Nvidia为何想把芯片卖给中国、回应对Intel的质疑，以及《KPop Demon Hunters》征服全球

Hello, and welcome back to another episode of Sharp Tech. I’m Andrew Sharp, and on the other line, live from the United States of America, Ben Thompson. Ben, how are you doing?

I’m doing fine. You’re framing this as though this is some novel thing: recording both in America—

You have—

—despite the fact that we’ve been doing this all summer.

You’ve technically been here all summer long, but now you’re officially a resident, you know? And I’ve been waiting 10 days to—

No, for the record—

—celebrate live on the podcast.

—and for the IRS commissioners who are listening, I was here for a few weeks in June—

Yes.

—and then I was traveling, and I was back in Taiwan finishing my house. I did not move here until August 1st, so let’s be super clear about that.

Well, you vacationed here in summers past.

Yes, vacations.

We did. I had to return to 9:00 p.m. recording sessions throughout July, which wasn’t great. But now you’re back. We’re recording in the middle of the day. The globalist era is over. It’s a very exciting new era. How are you feeling in the midst of all this?

Frazzled. I actually have this whole fancy setup that we’re going to be using for Sharp Tech going forward. It looks great on video. I couldn’t get it to work, so we’re looking at a blank wall in the background, and it’s—

Yeah.

Very unfortunate.

A little behind the music.

But I have extra cameras now. It’s great because I have—

Mm-hmm.

—a setup in the U.S. and one in Taiwan. Now I just have too much equipment that I don’t know what to do with. But we’re working on it.

High-quality cameras all over the house. Eventually, I’m going to be projected into your basement on a big screen. It’s going to be great.

1. Nvidia Earnings Face Supply Reality

For now, we’re going to kick things off with Nvidia. Always a lot of anxiety surrounding Nvidia earnings. Those came earlier this week, and I’m going to read from The Wall Street Journal. They write:

“Nvidia’s sales set a fresh record on Wednesday as the world’s most valuable public company continued to capitalize on strong demand for AI computing, but the company’s lackluster outlook stoked jitters about future demand, sending its share price lower. Sales in the July quarter hit $46.7 billion, roughly in line with analyst estimates. The company predicted revenue of $54 billion for the third quarter, slightly higher than consensus estimates by Wall Street analysts. After several blockbuster quarters, the revenue projection was seen as underwhelming and stoked worries that growth in demand for AI chips might be hitting a plateau.”

So, Ben, there were a couple of different interesting threads from these earnings, which you wrote about on Thursday. We can take them one by one. First of all, how should we be thinking about Nvidia’s results generally, and specifically those worries alluded to by the Journal that demand for AI chips might be hitting a plateau? What do you think?

Well, number one, who knows? It’s always speculation about market sentiment. The issue was that their revenue projections came in over consensus but were short of the whisper number. It’s just a bunch of nonsense, to be honest. It doesn’t really matter.

I think this is a short-term concern for people who are day trading Nvidia stock, which is probably not the best idea in the world. I’m going to advise against that.

Okay.

The reason I say “who knows” is that until we have good evidence that there is some sort of supply-demand balance, it’s impossible to know what’s actually going on.

Mm.

The point I keep coming back to with all these Nvidia earnings is that they’re fairly pointless, in the way that Apple’s earnings were fairly pointless for the first 6 or 7 years of the iPhone. And the reason is that you could actually—

Yeah.

People at the time were remarking, “It’s incredible how Apple hits their numbers every—”

Yeah.

Or they were always X amount over. It was like, “This is the best finance department in the world. They really know what they’re doing.”

Actually, what it turned out was that Apple was slowly increasing distribution of the iPhone around the world, which was a function of signing up new carriers.

Mm-hmm.

They would sign up a new carrier, and as part of signing up that new carrier, there’d be a purchase guarantee: “You have to sell X number of iPhones.” And guess what? That gave you a very predictable view into how many iPhones you’re going to sell over the next quarter and the next year.

Right.

And so they basically nailed every single quarter all the way through the iPhone 6. The iPhone 6 was notable not just because it was the first large-screen iPhone, but that was the phone that launched on China Mobile.

I think it was the one on NTT Docomo. NTT Docomo might have been the iPhone 5 or 5s; I can’t remember. But those were the two big holdouts as far as being broadly available all over the world.

Mm.

And the first year after the iPhone 6, they wildly missed their projections.

Suddenly they’re not so—

Like, the smartphone with—

—great anymore. Ben Thompson

Right.

What the hell happened to those accountants?

It turned out they actually had no idea how to break their market. It was actually, I thought, one of the more embarrassing performances by a CEO that I’ve ever covered—

Mm-hmm.

—Tim Cook getting on the call. I think it was after the iPhone 6s, or it might have been the iPhone 7. I think it was the 6s, but it was the first quarter, and they wildly missed their numbers, and he’s just talking—he clearly has no idea what’s going on.

Yeah.

He’s just talking nonsense: “Oh, yeah, this is why. It’s so blah, blah, blah.” And then they had a terrible whole year where they missed projections, and it was a big deal, blah, blah, blah. Eventually, they figured it out. They’ve gotten better at their projections since then.

Mm-hmm.

I actually think they had no idea, because the problem was that their sales were gated by demand—

Right.

—in that particular case. As they increased demand, their sales became gated. Nvidia is kind of the opposite. It’s gated by supply. All the hyperscalers are out there, number one, massively increasing their CapEx—

Mm-hmm.

—and number two, saying demand exceeds supply, and we can’t get enough GPUs.

We will buy everything that you’re willing to sell us. I mean, you wrote—

Right.

—the sales are governed by the number of GPUs they can make, and—

Well, it’s also governed to an extent by Nvidia’s choices. They choose to sell to CoreWeave because CoreWeave is an Nvidia shop, right?

They’re wary of, say, an Amazon that is really working to build up Trainium as an alternative. They’ll sell to Amazon, but Amazon does not get nearly as many GPUs as they would like because Nvidia doesn’t want to sell them the GPUs.

Mm-hmm.

So there’s also an aspect here. It’s in Nvidia’s interest. On the one hand, they do want to dominate, but keeping a leash on some of these companies and having leverage over them is also something they enjoy having and work to preserve their moat and preserve their margins.

And then we have this whole H20 thing, which we’re going to get to. They could sell H20s to other people, or the production-line capacity they’re using for H20s could be used to build Blackwells.

Yeah.

So Nvidia is also not maximizing supply. It’s not doing everything it could do to have total supply.

Mm-hmm.

There’s a lot of internal deliberation at Nvidia, choosing customers and deciding who gets it. There are also real capacity issues, which is why I don’t know what’s going on. I think anyone trying to read too deeply into these numbers is making a mistake.

Then you also have the fact that, presuming it’s supply-constrained, they’re bringing new fabs online, new packaging facilities online—all these sorts of things. But they’re not coming online as fast as demand is increasing.

Also, Nvidia—it’s like the law of large numbers. If you add 100 units of capacity to a company that’s selling 100 units, you just doubled its sales potential. But if the company is now selling 1,000 units and you bring on 100 units, you’ve only increased its growth potential by 10%.

Mm-hmm.

Nvidia’s base is now getting so large that, of course, growth has to ameliorate because they just—

They can’t possibly compete with the exponential curve that they were on 2 years ago.

Right. So their absolute increase in sales may be higher than in previous quarters, but the percentage increase might be lower.

Yeah.

To the extent their sales are gated by supply, by definition, the growth rate, if it’s truly limited by supply, has to flatten out.

Has to go down.

Yep.

So, number one, people reading too much into this, I think, are dumb. Number two, I'm saying they're dumb not because I know the answer, but because you're trying to read things into something where there are so many variables. The core variable is that, if your assumption is that there's more demand than supply, it's impossible to know anything about the nature of that demand. We won't truly know what AI demand is until there's sufficient supply to meet it.

2. China Still Wants H20 Chips

Okay. Well, speaking of demand, the revenue projections for Q3 do not assume any shipments of the aforementioned H20 chips going to China. CFO Colette Kress said, “We're still waiting on several of the geopolitical issues going back and forth between the governments and the companies trying to determine their purchases and what they want to do. So it's still open at this time, and we're not exactly sure what the full amount will be this quarter. However, if more interest arrives, more licenses arrive, again, we can also still build additional H20s and ship more as well.”

So what do you think of what's going on there? Nvidia seems very determined to sell into China right now, but Chinese companies don't seem like they're willing or able to buy H20. What do you make of this situation? Regardless of what happens with the H20, there were also reports on Thursday that Nvidia is in talks with the White House to sell a modified version of the Blackwell chip to Chinese companies. This story is going to continue to unfold over the next 3 or 4 months.

Needless to say, there's a lot to unpack here. Companies in China are very willing to buy H20s.

Mm-hmm.

Even though it's a relatively limited chip, particularly for training, because the bandwidth is constrained, particularly the chip-to-chip bandwidth between the different chips. You can't build the huge systems. It is a very capable chip. It has a lot of memory onboard, particularly for inference, so it would be very useful, and it's better than anything that these companies can get in China.

Huawei is doing some interesting things, but they're fundamentally limited, first, by the chips they can get, which are mostly made by TSMC.

Yeah.

Of course they'd be willing to buy it. The problem is, this chip was designed to the Biden administration's specifications for exports. I don't understand why people in Washington get mad at Nvidia for this. The Biden administration set down some rules.

Mm-hmm.

So they built a chip that fit within those rules. Then people said, “That chip's too good.” It's like, well, then you should have done a better job making the rules, right? It's a little unfair.

What you should have done is institute a blanket ban on selling chips into China if you wanted to curtail that. By setting out these technocratic parameters, it incentivizes a company like Nvidia to—I mean, honestly, Nvidia would be irresponsible not to build a chip that works around the parameters they laid out.

Right. But also, such a ban would have been a terrible idea for reasons that we've talked about, which gets into why Nvidia wants to sell into China. So, just to review the timeline here: Nvidia makes this chip to the Biden administration's specifications. The Biden administration, I believe, was fairly annoyed at the H20, but again, what do you want Nvidia to do, to your point?

The reason they want to sell to China is because Nvidia's moat is about software. It's about a lot of things. It's about networking and overall performance, which is really important in the US because we're going to be power-constrained soon. That means the number of tokens you can generate per watt of power is going to be your limiting factor, and Nvidia will be able to charge a premium if they can stay on the cutting edge there.

But staying on the cutting edge is hard. Nvidia has done it for a long time, but the other thing they have is CUDA.

Mm-hmm.

When everyone is used to using CUDA, and CUDA is used to program these chips—to use them as a computer—it's a real lock-in. CUDA is clunky and it's hard to use, but it was totally groundbreaking. Nvidia created this entire space of programmable GPUs and being able to use them as a supercomputer, and it's a real lock-in.

Now, companies will work to get alternatives and build their own software stacks, but this is the same thing as the whole Intel x86 versus RISC debate, all the things we've talked about. This low-level software is very, very powerful as a lock-in. No one wants to do that work all over again.

Right.

If you want to hire someone to work on AI models, everyone knows CUDA. If you're a company building something, you have optionality in which cloud you go to because everyone has Nvidia. There are real ecosystem effects that are very, very powerful, and the reason why limiting Nvidia sales to China is such a risk for Nvidia is that life will find a way, right?

Mm-hmm.

The Chinese developers will build an alternative. It's a sufficiently large market that the alternative, particularly to the extent that it's open source, could gain traction. Actually, the best thing that's happened to Nvidia is that Huawei keeps trying to do its own version that's proprietary to Huawei.

Mm.

They should be doing a fully open-source alternative to CUDA and really leaning into that. But if that gets scale and traction in China, it's not just that Nvidia's ability to sell to China will be limited in the future. That alternative will spread to the West, and suddenly there's a meaningful alternative with real developer support and widespread use compared with CUDA, and it's going to work on chips other than Nvidia chips.

From Nvidia's perspective, China building up an alternative—or them not being able to sell to China—is a big problem. Right now, everyone in China would rather buy Nvidia chips because they're better. But if they can't buy Nvidia chips, then they will use something else. In the process of using something else, they will be incentivized to destroy Nvidia's software moat, and that destruction will not be limited to China. It will spread all over the world.

So this is a big problem for Nvidia. It's not just a China problem; it's a future-of-the-company problem.

Yeah.

This is why it's really important to them to sell to China. From my perspective, it should be important to the United States to sell to China.

There's this whole mindset about going into China where, for a lot of people who have been appropriately skeptical about the whole outsourcing thing and have, in many cases, been proven right, they're immediately like, “Oh, yeah, another company selling to China, and now they're getting killed, da-da-da-da.”

No, that was transferring technology that was limited to a single object—car technology, metallurgy, whatever it might be. We're talking about network effects. Network effects are totally different from the technology that we're talking about. What Nvidia is trying to preserve are the network effects around CUDA. Network effects depend on everyone using the network.

Mm-hmm.

That's why they want to sell there. That's why I think the US government should allow them to sell there.

And they want to preserve it not only in China, but all over the world. That's the long-term existential concern.

No, because if they lose it in China, it's going to seep out. Yeah.

Yeah.

The software will seep out. Even if you put a ban on Huawei chips in the US, developing a CUDA alternative that has real mass and momentum behind it is going to diminish Nvidia's advantage, pricing power, and all the sort of things in the West.

Mm-hmm.

From an AMD perspective, this China ban stuff means they're losing sales in China, but in the very long run, it's probably a good thing for them because it's creating the conditions to break down the Nvidia moat, not just in China, but in the West. Once the moat's broken down and you get to more general-purpose software that runs on anything, they'll just buy it in an AMD chip instead.

Yeah.

This is why it is actually a really big deal and a problem for Nvidia. That's why they're very, very concerned about it, above and beyond the fact that, of course, they'd like to sell to China because there's a whole lot of money to be made. It's the second-most important market in the world. But that one's easy to understand.

I felt like an idiot asking that because obviously it's $10 billion in revenue, or whatever the number is, available in China every quarter if they were allowed to sell into it.

It would be massively more than that if they could sell unconstrained to China. All this money they're getting from selling to these hyperscalers in the US, China would get a whole bunch, too.

But it's interesting because, to the extent that their sales are gated by their ability to create enough chips to meet demand, China has been an existential priority.

When the H20s were banned earlier this year, back in April, there’s been a lot of back and forth over the last 6 months. But the next day, Jensen Huang got on a private jet and flew to Beijing to meet with PRC leadership to smooth things over. It’s been a real driving concern for 2 years now, and it’s happening at a super interesting time in China because the Chinese government may realize some of what is happening here. They’ve banned companies from buying the H20s, and it’s not clear why they’re doing that or what the endgame is.

3. Washington Chip Controls Backfire

This is where the Trump administration really screwed up, and they want to put all the blame on the Biden administration. I think right now they got to the right place, which is: You should let Nvidia sell into China. You should want the American technological stack to be dominant all over the world. The problem is, they banned the H20 sales in April. It was dumb.

Yeah.

It raised the prominence of this issue, and it’s like they’re trying to just undo it. You talk to them, and they’re like, “We didn’t ban it. We just imposed a licensing regime.” What? The pedantry they want to engage in with these arguments instead of just being like, “Yeah, we screwed up”—that was a bummer.

Yeah.

If they just let it flow through and let Nvidia sell H20s—

Mm-hmm.

Again, I hate ascribing Chinese actions to the U.S., right? China is an independent actor. They do things for their own reasons. They’re a smart, talented, powerful country. All those things are true. Way too much China analysis just ascribes everything to the U.S.

A response to what we’re doing, right?

Yeah. Everyone else in the world is an NPC. We are the only actors, and that is a mistake. I say all that to say: Why did the administration do this?

Well, a couple things happened. They—

And then they undo the ban, and they’re fighting domestically, and they’re like, “Oh, don’t worry. We’re selling to China because we’re going to own it all.” The Chinese government is like, “Wait. This sounds like a big problem. Maybe we shouldn’t allow this to happen.”

That’s what’s funny about it. Obviously, I think your point is, had they just continued selling H20s all year long with no back and forth on whether these are banned or export-controlled, the Chinese government never takes note of whether these chips are strategically beneficial or not. Then they unbanned it in July, and Howard Lutnick said on CNBC, “We don’t sell them our best stuff, not our second-best stuff, not even our third-best. You want to sell the Chinese enough that their developers get addicted to the American technology stack. That’s the thinking.” So he was explicitly laying out the strategy on the American side.

Which is all totally correct; it’s what you should do. But you probably shouldn’t talk about it on TV. And, again, I think the administration got to the right point in my estimation. I wrote before they banned it. Remember, we had the long discussion about chip precariousness, and then they... The whole thing was about how I think they should let China build on TSMC, and I think they should let... The chip ban is a bad idea. I’m fully opposed to it, and I laid out all the reasons why. But the problem is, I’m glad they listen to me now.

Mm-hmm.

But—and it’s, again, it’s frustrating talking to this team because they... Sorry, you screwed up. You don’t get to undo it. The April ban was a problem.

Well, that was the most frustrating thing about your chip piece. I think the thinking there would’ve been really effective if it were implemented in 2018 and 2019, before moving forward with various semiconductor controls, and obviously it intensified under Biden. Now it seems like the horse is out of the barn to some degree.

Yeah. It’s tricky. Is it already too late? The one single most effective chip control was not allowing EUV machines into China. And so was that one control sort of worth all the other ones that made me make the case?

It could be. Yeah.

It’s sort of a baby-and-bathwater thing. It’s really easy, especially when you’re dealing with government, to... You can criticize literally everything that happens because it’s never going to be as sophisticated or perfect. This is a mistake people make all the time because it’s easy: You could point out the problems with everything.

You have to take the package as a whole. So if you take the package as a whole, including the Biden administration, including the Trump administration, has the overall thrust been correct simply thanks to the EUV decision? Which, by the way—

Hmm.

—was one of the very first ones. Everything else has been bad, arguably. You could make a case that it is.

Yeah.

And by the way, a hulking piece of equipment that is easier to control and restrict than—

Exactly.

—chips as they’re transiting the world.

Exactly. Exactly. So, just to review, I’m happy to give China the best Nvidia chips. I’m happy to let them fab all they want at TSMC, and I want a total ban on semiconductor equipment because I want them dependent on Taiwan. And by the way, if the administration wants to undo their H20 screw-up, guess what might actually overcome the Chinese government’s consternation at the U.S. flaunting its technological superiority?

Selling black-box chips?

Actually giving them technologically superior chips, right?

Yeah.

They’re just so much better than what China can do domestically that the government has to be pragmatic and think, “Look, the benefits we will get from AI, instead of waiting an indeterminate amount of time till we can do better, we’ll take it.”

Well, it’ll be an interesting experiment because it’s hard to know what’s driving the thinking in Beijing on some of these issues. I think comments like Lutnick’s, in terms of ascribing too much agency to the American side of these dynamics—I think what Lutnick said was perceived as insulting, according to a Financial Times report. But the idea that that alone drove the policy shifts strikes me as—I don’t know. I don’t really buy that.

You can argue that China’s doing the right thing. And I—

Well, that’s the thing. I would say—

No, because I’ve made this... I talk about this in the context of chip controls. The best thing that China could do is—when China released that chip that everyone flipped out about in Washington that had bad yields or whatever, I said, “This is actually bad for China because what they need to do is build up their semiconductor equipment manufacturing industry from the ground up.” They need to be fully indigenous because the U.S. has—

Mm-hmm.

Obviously, I’m on the U.S. side, but the best thing for China in the very, very long run is to fully control the stack. Clearly, it’s the most important technology in the world, and the U.S. is going to leverage it. They already have, right? So China making its own chips with a bunch of U.S. equipment is actually preserving U.S. control over its industry. To the extent it can build up its own lithography, build up its own etching, build up all the actual pieces that go into photoresist, that go into these processes, that would be good.

The problem is that’s economically insane, because if you can just buy it, why would you do it otherwise?

Yeah.

And in the “Hey, we’re making our own chips” narrative, don’t look at the fact that it’s all Chinese and Japanese chemicals and all American equipment and European equipment, or whatever it might be. That’s sort of the reality today.

And so this is a similar situation where there’s a pragmatic decision facing the Chinese government. Do we want the best chips for AI right now, or do we want to fall behind potentially for years or decades so that we bring our own system to bear, but in the long run we’ll be truly independent and not dependent on the U.S.? That’s a tough decision.

Yeah. Well, I think the reality is there are probably people within the Chinese government who want to just rip the Band-Aid off and go that direction now, and comments like the ones we got from Lutnick in mid-July probably—

Yeah, it gives them—

—help them make their case.

—gives them real ammunition. Yep.

Yeah.

Absolutely.

And there’s also a possibility that they’re just putting the screws to Nvidia and Jensen Huang, saying, “Go back to the U.S. government and lobby for relaxing restrictions on Blackwell chips.” But if they are in fact allowed to sell Blackwell chips, that’ll be the tell.

They should be allowed to sell, and they might have to be allowed to sell to undo the April screw-up.

Yes. Well, more will be revealed on that front. For now, 2 other notes from the call.

CFO Colette Kress again responded to a question about Nvidia’s deal to pay the U.S. government 15% of its China revenue:

“When I first heard it, I said, ‘I’m sorry, where’s the regulation?’ I said, ‘No wired money goes out of this company based on someone told me. I cannot do that. I have policies, procedures, and rules—SOX, everything else—that will say there is nothing you can do there. They understand, and we have been communicating that. I’m just saying, if nothing shows up, I’ve got licenses. I don’t have to do this 15% until I see something that is a true regulatory document. So just keep that in mind.’”

I’m just noting that for the record. We didn’t record last week, and we didn’t mention it 2 weeks ago. It’s such an insane structure that the government has proposed here. It’s not clear whether it’s legal, and it’s not clear whether it will even move forward, at least based on the comments yesterday. I don’t know whether you have any thoughts, but I’m just closing the loop on the bizarre story from the last 2 weeks.

Yeah. I don’t know. I guess the administration’s position—because a lot of people, including me, were just like, “This seems clearly unconstitutional. You can’t tariff exports.”

Yep.

The reason for that was the South insisted that be put in the Constitution because they were worried the North was going to tariff their cotton exports and tobacco exports. Literally, this isn’t an amendment, and it isn’t a law. It’s in the Constitution.

Mm-hmm.

I think the argument is that the Nvidia chips are made abroad. They’re never actually entering the U.S.

They don’t touch our shores.

That’s right. I don’t know. That’s for the courts to decide. But who knows?

Yeah.

Who knows?

Who the hell knows?

I guess I’m touching questionable-legality topics on a regular basis here, but it is what it is.

It’s all part of the fun. Speaking of controversial plans of questionable legality, in a Reuters report this week, Treasury Secretary Scott Bessent was asked about the U.S. taking a stake in Nvidia, and he said, “I don’t think Nvidia needs financial support, so that seems not on the table right now.” That was on Fox Business.

So we can all rest easy. All the libertarian babies can have their bottle. We’re not nationalizing Nvidia this week. But that brings us to segment 2.

4. Intel Needs Government Backing

For anyone who’s not familiar with the particulars, I will read from CNBC:

“Intel on Monday warned of adverse reactions from investors, employees, and others to the Trump administration taking a 10% stake in the company in a filing citing risks involved with the deal. The deal, which was announced Friday, gives the Department of Commerce up to 433.3 million shares of the company, which is dilutive to existing shareholders. The purchase of shares is being funded largely by money already awarded to Intel under President Joe Biden’s CHIPS Act. Intel has already received $2.2 billion from the program and is set for another $5.7 billion. A separate federal program awarded $3.2 billion, for a total of $11.1 billion, according to a release. Trump called the agreement ‘a great deal for America,’ and said the building of advanced chips is fundamental to the future of our nation.”

You wrote about this deal at length on Tuesday. We’ve mooted the idea of a U.S. government–Intel partnership at various points in the past on this podcast.

No, I think on Sharp Tech specifically, I’ve said Intel is barreling toward nationalization. This might be one of the all-time being-right points. I was surprised at how surprised people were about this happening.

So was I.

Again, it’s just because we’ve been talking about it so much. Sadly, not everyone listens to Sharp Tech. I believe we used the specific words: Intel might need to be nationalized.

Yeah.

So we can get to why, but this isn’t out of the blue by any means.

Yes, and I made fun of you when you raised that possibility. I called you Comrade Ben on the podcast. So, again, you foretold a lot of the discourse that has been swirling for the last 10 days or so.

Joshua says, “Ben, why is the U.S. government taking a stake in Intel a better solution than further incentivizing investment in U.S. fabs, whether by TSMC, Samsung, or Intel? Your writing has shifted my opinion on how we should engage with China on chips. It’s also convinced me that Intel is more likely than not a lost cause, but that incentivizing others to build U.S. capacity is feasible.

“In the event of military conflict with China or Chinese action against Taiwan, wouldn’t TSMC and Samsung be even more incentivized to support their U.S. plants? And in this downside scenario, nationalizing those plants would then be on the table. To put it even more simply, if China bombed TSMC and Samsung plants, who would you trust more? Intel rushing in to adapt and meet bigger demand, or TSMC and Samsung executives rallying to accomplish this?”

What do you think? We got a number of fun responses, but we can start there.

5. Foundry Knowledge Cannot Be Replaced

I think this is a misunderstanding or an incomplete understanding of what we’re trying to preserve. It’s not the equipment, right? This is sort of the same answer you see from a lot of people: they get a cursory understanding of the semiconductor value chain, and they’re like, “Well, actually, ASML’s the most important company in the world, because they’re the only one that makes EUV, which TSMC needs to make chips.”

And then they start talking about ZEISS, the camera company.

That’s right. “Oh, actually, ZEISS is the most important. Actually, no, it’s DRUMPF,” which is hilarious, this whole thing. The U.S.’s key point of leverage over ASML is a laser that’s made in San Diego by a company that is called DRUMPF, D-R-U-M-P-F, which ASML acquired. But it’s all, you know.

Almost all of EUV is U.S. technology. Is it called DRUMPF? No, then I’d be wrong about DR— It’s something like DRUMPF. I don’t know. It’s something quick.

DRUMPF, the John Oliver meme? I hope it’s not spelled the same way.

No, sorry. I got stuck on the John Oliver spelling. No, it’s TRUMPF, T-R-U-M-P-F. So this whole thing might be lending credence to this all being a simulation. It is a little ridiculous. The U.S.’s point of leverage over ASML is a company called TRUMPF—

Yep.

—that is based in San Diego that ASML acquired.

Which then gives them the leverage to invoke the foreign direct product rule.

Yeah, TRUMPF lets Trump trump ASML’s ability to sell. I’m literally not joking.

Wonderful way to close out the summer here.

It’s totally true.

It all descends into chaos on the podcast.

There is tremendous expertise—and the most irreplaceable expertise in the world—in actually putting all this together. A lot of the equipment, including ASML, comes from Intel. Intel invented EUV, like way back when.

TSMC co-designs a lot of these processes with its partners. The point of integration is the foundry: knowing how to do this, getting all this stuff together, and coordinating with all these different companies helps them figure out these problems. There’s all this co-development, and there’s tacit knowledge that builds up over decades. You can’t just learn it overnight.

You just know what to do. You know how to solve this problem. You’ve seen it before. It’s the pattern matching and the understanding that goes into it. It’s irreplaceable. It cannot be replaced.

So, number 1, your number 1 problem with TSMC in a war scenario is that the people with that knowledge in their heads are probably dead.

Mm.

Right? Now, there’s the idea that, “Oh, maybe we’re going to get a plane and get them out of there.” Well, maybe you will, maybe you will not. That’s problem number 1.

Problem number 2 is TSMC and the Taiwan government in particular, which obviously has a say in TSMC. Trump complaining about TSMC—or, sorry, Taiwan—holding up the U.S. by virtue of chips is totally correct.

Mm.

That is basically the foundation of Taiwan’s defense. They don’t spend enough on defense. Their army’s kind of a mess. They’re like, “You need TSMC. You’re going to defend us.” That’s maybe a little harsh, but not that harsh, if we’re being totally honest. I’m now in the U.S. They can’t come get me, so I can be honest about this, right?

There you go.

Welcome home.

So the problem is that all those plants could be nationalized. But the problem with nationalization, which we talk about in socialist countries or communist countries, or apparently now the United States, is that you have a bunch of equipment. What are you going to do with it?

Right.

There are all these stories of them nationalizing the oil company in the Third World and saying, “We’re going to get all the money and profits.” Then what happens? It all breaks down, and their production falls.

Actually, because they don't know what they're doing. They don't know how to use it, and all the talent fled. Right? So the problem is you can have all this stuff in the U.S., but if you don't have the knowledge and the understanding of how to use it, it might work for a while, but you've lost it, and you're not going to rebuild it.

The danger of losing Intel is that you lose a source of the knowledge—the understanding—of how to use this stuff. And yes, TSMC and Samsung might build foundries in the U.S., and that's definitely better than nothing. I compliment both the Trump and Biden administrations for the extent to which they have incentivized and pushed TSMC, which didn't and doesn't want to do this. The Taiwan government does not want them to do it, but they've had so much pressure applied to them that they have done it, and they are getting economies of scale from going into Arizona. That is a good thing from the U.S. perspective.

It doesn't change the fact that this is a Taiwanese company and this is a South Korean company. And number 1, they may be allies, but they're not ultimately loyal to the U.S.

Mm-hmm.

Number 2, their most important people, most important source of knowledge, most important R&D centers—all those sorts of things—

On the other side of the world.

—are within, like, a 5-minute missile flight from China.

Yeah.

And the reality is, if we're in a real war, what—

There's certainly incentive to take out TSMC.

—you don't get to coordinate an evacuation and manage this sort of exit. This is just the root of it. And it's a hard thing because, of course, our hope is this never happens.

Mm-hmm.

But you can't plan that way. If Intel goes away and gets out of the foundry business, at the end of the day, you are basically committed forever to Taiwan and South Korea funding the most important technology that exists.

Yeah.

Right now, if TSMC went away and Samsung went away, Intel, despite the mess that they are, is by far the best option to fill in the gap. It's not even close. It's a much better option than a bunch of TSMC plants in Arizona where all the head engineers are in Taiwan. It's just—and so this is the core.

I completely understand and, by and large, agree with all of the objections to the deal. But what I find intensely frustrating about this discussion is that sometimes—this goes back to my political discussion before—you are presented with an array of horrific options, and you have to pick one.

Did I enjoy writing this article? Not really. I'm not a socialist. I don't like it. And, of course, the Trump manner is objectionable. "Oh, yeah, we're going to take shares of lots of companies. We're TSMC." It's like, can you just shut up?

Yeah.

Then you get lumped in with defending all this sort of thing, and Intel doesn't deserve it. They're a poorly managed company. Their culture is a problem. They've lost a lot. All these things are totally true. I just do think that chips are different.

Now, why ownership? The problem is, I think a year ago you could have gotten there with purchase guarantees. "Your 18A is going to be viable, so we're going to buy X number. We're going to have these incentives attached." The problem is that since then—and, by the way, Lip-Bu Tan might have completely played Trump, to be totally honest—

How so?

When he comes out on the earnings call and says, "We're not going to do 14A unless we have an external customer," he's basically guaranteeing that Intel does not get an external customer.

Totally. I mean, it would be irresponsible for anybody to sign up to be Intel's customer when they might not be in that business in 3 years. You're not going to dedicate a bunch of resources to designing chips for a foundry that might not be producing chips in 24 to 36 months.

Yeah, and I don't want to overly accuse Tan, because he's—it’s also likely correct. If Intel tried to do 14A and their only customer were themselves, their business is under massive pressure. They're getting their ass kicked by AMD, in particular, and x86 is just fading in importance generally, particularly as Arm becomes a bigger thing in the data center and all these sorts of things, and all the money's going to GPUs.

This is kind of like the Lutnick thing. He might be saying the truth, but he's also sort of guaranteeing a certain outcome by saying that.

Yeah.

That's why, in my estimation, it had to be ownership. You need some sort of credible guarantee that Intel Foundry is going to exist and will continue to exist.

And at this point, beyond the fact that Intel needs financial help, as we've been saying for over a year—

Mm-hmm.

—or maybe even longer than that, but very clearly for over a year—is the fact that they now need to credibly promise this would-be external customer that they will be around, and they can't do that. Intel cannot do that on their own.

Yeah.

And now it's more credible. Then, some of the objections: "Oh, Trump's going to browbeat companies into using Intel, and they're better off using TSMC." Yeah, that's kind of what needs to happen, right? This is the other thing: There are 2 pieces. It's the national security bit, and there's the fact that no company wants to choose Intel.

Yeah.

Because it's not worth it. TSMC is there to serve their needs. It's reliable. Everything about the experience is better. And so it's not a market failure. It's a market success.

Mm-hmm.

The problem is, sometimes market successes are in direct conflict with national security concerns—not just in the next year or 2, but in the next 10, 20, 100 years. That's the time period we're talking about.

Yeah. TSMC is the rational outcome of a globalized free market, but the problem is that—I don't know how many miles it is from—

It's actually totally normal that TSMC won. The costs are so astronomical, it makes sense that one company does this, and this has happened. There used to be 5, 6, 10 foundries.

That's the thing. It makes sense for there to be one company doing this.

And every time it gets more and more expensive, foundries drop off. AMD split, and then GlobalFoundries fell off, and then IBM fell off. This has been happening for a long time because it's so expensive that whoever gets there first and is in the best financial position survives, and other people throw in the towel.

The economically rational thing for everyone involved is for Intel to give up.

Right.

The problem is that's bad for the U.S. So what do you do, then? What's the solution the people who hate this suggest?

Yeah. Well, that's the question. Do you want to read another objection to your takes here?

Bring it on.

This invokes antitrust.

Bring it, Phil.

T.K. says, "Subject: The road to hell is paved with U.S. v. Intel. Andrew and Ben, the opening blurb from Ben steelmanning the case for U.S. Intel spurred me to strawman Ben's admittedly eloquent argument.

"The outline of the argument for partial nationalization of Intel parallels the original argument Ben had on antitrust intervention against Facebook regarding it acquiring social networks. That is: I am generally wary of state intervention, but this time it's different because the target company, be it Intel or Instagram, is in this particular circumstance. And 3, intervention could be the best path in this case to get to this particularly targeted outcome, be it preventing social networks from acquiring other social networks or maintaining U.S. chip fabrication capabilities.

"Ben seems to have recanted his position on antitrust because he found that government intervention is a blunt instrument, and he has learned that he cannot count on the execution of policy to be nuanced enough to be actually effective.

"Given that, I'm keen to hear how he has thought around the long-run second-order effects of having a U.S. government that is taking equity stakes in private companies for the fulfillment of policy goals. Does this precedent establish a slippery slope? How far could the concept of national interest be stretched? Is this a power that Ben would trust future administrations to wield regardless of their political stripes?

"Any thoughts on that, Ben?"

6. Nationalization Sets A Dangerous Precedent

Great email. I think it's a totally fair analogy. I do think it's different, which, of course, I will explain. But to his point: Does this precedent establish a slippery slope? Yes. How far could the concept of national interest be stretched?

Pretty far.

Unfortunately far.

Uncomfortably far.

Is this a power that Ben would trust future administrations to wield regardless of their political stripe? No. Do I trust this power with this administration? Also no.

Mm.

I think the challenge—that was why I opened with the steelmanning sort of bit—is that I acknowledge and agree with basically almost every single objection to this deal. It’s terrible. All these things are true and bad.

Mm.

The distinction—and so the point about whether you’re going to get appropriate nuance—I have learned that lesson, and I think his point that we’ll usually have appropriate nuance here is fair. The difference is that the stakes for the antitrust stuff—another shift in my thinking—are lower.

Yeah.

To the extent that I think the way TikTok took over and threatened Facebook was a learning for me, too, right? Facebook’s dominance was pretty short-lived. The market actually fixed the problem.

Mm-hmm.

And so that was a—my recanting wasn’t just, “Do I think the government should have banned the acquisition of Instagram?” Yes, I actually still think that is the case. The reason why I’m opposed to it is because I’ve seen, number 1, they can’t distinguish—

It wouldn’t stop there.

—between acquiring Instagram—

Yeah.

—and all these other things. And number 2, given this, the stakes aren’t high enough to risk getting it wrong and screwing it all up. Actually, the market works out pretty well anyway.

I’m Mr. Pro-Market, questioning it now, but most of our emails over the years are annoyed at me being too pro-market in many respects. I realize that the stakes aren’t high enough to run the risk of the government getting it wrong. The distinction I would draw here is that I acknowledge and accept all these things. It’s probably going to go badly. It’s going to be abused either by this administration or future ones, but the chip issue is so fundamental, and Intel is so irreplaceable, that I am acknowledging and accepting all of that and making a hard choice.

Mm-hmm.

All I would encourage everyone who pushed back on me to do is make sure you’ve thought it through. Make sure you’ve steelmanned the case. And the steelmanning can’t be—I was in an argument in a group chat about this, and someone’s like, “China’s not invading Taiwan anytime soon.” I’m like, “No, I agree.” But the problem isn’t in 2025 or 2026 or 2027; it’s in 2035, 2036—

Yeah.

—2037. You’re not going to turn around and rebuild a foundry at that point. This is actually a 10-, 20-, 30-year decision.

The other piece of this is that a lot of people who are advocating for the free market to solve these problems will say it’s healthy to let companies die. That’s been sort of a consensus view.

Which I agree with.

Yeah, but if Intel dies, there’s not going to be an American chipmaker that takes its place over the next 10 years, and that is a critical problem.

Nope.

Yeah.

People—this is a real problem—conflate NVIDIA, AMD, and Intel. NVIDIA doesn’t make anything.

Yeah.

TSMC makes it. There is a distinction between those companies. What we’re seeking to preserve is not—

Intel’s rational course is probably to spin off the foundry and just focus on its x86 business. It’s profitable, it’s shrinking, and, again, x86 in general is being threatened. AMD is kicking Intel’s rear end in x86 specifically. But it’s still a profitable company that could be a P&L throwing off lots of cash. It has lots of locked-in customers, all that sort of thing. That’s probably the very rational thing to do. From the U.S. perspective, it doesn’t give a flying fuck about x86.

Mm.

Actually, my preferred approach here is splitting off the foundry from Intel. Intel is the problem here. No one wants to build stuff on Intel because then Intel’s going to steal their designs.

Well, and also, I’m wary of endorsing what’s happening here because, based on all of our previous conversations about Intel’s broken culture, I’m skeptical that any approach is going to work at Intel. I understand why the government’s doing this, and it’s coherent, but—

This has been my maturing.

Okay.

What the Trump administration did this week is not the way I would have handled it.

Okay.

But it’s what was done. So do I support it or not? At the end of the day, I would have done it differently. I would have acted a long time ago. A year ago, in “Intel Honesty,” I said very explicitly that the government has to guarantee Intel demand.

Mm-hmm.

And everyone today is like, “Why does the government guarantee Intel demand?” Because it’s not 2024. That’s the problem. The problem right now is that Intel basically signaled its death on the last earnings call. That’s just the reality.

Again, was it intentional? Was it not? I don’t know. It certainly was, I think, overall, played very well. The choice isn’t—everyone, when they present these arguments, wants to choose the ideal outcome of what they would do versus what happened. That’s not how the real world works. The options in front of us are not taking a stake in Intel and Trump, through whatever means, in his own certainly unique style, taking a stake. That’s the choice.

Yeah.

And this is the better choice, in my estimation, even though it stinks up and down.

And without the government backstopping Intel and lending a bit of credibility to the idea that it’s still going to be in the chipmaking business 5 years from now, that business—on the leading edge, at least—just sort of withers and dies on the vine in that scenario. That would be the fear.

Right. And this ties into all the trade stuff, right?

Oh, it ties into everything. Honestly, the conversation around this was frustrating for me to live through over the last 7 or 8 days. Granted, part of this is that you and I have been talking about it for a year and a half, so maybe I’m just closer to it and most people haven’t really been paying attention.

But on the right, there was a consensus horror because of what this means in the context of capitalism, and the idea of nationalizing anything is a problem. The left is just happy to take shots at the deal because it’s Trump. It’s crazy to me to see all these people in D.C. In D.C., everyone on both sides is very concerned about China. They see the relationship with China as an existential issue that should drive all sorts of different policies around this town. And then you—

And then when someone actually does something—

Well, you see news like this that produces these hysterical reactions, and it’s like, have any of you guys thought through 3 steps of what a conflict with China would actually look like—

Yep.

—and what the impact on American life would actually be?

The antitrust email is actually a good example, because capitalism is awesome, and the free market does deserve deference, but there are certain cases where companies start optimizing for market share as opposed to profit. At a certain point, that can break capitalism, or at least be a threat to capitalism, and produce worse outcomes—

Speak for yourself, but yes, continue.

—than capitalism generally does.

The same thing is true when countries are not rational actors. China, literally in industry after industry, is optimizing for market share as opposed to profit. But also, China could bomb TSMC. That would not be an economically rational decision, but it would be a big problem for the entire world.

Well, it was all economically rational to outsource all our manufacturing to China.

Right.

Would anyone like to undo that at this point? It sounds like it. It seems like it. So now you’re presented with a very similar sort of dynamic, and suddenly no one wants to do anything, right?

Yeah.

The problem is that they’re all bad choices, and people only want to endorse good choices. It sucks to endorse a bad choice, right?

Again, this isn’t the sort of article that I enjoy writing. I was up till 4:00 a.m. writing it, in part because I knew my take, and I just didn’t want to write it.

Yeah.

Well, nobody wants to make a full-throated defense of what’s happening here. It’s an unfortunate situation, and it’s not like either one of us is Team Intel at the end of the day.

Right. And by the way, I do respect and appreciate everyone who is still opposed to it.

Yeah.

It’s not an easy issue at all, and I appreciate and respect people on the other side of the chip controls and all those sorts of things.

My hope is that the people who do oppose it—and I think part of my frustration was born, like I said, from some group-chat arguing that was happening over the weekend about this—the people who were just declaring this bad from the top rope and not even being willing to engage with the reasons why it might be justifiable, because they’re living in theory land. “This is bad in theory, therefore I don’t even need to listen to your argument, because it’s bad.”

You see this in politics, in political discourse, a lot.

Just not even engaging with the other side because, “I’ve decided there is a particular principle on which it is bad.” And therefore—

Yeah.

—I don’t even need to entertain counterarguments. All I would ask of the people who disagree is that they make sure they’ve thought through and articulated the counterarguments, because they do exist. Don’t be that person who’s just like, “Nope, this is socialism. We don’t do this in the US. It’s bad.”

Yeah. Well, I would argue that the government is going to have to make these compromises in a number of areas over the next several years, and will do so to preserve the long-term viability of the US system, the Western capitalist system.

If you want to undo globalization and our total dependence on China, it’s not going to happen in the free market.

Right. Look at what we’re doing with rare earths right now. We’re going to have to do the same thing with chip building.

It’s the exact same thing.

Yeah.

It’s the exact same— No, rare earths is actually a perfect example. By the way, where were all the objections about rare earths? The Trump administration did the exact same thing three weeks ago.

Yeah.

It’s literally the exact same problem. There is no economic rationale to build rare earth mining and particularly refining capacity. It’s extremely expensive. It’s bad for the environment, so you’re going to deal with tons of lawsuits. It’s just a horrible industry to go into.

If you did go into it, China can flood the market, and you’re not going to make any money. So the rare earth issue, on which China has the US, is why people complain about Trump going easy on China versus the EU or whatever. It’s because China has leverage over the US. This isn’t kissing up to China. It’s a reflection of reality.

Yeah.

And because they have leverage over the US, the US needs to develop its own rare earth capacity. Now, should we go back and undo the Clinton administration selling it away, or the Bush administration, whoever it was? I can’t even remember which administration sold us up the river as far as rare earths. It might have been Obama. Whatever, it was one of them. Unfortunately, we can’t do that.

Yeah.

We have to accept the world as it is, which is that we don’t have access to rare earths. We’re totally dependent on China, and China has demonstrated a willingness to use that dependence. To undo that, there is no market solution. It has to be government action. By the way—

And by the way, China understands this.

This is our version of China. Yes. China—this is the China debate we talked about with chips. Are they going to accept the US having leverage over them, or are they going to brute-force an alternative?

Yeah.

This is just the reality. By the way, this is what should give people some comfort: all the things that drive us crazy about our dependence on China drive China crazy about its dependence on us. It is a two-way road. I can promise both countries have a lot of regrets, okay? So that’s actually the good news.

Mm-hmm.

This is why there probably won’t be a war. We probably will just continue on, and Taiwan is going to be in this weird middle situation. This is sort of my default position.

Yeah.

Nothing’s going to happen precisely because we are so enmeshed. Maybe this is where the grand world planners, particularly post–World War II, actually did get it right. If there were ever to be a world war, it would be the US versus China.

Mm-hmm.

And the reason it won’t happen is because we’re so freaking enmeshed that we can’t do it. We’re just going to muddle along.

Yeah. Well, it would be irresponsible to proceed on that assumption, so I understand trying to save Intel and do what they can.

What I would respect is an emailer emailing in saying, “Actually, the US should not do this, because the U.S. is increasing its dependence on China.” It would be like, “Ben, you said that you want China dependent on Taiwan and TSMC. Shouldn’t we also want the US to be dependent on China because—”

Chinese rare earths.

—so that they— No, exactly, right? That, by the way, there’s the take. Someone can take it if they want. The US should not be building chip independence. They should not be building rare earth independence. We should all be invested in economic enmeshment, because in the long run that is actually what it is. It’s basically like the nuclear strategy.

Well, that does require two to tango, though. It would require the PRC to stand down on its domestication efforts on the chip side. But in any event, yes, we’ll see what happens over the next couple of years. I’m not going to bet the house on Intel figuring this out, but I think it’s worth a try, and I understand where the government’s coming from. All the concerns from the emailer TK are very fair concerns to have.

No, I mean, I’m probably very annoying to argue with on this point, because basically every objection someone raises, I’m going to grant to them. “Yep, that’s right. That’s a new problem. Yep, you’re correct. Yes, I completely agree.” And you’re like, “Isn’t this a slippery slope?” It’s slipperier than you can imagine, okay? It’s going to be—

But do you identify with my reaction? Watching people process this news, I was like, “Have you guys just not been paying attention to anything over the last 10 years?”

Oh, yeah, completely. It’s maddening.

It’s pretty wild. In any event, to keep it moving and go in a completely different direction, I’m going to read a text message from my friend Ed, who wrote on Wednesday: “Please tell Ben I was very upset to see Demon Hunters in my inbox. My kids are obsessed, and I can’t escape it. It’s all they want to listen to. ‘Golden’ and ‘What It Sounds Like’ are now stuck in my head constantly.”

So there you go, Ben. Maybe next time you consider a trigger warning of sorts for young parents out there worried about K-pop-related daily updates. I don’t know. What do you have to say for yourself?

I mean, that was a very abrupt shift. We’re into Intel, political things, war, and then it’s like, wait, what are you talking about?

Catastrophe.

Yeah, and we’re talking about K-pop, baby.

Elsewhere in Asia.

That’s right.

Let’s talk.

Speaking of South Korea—yeah, KPop Demon Hunters. My kids roped me into watching it a few weeks ago, and I was hooked. It’s good. I recommend it.

What’s good about it is—I mean, the music is insane. It’s insanely catchy, basically every K-pop cliché just on steroids. But there’s an edge to it. My favorite scene is where the heroes are kind of failing, and the risk of them failing is people losing their souls to the demons. They’re in a train fighting, and they fail, and the train’s empty because everyone gets— They just depicted the massacre of hundreds of people. It’s kind of dark, which is great. It feels like there are stakes. Anyhow, I endorse the movie—

Edgy tween content.

—it’s a good view. Yeah. But it’s a fascinating business story.

Yes. I’ll get to the business story momentarily. Context for anybody who has no idea what we’re talking about: this is the most popular film in the history of the Netflix platform as we record this. 236 million views was the number earlier this week.

And by the way, the graph of those views is a straight line. It’s not leveling out.

Yeah. We’ll see where it nets out after the next couple of weeks here. It’s an animated musical film developed by Sony. The movie centers on a fallen K-pop girl group that secretly doubles as demon hunters. After their career falters, the group comes together again, balancing the flashy, competitive world of K-pop with the supernatural battle against demons threatening humanity. It blends K-pop culture, action fantasy, and musical storytelling.

I’m sorry. Have you seen this movie?

I have not seen this movie, and I’m all set.

Where did you— And number two, where did you get this summary of the movie?

I got it from some article. Why?

It’s a terrible summary.

What is it?

If you’ve seen it, they’re not a fallen K-pop girl group. They’re at the top of the game. They’re number one. Their career didn’t falter. It’s just a terrible summary.

Look at you getting all defensive on behalf of the K-pop demon hunters.

I know.

Okay. Fair enough.

Yeah. You’ve got to watch it.

Well, it was also released in theaters last weekend and was the biggest movie in America for the limited release that Netflix put together. Matt Belloni at Puck writes, “Sony likes to say that it won the streaming wars by sitting them out and refusing to shovel money into the furnace of a mass-market service. But here’s an instance where the incentive to sell to the highest bidder ended up causing Sony to potentially leave massive amounts of money on the table, at least for this one title.

Other vertically integrated entertainment companies would probably have held on to a theatrical window for a project that their executives believed could be successful, or hedged by sending it directly to their owned-and-operated streamer. In either scenario, there could have been a direct-to-streaming sequel that then could be elevated to theaters if desired. Remember, Disney developed Moana 2 as a Disney+ series, then refashioned it as a theatrical film, and it grossed $1 billion.”

So Sony didn’t do that. They shared the costs with Netflix. I believe Netflix bore the cost of $100 million to develop this movie.

Yep. Basically, the deal was that Sony gets a 25% guaranteed profit margin—guaranteed, very important point—or $20 million, whichever is higher.

Mm-hmm.

In this case, the film cost $100 million, so it was $20 million. So Netflix paid the $100 million—

Yep.

That was the cost to do it, and then Sony made $20 million on top of it. So Sony is making money on this. They’re not out.

But they’re not making $1 billion on this. But you wrote admiringly of Sony’s decision-making, so what do you think? Why is it ultimately a smart decision on Sony’s part?

I feel really embarrassed because I think I said Matt Belloni in my recording. Is it Belloni?

It is Matt Belloni. I have previously—

I interviewed him on Stuck at Home.

—screwed up his name.

So I’m pretty sure he’s told me it before. Man, that’s a real bad one by me. Sorry, Matt.

I made the exact same—

Great article.

—mistake and felt like a complete idiot about 9 months ago.

Uh-huh.

One of our mutual friends made fun of us in a group chat.

Super embarrassing.

So.

Yeah.

The water’s warm. Don’t worry.

Really great article exploring this tension, right? Why did Sony give away not just the profits from this movie, but the future one? Netflix has rights to it. Sony can renegotiate the deals for the next ones, and Sony can say, “We want more of a share,” or whatever it might be, but it’s Netflix’s movie.

Mm-hmm.

Netflix is going to own, I believe, the merchandising rights, things along those lines. There are some debates about what exactly is owned by whom and what, but this is massive for Netflix. They want franchises. Why are franchises so valuable? Because people flock to them without knowing what they are. They’re basically embedded free marketing.

Yeah.

They want to build up a game. Guess what KPop Demon Hunters do? They fight demons. Sounds like a good game.

Yeah.

They’re going to have this smash-hit game around this. It’s guaranteed.

Mm-hmm.

To Netflix’s credit, I’ve been super skeptical about the whole gaming thing. It seems like a waste of time. Well, they might have the last laugh because this is going to be massive, right? It’s going to be all over the place. There are going to be sequels. Are they going to open a KPop theme park? I don’t know.

It changes everything for Netflix. It’s a huge deal. It’s so easy—and I’m not saying that Belloni was doing this. He was pretty clear about what went into Sony’s decision-making and stuff—but it’s really easy to backward-mirror-drive and say, “Wow, Sony screwed that one up.”

Yeah.

The problem is, I don’t think they screwed up. They talk about how Sony could have put it on their streaming service. They had it. Guess what would go into that? You’d have to build the streaming service. You’d have to try to acquire customers. You’d have to watch all the customers churn. You’d have to be losing billions and billions of dollars along the way, doing things you’re actually not good at, and ultimately losing your shirt and selling to Netflix anyway, which is going to happen to everybody else except for maybe Disney, in my estimation.

And, oh, by the way, you got a hit along the way. Is that hit going to equal out all of what you did? Maybe. Maybe KPop Demon Hunters is going to be so large that it would be worth it.

Mm.

But as a Sony executive who decided, “We’re not going to do that. We are a studio that makes content. We’re going to make content, and we’re going to make deals, and we’re going to make a deal with Netflix,” you also had the COVID bit to consider, which Bethany touches on. Sony didn’t have to do layoffs because Netflix swooped in and bought all this stuff that was supposed to be for theaters, and so they kept everything together.

They have an amazing animation studio that does the Spider-Man movies, for example, which are amazing. It’s the best animation studio sort of going right now, in my estimation. There’s also the desire to assume every decision should have been made perfectly. But decisions are part of a systematic understanding of where you are and what you’re doing.

Yeah.

Sony chose, “We’re not going to be the next Disney. We’re not going to be the next Netflix. We’re going to do what we do well.” Implicit in that choice was that they reduced the risk. They don’t have the risk of a failing streaming service. They don’t have the risk of layoffs during COVID. They don’t have all these risks that they eschewed.

Mm-hmm.

But whenever you limit downside risk, you limit upside risk, and this couldn’t be a better example of it. And you know what? That’s a choice they made, and I think it was the right choice. You don’t get to look backward and say, “Well, you should have known that a movie about K-pop stars fighting demons was going to be worth $1 billion, and the franchise was going to be worth billions more.”

Not a fallen K-pop group, though.

Yeah.

The No. 1 K-pop group.

Yeah. No, if you could rewrite history, would Sony like to have it? Of course they would. But that raises the second question: Would KPop Demon Hunters be KPop Demon Hunters without Netflix?

That’s the most interesting aspect of the story. Can I just confess to you? The synopsis that I read earlier wasn’t from some website. It was from ChatGPT, and—

I knew it. I just wanted you to admit it. It was so clearly ChatGPT language.

It’s horrible. It’s horrible, and I had flashbacks to the mistake I made on the GOAT, like, 2 years ago, trusting ChatGPT. It will never happen again, I promise. I should have looked at an actual—

No, it was totally appropriate this way.

KPop Demon Hunters synopsis.

So it makes it that much clearer.

In any event, that’s the aspect of the story that’s so fascinating because obviously you can look at the world domination that KPop Demon Hunters now enjoys, and there are going to be sequels. There are going to be soundtracks. My friend Ed is besieged with the music all day long. I wonder whether any of this would be happening if not for Netflix as the massive funnel pushing this to the entire world, having invested $100 million in it.

No, this is super important. Read this Binary email because it actually ties into my perspective on this point.

Perfect. Okay, so Binary, as we’re talking about the entertainment business here.

“Hi, guys. Here’s a question for the party in the back portion of the episode. What are the economic and technological reasons that modern American movies are so derivative? Everything is a sequel or a remake of a popular movie from decades ago—i.e., 2024’s Road House—or a franchise: Marvel, Star Wars, et cetera. People have been bemoaning this for a decade now. Ross Douthat says that American society is now decadent, so we’re not producing new creative works. Stratechery is a business and tech publication, though, so let’s put the cultural arguments aside. From an economic perspective, what is going on here?” What do you think?

Well, my favorite topic: newspapers.

Mm.

Why were newspapers, to use Warren Buffett’s term, franchises and not businesses? Because it didn’t matter what chucklehead you put in charge; they were going to make a lot of money. Why were they going to make a lot of money? Because they were the default. Where else were you going to get something delivered to your doorstep with the news, the comics, the sports, all the sorts of things? There were classifieds in there. There were all these things. Of course advertisers were going to go there because they wanted to reach customers. It was a beautiful business.

Then what happened? The internet.

Mm.

You could access newspapers everywhere. At first, “This is amazing. We can reach everyone. This is so great.” It turns out everyone can reach everyone. You have to actually acquire customers. So most of them became beholden to Google: “Hopefully people search and find us.”

The ones that succeeded, like The New York Times, became destination sites. People didn’t go through Google and follow an article. They typed NYTimes.com in their URL bar.

Yeah.

Download the app or whatever it might be. And that's a great business. It's a business, not a franchise, because it depends on excellent management. It depends on maintaining your position. You're trying to influence the choices of all these people to decide to go to you in the midst of a sea of options.

Mm-hmm.

It's not that you move into a house and think, "Of course, I have to sign up for the newspaper." That doesn't exist anymore. It's a constant pursuit of acquiring customers.

And this is amazing. It's 2025. The extent to which I've seen some discussions about something and people still don't appreciate that it's not about the technology—it's about customer acquisition. When you're the default, it's super easy, and when you have to compete, it's really hard.

Mm-hmm.

What did we use to do in the '90s?

Just go to a movie theater.

Friday night.

See what's playing.

Let's go. Did we even know what movies were in the theater? No. We just went to the movie theater and thought, "Oh, I'll see that one." We bought a ticket. It was general admission. You didn't even choose your seat. One of the greatest innovations of all time.

Technically speaking, what we did in the '90s was open up a newspaper and see what movies were playing, and then go to the theater on a Friday night, no matter what was in there. Because that—

No, but some people would just go to the theater. But regardless, that was the default entertainment option.

Yeah.

Life was actually pretty boring in many respects. We didn't have a lot going on in the '90s. Maybe you don't remember. You might be too young.

No, I do. It was great, though. That was something to do with my dad: go out to a theater on Friday night and see whatever was playing. I just remember fondly looking through the newspapers to look at the listings, and then Fandango disrupted that business. But yeah, if you were in a theater, there was going to be an audience.

That's right. And if you want to make an original movie, why were there so many independent films and all these new sorts of movies and things in the '80s and '90s? The hard part was getting the movie made, right?

Mm-hmm.

You had to get funding for that. You had auteurs pitching a great movie, or what sounded like a great idea, and they would make it. And just by virtue of being made, that was the hard part. Now it was in the theater, and people would go because it was in the theater. The poster looked cool. You had an interesting star. "I'll go see that."

The hurdle was getting approved to make the movie. You weren't appealing to the general public; you were appealing to a very small number of people in Hollywood who made those decisions and, by and large, had good taste.

Mm-hmm.

They'd say, "That looks interesting. Go and make it." They didn't have to really worry about the economic outcome. Of course they wanted to make money, but by virtue of being made, most movies would make money. That was the real hurdle.

It's like back in the day with software: you had to get on the shelf at Circuit City. That was the hard part.

Yeah.

Right? The internet changed all this. It took away the gatekeeper function, and we celebrate this by and large. Of course, who likes gatekeepers?

I actually wrote an article about the Harvey Weinstein sort of thing and one of those gatekeepers—

Pro-Weinstein, huh? Real zag.

Well, no. But he was one of those people who chose movies.

Yeah.

And he exacted certain favors from people who were in the movies.

Mm-hmm.

Obviously objectionable. The point I was making in that article was: Why was he in a position to do that? Because the structure of the movie industry relied on those gatekeepers, and he was one of the gatekeepers.

Yep.

Today, to be a star on YouTube, we're not going to have that problem because anyone—there is no gatekeeper. Anyone can go on there and do it.

And the reality of movies today is that they exist in a world with YouTube. They exist in a world with TikTok. They exist in a world of Sharp Tech, of a gazillion things to do to occupy your time and to listen to, which means movies have become a destination site.

Mm-hmm.

People have to decide to go see a movie. And what's the best and easiest way to get people to decide to go see a movie? Make a sequel to a movie they've already seen before.

Something they know. Yeah, exactly.

That's right. And so, if you do original content, you have to spend basically—if you spend $200 billion, you spend $200 billion in marketing if you want to make money.

Mm-hmm.

That's the price. You're going to get movies that have to be widely applicable. You're not going to get niche movies, because you have to do it at scale. And that's the fundamental issue that's happening here, and this is how it ties back to KPop Demon Hunters.

KPop Demon Hunters is not getting the audience it got in this world. Who is making a movie about a fallen K-pop girl group that secretly doubles as demon hunters? Andrew, what are you doing Friday night? You want to go on a date?

Well, I've heard a lot about KPop Demon Hunters because—

No, no, no.

It's the most popular film in Netflix history.

No. Exactly. You've heard about it because it was on Netflix.

Yeah.

And on Netflix, everyone's already subscribed to Netflix. It's just there. So some number of people in the US watched it, then they told their friends. They watched it, then it got on social media, then people started doing covers.

I've watched way too many videos of—there's a whole thing on YouTube of voice actors or voice experts breaking down songs or whatever.

Oh, boy.

It's actually pretty funny with this one, particularly the song "Golden," because every single one has the exact same reaction: "How the fuck is she singing this song?" It seems completely impossible. The repetition actually kind of amuses me.

This was so interesting. I mentioned that straight-line chart. Every other movie on Netflix that has had huge numbers has shot up and then leveled out.

Mm-hmm.

KPop Demon Hunters did not shoot up. It grew very gradually. It was a very steady, linear increase that hasn't halted. The slope of the line is actually totally different from every other big movie. And the other big movies had big stars.

Yeah.

You're on Netflix and it's like, "Oh, there's a Leo DiCaprio in the Netflix movie. Wow, that's cool. I'll watch that."

Big stars, or released over a holiday weekend, where there's a big spike at the beginning—

Right.

—and then it kind of flattens from there. It's true. I didn't notice that about the chart.

Yeah. No, this chart shows this was pure organic growth. Not to brag, it kind of looks like how ShashiKri grew—

Mm-hmm.

—where it was just like—it was... People expect this sort of exponential story. You don't grow exponentially, because the problem with word of mouth is, yeah, one person can tell many people, but people run out of friends. They're in friend groups, right?

The chart of Stratechery's growth was just a perfectly straight line, with very, very little wavering. You don't get exponential growth, but you do get linear growth through word of mouth.

Yeah.

When I see that chart, I'm like, "I know that." That is pure word of mouth, and word of mouth in this case only happens because everyone already has Netflix. The risk of me watching this movie is just my time on a Friday night when I'm watching a movie anyway. I might as well watch this one.

That's totally different from having to go buy a ticket, drive to the movie theater, get popcorn, and stand in line. That's a commitment. That's a destination. K-pop Demon Girls was not a destination.

The great irony here is that all these studios wanted to take on Netflix and thought they could succeed because, "We have the content. Netflix doesn't have content." And the reality is they're actually all reduced to being IP factories, just sort of regurgitating content. The place to create original content ended up being Netflix, because the distribution gives the space for greatness—

Risk-taking.

—excellence—

Yeah.

—to emerge.

Original work.

Yeah.

Original work, as opposed to lowest-common-denominator stuff that you can count on to draw a baseline audience to the theater in 2025.

The counter here is: Why does so much content on Netflix suck if they have this? That's actually part of the answer. It's the opposite of Sony. You throw a bunch of crap against the wall; some of it will stick.

Mm-hmm.

And so the downside is higher, but the upside is higher.

And by the way, that's how the movie business worked for about 75 years.

It was not that no movies lost money, but the hits hit high enough that the upside covered all the losses on everything else, and there were people in movie theaters every single weekend. Now there’s far less risk tolerance in that industry than there used to be because there aren’t as many monster hits beyond a handful of blockbusters, which themselves cost $500 million to produce and market and everything else. So the dynamics of the business are just broken. They’re not—I would say the outlook is a little bit better than it seemed to be four or five years ago, because you can still succeed with IMAX movies. But in general—

Right. But IMAX is leaning into the destination aspect, right? You’re going there because it’s the only place you can get the big screen and all the speakers and all that sort of thing.

It’s a concert experience as opposed to—

Right. Exactly.

—a movie experience. Yeah.

Which is totally rational, and that makes sense that that is what is remaining because you have to deliver something people can’t get at home.

But it’s interesting, too, that KPop Demon Hunters, I mean, it was a hit in theaters the past week or so, and so it—

Because all those people going had already seen the film.

That’s what I— It created the event community experience—

Yes.

—that we all yearned for.

No, this is your concert point, right? Who goes to a concert and hasn’t heard any of the songs? The whole point is you’ve heard all the songs, and now you want to have the communal event and seeing it live. That’s what happened this weekend.

Yeah. Well, and we talked about Netflix as the last bastion of monoculture six to 12 months ago.

This is a killer example of that.

Yeah. Keeping it alive, one K-pop movie at a time. Maybe I’ll check it out. Probably not. Not going to make any promises to the audience. As we close out here—

No, no, you should watch it. It’s good. Seriously.

Yeah, for about 20 years or so, nerds have been telling me how good animated movies are, and they never are actually that good. But I’ve heard the music is catchy. I don’t know. Maybe I’ll get it stuck in my head. One question as we close out here.

Yeah, as we close out, one hour and 20 minutes in, so I’m just going to let that comment fly.

Fair enough. Sydney says, “I know this was a dithering digression, but I thought there was a better chance of you expounding on the virtues of dual monitors on Sharp Tech, so here we are. I’ve long thought that ultrawides and dual side-by-side horizontal monitors don’t make any sense. I pine for the return of 16-by-10, but I won’t hold my breath.”

Man, speaking of nerds. “Anyway, here’s my take,” Sydney says. “Why spend all day shaking your head no when you can add some subconscious positivity by saying yes to your productivity working vertically? Everyone who sees my desk”—this was easier a decade ago when I worked in an office, but here we are—“thought I was nuts, but I feel validated, vindicated even, to know there’s someone out there who’s taken it even further and cut a giant hole in his desk to sink the second monitor even lower. You inspired me to lower my lower monitor further, and I’m happy to report it’s just high enough off my desk to permit my keyboard cable to run underneath. The monitor is angled at a comfortable 38-degree angle, as measured by my iPhone measuring app.”

Ben, do you want to tell people more about your monitor adventures that you did tease on Dithering with, like, a 60-second digression?

So I figured this out with the LG 21-inch UltraFine. Turning your head right and left is uncomfortable and unnatural, and stuff’s out of your field of vision.

Mm-hmm.

I like multiple monitors. You have reference on one and typing on the other. I also don’t like one big monitor. If you have two monitors, you have four quadrants. You can have your writing on one set, a browser on the other. You could have ChatGPT in another, or whatever it might be, right? It’s all sort of in your point of view. It’s so much better to have it all in front of you.

The problem is, it’s really bad for your neck and uncomfortable over time to be looking up. I love those 21-inch LG monitors because they were small enough that I could have them on top of each other comfortably. They went bad because they’re crappy monitors in many respects, so I bought new ones. They were 24-inch.

Hmm.

You’d think that would be good. It was bad. So what I had to do was slope the bottom one at a really severe angle and then have the other one on top of it, and it mostly worked. The slope was a little too much. It would collect dust and, you know, my keyboard would just sort of cover the front—

So this is what you were doing with your week off. You—

In Taiwan. No, no, no. This is what I was doing in Taiwan. I had the two 24-inch monitors on top of each other, with the bottom one massively sloped and then the other one sort of on top of it, so it could be ergonomically accessible or acceptable, but I could still have my view.

Mm-hmm.

However, I knew coming back to the U.S.—I’d been planning this for ages. Because the bottom monitor—

What do you mean about this?

—is at an angle, you think about where your eyes are. Your eyes go past your keyboard, and you could actually have the monitor down inside, lower than the desk, and still be able to see it. So that’s what I did. I cut a big rectangle in my desk, and now the monitor is at an angle, sunk down into the desk. If I were to look at it from desktop level, I’m not going to see the bottom inch and a half of the monitor because it’s under the desk.

Mm-hmm.

I’m not sitting at desktop level. My eyes are a foot and a half above the desk. I’m looking down at an angle, so I can perfectly see my dock, which I put at the bottom of the screen, and I can have the whole thing. By the way, when you have them stacked with macOS, you have the menu bar all the way at the top one. It puts the dock bar all the way at the bottom one. It’s actually a great setup. I have my four quadrants. It’s amazing.

The main problem is that I have a standing desk. I measured this all out. All the support is on the sides. There are some bars that go across the middle, and I cut the hole just in front of that. However, the mount itself is what’s actually hitting the desk. This could go further down. So what I need to do is cut out a part in the middle that lets the support bars still run underneath. There’s actually more to cut out, and then I could sink it even further.

I will do this when I reach my ultimate plan of getting two Apple Studio Displays, because—the reason I did have a Studio Display in Wisconsin was that I used it when I was here, because my full setup was in Taiwan. Now that I’m back, I’m using this setup because it is better than a single Studio Display. The quality’s definitely much better on the Studio Display.

Oh, boy.

So the ultimate dream is to get two of those stacked. I’m not sure if I can pull it off with the size of the hole that I have, but yeah, life’s great. My monitors are so low. I’m so ergonomically comfortable. I have a better view than I did before.

Did you consider the idea that you’re encouraging subconscious positivity in yourself by nodding all day, as opposed to shaking your head no, as Sydney lays out there?

I can’t say I ever considered that. I do still have monitors to the side, but I don’t like working on them. I have my messaging apps there. Right now, out of the corner of my eye, I have our Audio Hijack. I can see the levels and things like that—

Okay.

—for our recording. So it’s good for status stuff, but all the work I do is just in front of me. I really don’t like turning my head side to side.

There you go. Well, my dream is for you to film whatever carpentry you undertake in terms of realizing your Studio Display dream. So maybe we can do a special YouTube video with you—

—underneath the desk, drill in hand—

No, no, no.

—peppering in some profanity.

I’m not drilling in my office. I was drilling in the garage.

Okay.

I’m not drilling anything.

Well, let’s do it. All I want is that on film.

Also, I want to read this one note from Spike. He says, “I retweeted one tweet from some tech magazine about Ben living in the United States, and now half of my For You page is founder bullshit.” My Twitter has been full of founder bullshit for two years as I host this podcast. I’m glad Spike gets to enjoy that as well. I enjoyed the TBPN shout-out to you draped in an American flag as you make your return to the States. So all is right in the world, you know what I mean?

All is right. It’s interesting being here. It’s interesting when you change anything, like all the bits and pieces in your life that were structured around certain realities that sort of get undone, right? I do think it’s generally useful. You’re forced to reconsider things from first principles on all sorts of things.

They’re first principles. I bet you see that in your tech founder bullshit—

Oh, boy.

Twitter feed.

A lot of learnings and first principles, man. It’s great.

Yeah. But, yeah, it’s weird. I’ve obviously been back in the US every summer, but when I was here in the summer previously, I was just trying to limp along until I got back to Taiwan and my usual routine.

And you’re doing a lot of vacation-y things.

Right.

It’s not like I live here and I’m structuring my life around work.

Yeah. It’s been really interesting. It was really good that I took the time. Sorry we didn’t warn anyone I was taking time off. I got to Monday, and I’m like, “This is reaching a crisis point. This whole—”

You have half a breakdown.

I was writing while going back to Taiwan and cleaning and getting stuff packed up and all that. This has obviously been a multi-month process, particularly intense over the summer. And like I said, I didn’t actually have a desk set up until I took some time off last week.

So my apologies for no warning, but you’re welcome for continuing to deliver content to you while this is going on. I’m here, and we’re recording at a normal time. And I think we went long.

Yeah, we went a little bit long, but we’re going to be back next week. In the meantime, you have to get out there and enjoy your first Labor Day in 23 years as a full-time American resident.

No, I was back for 4 years. You forget. I went to business school and worked at Microsoft, so, yeah.

Wow.

But it was 12 years continuously for this stretch, so it’s a big change. But, yeah, it’s good to be here.

Nothing better than a holiday weekend to celebrate. We will be back next week. Email at sharptech.fm if you have questions. That’s the address. Ben, enjoy the next several days, and I will talk to you on the other side.

Talk to you next week.