Kevin Xu started out in politics. He worked on Barack Obama’s first presidential campaign, then in the White House and at the Commerce Department. He moved into technology as GitHub’s lead for international expansion strategy, during the years when the platform became where the world’s programmers built software together, Chinese programmers included. Then he became an investor. He is now founder and chief investment officer of Interconnected Capital, a global fund investing in the hardware and software infrastructure of the AI economy. He also writes Interconnected, a bilingual newsletter on technology, investing, geopolitics and US–Asia relations. He calls the common thread restlessness: he likes starting over in a field he knows nothing about, and he prefers the overlap between disciplines to the disciplines themselves.
His work cuts against the usual story about China and technology. The common assumption is that Beijing chose open source as strategy, to commoditize the model layer and undercut America’s frontier labs. His essay Chinese Open Source: A Definitive History tells a messier story, which we discuss at length in the episode:
The mid-1990s: Linux reached China on floppy disks carried back from Helsinki.
The 2000s: Alibaba and its peers adopted open source to escape the licence fees they were paying IBM, Oracle and EMC.
An uneasy truce: the state tolerated open source with reservations. GitHub was once blocked for a day, and the country’s best engineers simply stopped working.
After the Huawei export controls: open source became a matter of survival, and only then did the state embrace it.
The 14th Five-Year Plan’s call for globally recognized open-source projects came some twenty years after the first floppy disk crossed the border. The government, he argues, arrived late and gets too much credit for foresight.
In this conversation, Kevin explains why the Chinese government didn’t plan for the importance of open source but rather embraced it. He explains that open-weight models open only part of the AI stack. They release the weights but not the training code or data. That is still enough to free a model from the hardware and jurisdiction it runs on, which is why Nvidia wants more of them. He sorts China’s AI landscape into three camps:
neolabs such as DeepSeek, Kimi, and MiniMax;
open-weight labs inside Alibaba, Tencent, and Xiaomi;
ByteDance, closed-source all along, which shows that openness is not a state mandate.
On the money side, he reads America’s data center moratoriums as pauses rather than bans. He traces the build-out’s shift from corporate cash to debt, the turn that took Google’s free cash flow negative for the first time. He expects the model layer to settle into something like the iOS–Android split in smartphones. He names China’s underrated weakness as shallow capital markets, not chips, which is why its AI firms still look to the dollar when they want to grow. And he describes a country that is climate-change-pilled rather than AGI-pilled and may win on clean power what it cannot win on compute.
His case throughout is that the US, China, and Europe are driven more by the same commercial incentives than by the grand strategies attributed to their governments. He also argues that China, with more than 100 million factory workers and tens of millions of delivery drivers, will be the first to face the social cost of the automation it is racing to deploy.
Kevin writes Interconnected and is on X at @kevinsxu.
Marieke: Welcome to the Currency of Power podcast, where we explore how money, technology and geopolitics are reshaping the global order. We speak with the people who are building, regulating and challenging the new financial infrastructure, from currencies and stablecoins to AI, trade, energy and the competition for technological power. Our goal is to understand how the financial system is changing and who gains influence as it does.
I’m Marieke Flament.
Nicolas: And I’m Nicolas Colin.
Marieke: And today we talk about AI, open source, and what’s at stake between China, Europe and the US. And we have the perfect guest to talk about this: Kevin Xu. Super nice to have you here. Thank you so much for joining us.
Kevin: Thank you so much for having me.
“I love looking at the Venn diagram of things.”
Marieke: A few months ago, I read your piece on Substack, Chinese Open Source: A Definitive History, and I found it absolutely amazing. Highly recommended; we’ll have it in the show notes, and we’re going to talk about it. Then more recently, I saw posts from you saying that you were open to recording podcasts, so I just jumped on it and reached out. It’s really amazing to have you here.
You’re the founder and CIO of Interconnected Capital, a global technology fund investing in the hardware and software infrastructure of the AI economy. You also write Interconnected, a bilingual newsletter on technology, investing, geopolitics and US-Asia relations. What I think is really unique, and I’m sure this will come through as we discuss, is your perspective, which is very cross-disciplinary. You started your journey in politics. You worked on Barack Obama’s first campaign, and later in the White House and the Commerce Department. You then became a technology operator, including as GitHub’s lead for international expansion strategy, before becoming an investor and a writer. I think that’s an incredible background for sharing a lot of thinking about what’s happening right now, because you’re at the intersection of AI, open source, chips, industrial policy and the US-China relationship. Your work is cited and published across many, many outlets. I won’t name them all here, but whatever important outlet you can think of, I think you’ve been quoted in it or helped write for it. So that’s awesome, and we are really super happy to have you with us today at Currency of Power. Thank you.
Kevin: Of course, happy to be here.
Marieke: Maybe let’s start with your journey, actually. Tell us about your journey from Obama to GitHub, and now investor and writer. What’s been the common thread?
Kevin: I think the common thread, on a personal level, is just a healthy level of restlessness and easy boredom, in a lot of ways. With a lot of these disciplines, you can dive in very, very deep. But if you put your head up above the water, sometimes you see other things on the horizon, other things that are more interesting. From a personal perspective, I love to start over. I enjoy knowing nothing and jumping into a new field. You learn a lot very quickly that way. It can feel very scary from time to time, but ultimately, at least so far, it’s all turned out pretty well for me.
The other thing, which is probably more of a personal preference, is that I love looking at the Venn diagram of things as opposed to the circles themselves. The circle itself is fine. You can dig very deep and become a longtime domain expert in a particular discipline. But I do think the intersection, the interconnection, the overlap between circles is just as important in terms of informing what’s going on in the future. I find that underexplored, and I personally just enjoy looking at those things anyway. That turned out to be more or less what I do every single day, whether on the writing side or on the investing side.
Nicolas: Does your writing inform your investing, or is it the other way around?
Kevin: It kind of depends on the day. It’s very hard to actually tie those two things together. I would say one thing: I started writing before I started investing professionally. One common thread among almost all the investors we’ve come to know and respect, whether it’s the Warren Buffetts of the world, Stan Druckenmiller, a lot of fantastic European investors, Chinese investors as well, is that they’re all very good writers. They write differently, they don’t all write the same, but they are all good, compelling writers in their own way. Writing, to me, is still a very important discipline for thinking clearly, and I don’t think any investor can be good for a long time without being able to think clearly. He or she can be good for a year or two, a quarter here and there. But to sustain yourself over the long term, you need a very clear process to flush out the silly stuff in your brain, really distill the important stuff, and then hopefully act on it in a way that makes yourself and your investors some money. So I saw that natural connection, and that led me to do both. But as far as whether today’s writing informs my investing decisions tomorrow? Honestly, not really.
Marieke: I think that’s really music to our ears, because Nicolas and I have been writing. Nicolas, you’ve been writing for over a decade. But it’s true that you can only express something clearly, and therefore do something with what you’ve thought about, if you’re able to write it down.
Should we move on to open source and your essay? Do you want to start, Nicolas?
Nicolas: Yes. I wanted to start with a very, very basic question, because we have to assume that part of our audience is not familiar with the world of software development and what open source is exactly. Can you explain what open source is about, the origin of the open source movement, how open source differs from free software and obviously from proprietary software, and why it’s been so successful in being deployed all over the world and powering, or supporting, most of the software applications we use on a day-to-day basis?
Kevin: Absolutely. Open source, in a nutshell, is actually not a kind of technology. It’s more a mode of building technology. You can literally visualize a developer sitting in Brazil logging onto the internet and another programmer sitting in France logging onto the internet, and they’re all building this thing called, say, Linux, which is the largest open-source operating system in the world right now and originated on the campus of the University of Helsinki in the 90s. And these things persist.
Now, why did this happen? Originally, it was a response to all the closed-source, proprietary software that dominated our world in the 80s and 90s. You can think of it literally as the CD we used to buy and put into our computer to get software or a floppy disk. Then it became something you could download from the internet, but you had to pay for a copy. It could be very expensive, which made a lot of tech companies, in the US in particular, a lot of money selling what we call closed-source or proprietary software.
The pushback started in academia first. I believe it was initially on the campus of MIT, where people said, “Why do I have to pay a company to update my printer software to make it print better? This situation is ridiculous. I’m a computer science professor. I can write this software myself. I’m going to write a version of it and put it out into the world for everybody to use.” That’s the free software era of open source. Free software actually predates open source slightly.
But then people said, OK, free is good, but how do we change it? How do we improve it? What’s the mechanism through which this software can evolve? Because software has to evolve almost in perpetuity. If you buy a machine, a laptop, or a printer, it just sits there until you want to buy a new one. But the software is what runs underneath. That evolved into the organic global movement we now call open source, where everybody has the right, the privilege, and the access to download, customize, redistribute, and actually sell this piece of free software that is just sitting on the internet, literally. That eventually became the dominant way we all use software, whether we know it or not. Everything you use on a daily basis has at least one or two, probably fifteen, pieces of open source software sitting inside, powering it.
Nicolas: Can you say a few more words about the legal side of things? Every open source component has a license attached to it that protects its openness in perpetuity. You can’t take something that’s open source and turn it into closed-source or proprietary software, and that’s because of the license, if I’m not mistaken.
Kevin: That is mostly correct. I won’t bore your audience too much with the legal side of open source, but there is actually a huge distribution of licenses governing different open source software, and you, as the creator, can choose what you want your license to be depending on what you want out of it. There are many licenses, and this will be very relevant when we talk about Chinese open source; that is what we call permissive, meaning you can download the software, change something, turn it into closed-source software, and sell it for money. That’s totally OK if it’s licensed under these permissive licenses. Very, very hands-off. Then you have other licenses that are much more restrictive, in the sense that you cannot do what I just described. Or you cannot resell the software on a cloud platform like AWS or Azure, for example, but you can do whatever you want with it in a nonprofit or academic setting. That variety is actually something open-source people fight about a lot internally as well. What is really open source? What is not open source enough? I won’t get into the nitty-gritty of the cultural challenges with legal licensing, but it is very much a spectrum, and it’s a live issue in the industry.
Nicolas: Super interesting. Maybe one last question on open source. My eyes were opened to open source, and I understood most of it, when I read the landmark article by Tim O’Reilly published in 2004, called Open Source Paradigm Shift, which didn’t really appear in your essay on Chinese open source. Does O’Reilly in particular, or the whole community connected to him, have some influence or some audience in China, or are those completely separate?
Kevin: I think the influence is completely there. I chose to under-represent, if you will, some of the typical Western voices in open source, because I think the story of what happened in China is fascinating in its own right. But the undercurrent of how China even started doing open source to begin with all came from the West. In fact, even today, a lot of the norms, industry standards, and best practices in open source are still mostly learned from the West. Even these licenses we talked about: DeepSeek R1, which became the “DeepSeek moment” a little less than two years ago and shocked the world. One of the reasons it became such a shock is that it was licensed under one of these super permissive licenses where you can do whatever you want, and the name of the license is literally the MIT License. It came from the MIT campus.
What’s interesting about open source is this. When we think about the geopolitics of AI, or tech in general, borders matter a lot. One country versus another, this country doing its own thing, the other country imposing export controls. That’s all fine and good. But when it comes to open source, the permeation of information and know-how is seamless across borders. It’s almost borderless by design. So to your point, they’ve consumed everything Tim O’Reilly has written, everything all the open-source leaders have written, and then made it their own and adapted it to their own country’s conditions, situation, and needs. That’s how their trajectory became a little bit different.
“It became a matter of survival.”
Marieke: Maybe this is a good turning point to deep dive into your essay, Chinese Open Source: A Definitive History At first glance, there seems to be a contradiction. How would a centralized, more controlling state embrace open source? What’s fascinating in your essay is the story of why China actually has a very long history with open source. So maybe let’s start with the history of China and open source, if you can walk us through it.
Kevin: Sure. The one top-line takeaway I hope all your listeners take away, before I get into the stories, is that this is an example where the Chinese government came very late into the picture. Open source, as the form of technology development I described, more or less existed organically in China until, essentially, the US export controls that put Huawei on the blacklist. We’ll explain the logic of that. That’s the big takeaway, and I think it’s very different from what a lot of people think about China: that the government has big strategic thoughts, tells everybody what to do, everybody just does it, and then you get a lot of EVs on the streets, batteries in your home, appliances, and whatnot. Some of that is true, depending on the sector. But for open source, the government was late to the picture.
Open source entered the Chinese economy organically in the mid-90s, not long after Linux itself was born at the University of Helsinki. In fact, I think the story is that a Chinese researcher happened to be studying at the same university, learned about Linux, carried a copy of it back to China on a bunch of floppy disks, and started playing with it.
What was attractive about open source initially is that at the time, the late 90s and early 2000s, there was already a corps of internet companies that came up in China as part of the dot-com boom, like Alibaba, Tencent, and a bunch of others that no longer exist. They were paying a ransom’s worth of fees to closed-source software companies, particularly from America, just to keep their websites running: IBM, Oracle, and EMC in particular. They saw open source as a way to get rid of Western proprietary software so they could run their own stack more cheaply and also have more control over their own software and IT stack. That’s phase number one.
Then, as that became successful and Alibaba became successful, everybody started throwing their energy toward open source as a way to keep building. The important part is that China hadn’t really contributed back to open source at this time. They were takers of whatever had been produced in the West. They adapted it, but they didn’t contribute back.
This was also the time when GitHub, where I used to work, became a mild global sensation among hackers, because it made open-source collaboration that simple. It used to be very hard. You had to be on some obscure email listserv to coordinate, then download the software, make some changes, upload it to some other source-control thingy, and somebody else had to go in and do something else to make it work. GitHub made it that simple for everybody around the globe, including China. So that’s phase number two, where open-source norms became much more democratized by an American company funded by VC money like everybody else, but with global impact.
Then you move on to a more domestic phase, where China started making its own open-source communities, its own meetups, and its own hacker groups. They started grabbing what was amazing from the West but also assimilating it into their own setting to make it work for themselves.
There was also an interesting period when open source became almost a vehicle for social advocacy, and this is where the nuance between the Chinese government and open source really came in. As we all know and appreciate, the Chinese government does not really appreciate tools that let people organize online very, very easily. That’s why you don’t have Facebook in China. That’s why you don’t have Twitter in China. And for a little while, GitHub was also in the crosshairs because of this possible social dynamic, where people could organize on GitHub, since it’s a social network, in a way, just for programmers and geeks. GitHub was actually shut down for 24 hours, and all hell broke loose in the Chinese tech ecosystem because nobody could work. Your most productive, talented programmers in universities and tech companies were just saying, “OK, there’s nothing to do; GitHub is down.”
That became a tug-of-war, an uneasy relationship, as I call it in my piece, where the Chinese government probably didn’t like some aspects of open source, to be perfectly honest. But it’s also such an important engine for the country’s most talented people to build technology that you need to not be so, I don’t know, authoritarian about it, if you will. So that’s the first half of Chinese open source. I’ll pause here to see if there are any follow-up questions before I keep rambling on to the other chapters.
source that the
Kevin: There have been attempts at a split. What ended up happening, in response to these elements of open source the government may not appreciate, is that a local alternative to GitHub started being propped up. It’s called Gitee, G-I-T-E-E. You can think of it as China’s own GitHub, which is all fine and good. And then what ends up happening is exactly what you said. Gitee is used only by Chinese programmers. Does that work? OK, fine, you can upload some code onto Gitee and do open-source collaboration with only Chinese programmers. But what does that do to help your software advance when you could get the best eyes and the best talent from around the world to look at your software, file issues, improve it, give you use cases you’ve never thought of, give you better ideas, and interact with them all seamlessly on GitHub? The magic of open source is that it’s global by default. As soon as you nationalize it, that part is gone, and it’s just another software development tool without much to it.
So that’s still a tug-of-war. It’s not a resolved question, but China is still attempting to have alternatives to the Western platforms that host important open-source software. We have GitHub, and now we have Hugging Face, which is the platform that hosts AI models, datasets, and so on. China has its own version of that too. It’s sponsored by Alibaba and called ModelScope. It’s fine. But it’s just not going to get the global traction that Hugging Face has by default.
So there’s always going to be a little bit of tug-of-war, which is why, again, for audiences who don’t follow this super closely: We hear a lot of narratives saying China loves open source and is going to push open source forever for all these big strategic reasons, whether it’s overcapacity or destroying the US frontier models. What I’m trying to say is that the relationship has always been, and still is, very, very uneasy.
Nicolas: I was about to say that China is a large country with lots of software developers. Can’t you reach critical size with China alone to deliver high-quality open-source software, without relying on developers in the rest of the world?
Kevin: That’s a very interesting question, and it gets into the second half of the history, which is when China started producing its own open-source software. Everything I’ve described up to this point, which takes us to about the mid-2010s, give or take, is China taking, taking, taking. And then came US export controls, in particular during the first Trump administration, which put Huawei in the penalty box. I’m a hockey fan, so I like to use random hockey references here and there. They got put in the penalty box. That meant open source wasn’t just a nice-to-have for building your technology company anymore. It became a matter of survival for your company and possibly your country.
So Huawei, and all the other companies that had been playing with open source longer than Huawei, because Huawei came late to the open source picture as well, all had to throw more energy into producing software using open source. They could have kept it closed for themselves, but they also saw the magic and the power of distributing open source globally. So Huawei started not only making its entire stack by itself, because it couldn’t access Western technology anymore, from operating systems to chips to databases to servers. It also open-sourced all of that through its own open-source foundation to push it out to the rest of the world. One, because you just make better software if more people have eyes on it, period. You could make it into your own little ecosystem, but who’s really going to care if no one else can try it, improve it, raise questions about it, or find bugs in it? That’s a long but very important iterative process that makes open source work the way it does.
Two, and this is where the government does come in, in the early 2020s: The government finally issued a directive from the very top saying we want China to have two to three globally recognized open-source projects, as part of its five-year plan. That was the 14th Five-Year Plan, so not the most recent one, the one right before. So the government came in 20-ish years after Linux first crossed the Chinese border and said, “We want open source; we want to promote open source because it’s such an important part of our technology self-sufficiency story vis-à-vis external pressure. It’s obviously making our companies better too. There are some problems with open source that we don’t like, but net-net, it’s a very good thing, and we now want to lean into it.” And this was all before AI. The AI proliferation that came afterwards, with China dominating the open-source AI scene, made open source even more self-evidently something they want to support and should support, even though there are still reservations about elements of how it works.
