this deal. They'd rather shut it down than do this. Have they changed their mind for some reason? Is there a new perspective from their side?

Well, remember, if we go back 6 months, there was a camp that said, “Shut it down,” and there was a camp saying, “We'll just take it,” right? And I think the company's perspective—Zhang Yiming, the founder of the company—basically said there's no way to separate the algorithm between TikTok U.S. and TikTok in the rest of the world, because creators in the U.S. create content that goes to the rest of the world and vice versa.

And so, if you took away all the U.S., it does so much damage. He said you would be better off shutting down TikTok U.S. and just inviting the U.S. creators onto the French platform, or the United Kingdom version, or the Australian version of this, via a VPN or something. So I think the big change here, Bill, is this idea that U.S. TikTok and global TikTok will continue to use the same algorithm, and it's just a license to U.S. TikTok. That would be my guess as to what was part of that bridge or breakthrough.

I think a key thing here is: How does Altimeter or Sequoia or other U.S. investors—remember, 60% of the investors in ByteDance are U.S. investors—and the investors in places like Altimeter, they're pension funds, they're teachers, they're firefighters? If you think about the fair value for ByteDance, I think most people, although it only trades at—let's call it—$300 billion, think the fair value of this is closer to $1 trillion, or certainly to $800 billion.

So if you take 60% of $1 trillion, that's $600 billion in locked-up venture-capital value for all of the endowments and pension funds, et cetera, for U.S. investors. That's more than almost every other unrealized venture gain put together, Bill, right? And so, if you're able to take this company public, that turns into DPI—hundreds of billions of dollars of DPI—that goes out to the investors in these venture funds.

This company being ByteDance, we had to get the TikTok deal done as a condition required to get ByteDance public, or ByteDance out the door. Remember, about 90% of ByteDance's business is not TikTok U.S. Ninety percent of the value of the company is things like Douyin, which is the Chinese version of TikTok; Doubao, which is the Chinese version of ChatGPT; and TikTok around the world.

So there's a huge and profitable business inside of China and in the rest of the world, and we're just debating this piece in the United States. As a shareholder, I will tell you that whatever dilution is caused by this, it's nominal relative to the total value.

What I really want to see get done is just certainty, right? Certainty for the company. I think it's good for the U.S. that TikTok will remain. My kids love it. I'm glad we're going to make it abide by the rules and regulations. I think it's a win for Team Trump. I think it's a win for ByteDance.

But remember, we just hit them today, Bill, with 54% tariffs. So there may be a conversation that has to occur between Xi and Trump. I thought this deal would get approved by China. Now I'm not so sure.

And the Chinese government could probably block the deal.

Exactly. So, just because we announce a deal—if we do hear a deal announced over the course of the next few days or over the next week—doesn't mean that it's a done deal. But I'll leave on an optimistic note.

Okay, let's do that. I think that the president wants to do a deal with Xi. I don't believe we're going to have 54% tariffs against China. It's too important to the rest of the world that we can cooperate with China on things like ending the war in Ukraine and things in the Middle East.

Yes, there is a great competitive struggle between the 2 countries, but I think that ultimately the president will cut a deal. He said that he likes Xi, invited him to the inauguration, and as we know, he's a dealmaker. Now we've got everything from the Panama Canal to negotiate over, to TikTok, to all the other trade deals between the 2 countries.

So, I suspect that when we get back to what really came out of Liberation Day and what really matters, I think the most important thing that matters is U.S.-China bilateral trade relations. I think that's going to really dictate the direction of global growth and the direction of U.S. and Chinese economic growth over the course of the next few years. Important to watch.

All right, man. Take care. Great seeing you. Have fun at the games. Take care.