Kyle Harrison
← All podcasts
podcast August 5, 2025

TBPN — Tuesday, August 5th

TBPN · hosted by John Coogan & Jordi Hays

Watch on YouTube ↗

Summary

Kyle joins John Coogan and Jordi Hays on TBPN to unpack Contrary Research’s latest deep dive, Building an American TSMC. The core argument: the US depends on Taiwan for 90%+ of advanced semiconductors, and that is “probably the dumbest bottleneck we could possibly be in” — an existential vulnerability given a plausible China 2027 invasion timeline that Xi has pointed to. The startup ecosystem, however good, can’t reach leading-edge scale fast enough, so Kyle argues Intel — despite “sending up emergency flares” — is America’s only realistic path to sovereign leading-edge capacity.

His through-lines: the “American TSMC” is a mindset and an ecosystem, not a single company; policy should incentivize American companies to buy American-made chips to build up domestic “reps”; Intel Foundry must be treated as a growth engine to invest in, not “an ugly stepchild” (four foundry leaders in three years, a title demoted from president to general manager buried under two other jobs); and the deepest missing ingredient is a founder-mode visionary leader plus a patient investor base — he contrasts Lip-Bu Tan’s cut-after-cut posture with Jensen Huang, and Intel’s “Wall Street weenies” with the extraordinary Tesla investor base Elon Musk has assembled. Companion to Kyle’s essay Chips For America and the Silicon Shield framing that runs through it.

Transcript

Lightly cleaned for readability (speech-to-text artifacts corrected, names normalized). TBPN, Tuesday, August 5th, 2025.

Host: Anyway, we have our next guest in the Temple of Technology — Kyle from Contrary. How you doing, Kyle? Good to see you. Welcome to the stream.

Kyle Harrison: How’s it going, guys?

Host: What’s new in your world? Break it down for us.

Kyle Harrison: We’re just launching our latest deep dive from Contrary Research. For folks who don’t know, Contrary Research is our private-markets research database we launched a couple of years ago. Contrary is a venture fund, but our focus is on talent and on research — finding the best ideas, the best people, and bringing them together. Contrary Research has published research on 500 different companies; I think we’ve got most of your sponsors on the list, so we’re working our way down. We just published our latest deep dive on building an American TSMC. The TL;DR: we’re completely dependent on Taiwan for 90-plus percent of advanced semiconductors, and it’s probably the dumbest bottleneck we could possibly be in. It feels like we’re basically just hoping it doesn’t become a problem — when it could become the most significant geopolitical flashpoint and effectively lead to World War III. So it feels like there’s something we should do about it. We go through the options: Do we build it from scratch? Do we borrow it from TSMC? Do we try to buy it from some willing sellers?

Host: It seems consensus that Intel is not the American TSMC. Is TSMC the American TSMC? Is Samsung the American TSMC? We’ve seen a lot of progress from both — Elon doing a big deal with Samsung, TSMC obviously seeing great results out in Arizona. What’s your take on just porting or expanding the current set of fab companies that are doing quite well?

Kyle Harrison: Our biggest takeaway from the research is that the American TSMC is more of a mindset than a specific company. It’s an ecosystem — we have to build a lot of different aspects of this. There’s a ton of startups building advanced stuff in chip design and using AI, and TSMC is pouring $165 billion into the US, so there’s a lot we can take advantage of. The biggest limitation is that when you try to build a lot of this from scratch, you just don’t get there quickly enough for it to make a difference. There’s this idea — the folks at SemiAnalysis have it internally — of “China 2027”: if China is going to invade Taiwan, it’s probably going to happen over the next four or five years. Xi has talked about 2027 specifically. So this is an urgency thing, not a long-term bet. The question is: how do we get to scale where, if China were to take Taiwan tomorrow, we’d have some capacity here? The unfortunate reality is that the startup ecosystem, as great as it is and as much as it can benefit us in the long run, is not going to get there quick enough. As much benefit as we get from TSMC partnering — a lot of the results they’re having in Arizona — there’s a huge cultural blowback in Taiwan, because they have this “Silicon Shield”: they think of our dependence on them as their protection from China. So yes, TSMC wants to pour a bunch of resources here, but whether we get those leading-edge nodes that let us compete for things like AGI and really compete with China — that’s a much longer battle. Despite the fact that Intel is struggling, it’s basically sending up emergency flares. You guys talked about it on the show — in their last quarterly earnings they said, essentially, “we’re going to do our best, and if we don’t get material customers, it may not be economical for us to continue to compete in leading-edge nodes,” which would be a huge hit for the US staying competitive. So our thesis in the piece is that we basically have to take Intel as our only hope of getting to scale quickly enough. Intel’s not going to do it on their own — there’s a bunch we have to do to make that possible.

Host: What’s the biggest request from TSMC or Taiwan? Should we be pulling the “hey, we’ll invest alongside you, we’ll subsidize, we’ll give you tax cuts to build the second, third, fourth fab in America”? Or should we be more focused on committing more defense resources, more government spending to Taiwan to create an American shield that’s independent of the Silicon Shield? Or maybe both?

Kyle Harrison: There’s definitely — when you talk to all the major defense companies — Palmer Luckey has a great line where he talks about turning Taiwan into a prickly porcupine that people don’t want to step on. That is absolutely priority number one: to give them all the capabilities they need. The folks at Anduril are on the ground in Taiwan all the time, so that’s a huge component of sharpening the partnership. I also think you need to emphasize that the government needs to incentivize American companies to purchase American-made semiconductors as much as possible. As you reshore more of this — whether it’s TSMC’s efforts, or whether we literally nationalize Intel and take the foundry business private — Dylan Patel, who you guys have had on the show and who’s an absolute rockstar, has made it abundantly clear that the survival of Intel Foundry is critical to America’s economic and technological dominance. We have to take that seriously. There are a bunch of incentives we can build around getting people access to semiconductors made by Intel in the US, and by TSMC in the US. We have to incentivize people to use our own ecosystem so we get the reps we need to get as far away from dependence on Taiwan as possible.

Host: Do you think Lip-Bu Tan is the right leader for this transition for Intel? It seems like they need to reinvent themselves with broader support, and yet hearing Lip-Bu Tan on earnings calls just feels like cut after cut. He’s good at firing people, but is he really good at inspiring?

Kyle Harrison: I haven’t been inspired. He’s definitely not Karp on the mic — he’s not exactly inspiring people.

Host: He also has like 1,000 employees at Palantir, and I’m pretty sure Lip-Bu Tan has like 100,000 employees or something.

Kyle Harrison: Intel has more employees than TSMC and Nvidia combined — so he’s got plenty of people he can fire. The biggest thing that’s compelling about Intel is that they’re already in this process of cultural change. Back in 2005 was the first time they had a CEO who wasn’t an engineer by trade, and basically they went through 15 years of financialization — literally the Jack Welch playbook of dumping money into stock buybacks and robbing R&D for decades. Those CEOs put them on the path of stumbling and losing their lead. So even with Gelsinger, that transition was already underway — putting engineers back, “bringing the nerds back.” That’s important. The cultural leadership you should pay more attention to is the leadership level at Intel Foundry specifically. When you think about Intel’s two core businesses — Intel Products and Intel Foundry — Intel Foundry has had four different leaders over the last three years, some lasting no more than two months. What’s compelling is this idea that you can have a within-a-division leader who’s really inspirational. The folks at Waymo have a co-CEO setup — their former head of policy and their former CTO. You have the technology, you have the policy — that’s a great duo for the Waymo setup. Something at Intel Foundry that could be similar could be really compelling. The biggest concern for me: the first person they had once they started breaking out Intel Foundry into a separate P&L — they called him the “president of Intel Foundry,” and he reported directly to Gelsinger. Four leaders later, this person is the “general manager of Intel Foundry” — and they’re also the chief global supply chain officer and the chief technology officer of Intel’s core business. Meanwhile, Intel Products has its own dedicated CEO with the title of CEO. The biggest problem isn’t necessarily even Lip-Bu Tan; it’s more about whether we’re actually acknowledging that Intel Foundry needs to be a growth engine you can invest in to build back dominance — and that’s absolutely not how they treat it.

Host: Earlier this year there was a rumor that Elon was kicking the tires on Intel. How real was that? Did he kick the tire?

Kyle Harrison: It was the jet tracking — that’s how they all track it. But the jet tracking felt a bit silly, because everybody was at Mar-a-Lago. Everyone was there. You could spin a lot of compelling narratives based on all the people who were there at the same time. In terms of what’s actually official, the two things you know for sure: there were talks that TSMC was going to acquire Intel’s foundry business — their board flatly denied that, said it wasn’t true at all. Then there were talks that TSMC might invest heavily and run a joint venture where they take over operations of Intel’s foundry business even if they’re not buying it outright — and what’s funny is they did not deny that outright; they refused to comment. So the general consensus is that that was an active conversation.

Host: How much of the race for American semiconductors is driven by the question of “founder mode” — that abstract question of whether we need Elon in the seat leading? He said he was going to be walking the floor at Samsung. Obviously the founder-mode pattern has seemed to work in so many stagnating worlds — rocketry, electric cars — where Elon’s gotten in and jarred things loose. Is that something we should be looking out for or optimizing for?

Kyle Harrison: 100%. When you contrast Lip-Bu Tan versus Jensen, it’s night and day. You do not have visionary leaders in chip manufacturing in the US, because we don’t do it anymore — it’s not a fundamental part of what we do, there’s no core DNA there. Do you need somebody who is that visionary leader? Absolutely. And does Intel have any of that DNA? No. So when you think about what needs to change, our thesis after six months of research is that you cannot ignore the entire ecosystem — and that visionary-leader role is absolutely critical. That’s an unanswered question; we don’t have that person. Somebody retweeted our post today and said, “we just need an Elon Musk — a herculean act of God to make this possible.” The takeaway is accurate; the how is much more difficult.

Host: Yeah, it can’t be manufactured. But green shoots all over the place, lots of cause for optimism. Hopefully the piece starts a conversation and can be a catalyst for change. What are you hoping to see out of the immediately next quarter for Intel to get on the right track?

Kyle Harrison: The biggest opportunity is for Intel to stop treating Intel Foundry like it’s an ugly stepchild — this embarrassing accident they happen to have, that they’re trying to distract people away from or dress up like it’s something fancy. This goes back to a great conversation I’ve heard a couple of folks have — I think it was Bucco Capital — talking about how the greatest thing Elon Musk has built, independent of even Tesla or SpaceX, is the investor base at Tesla. It’s got to be the most incredible investor base any public company has ever had, to let him do what he’s doing and to have the vision. Intel is just smattered with what Palmer would call “Wall Street weenies” — everybody’s just thinking about how to optimize this thing as effectively as possible. That is not how you build the future, nor how you build American dominance. So you really need that investor-base turnover, and it feels like that comes from narrative: if you’re telling the story of Intel Foundry — we can get a visionary leader, we have this opportunity, they’re already at scale, they have a lot of the technology, they’ve caught back up in many ways (they’re still failing at operations pretty miserably) — the opportunity is to tell that story and get an investor base that’s stoked to be investing in the American TSMC.

Host: Well, thank you so much for joining. It’s a pleasure catching up. We’ll talk to you soon — have a good rest of your day. Thanks for coming on, Kyle. Great work.

Connections

People

  • John Coogan & Jordi Hays — the TBPN hosts.
  • Elon Musk — the founder-mode archetype; Kyle’s contrast case for visionary leadership and the Tesla investor base.
  • Jensen Huang — the counter-example to Lip-Bu Tan: “night and day difference” in visionary chip leadership.
  • Lip-Bu Tan — Intel CEO; Kyle’s “good at firing, not inspiring” critique.
  • Pat Gelsinger — prior Intel CEO who began the “bring the nerds back” engineering-culture turn.
  • Palmer Luckey — “prickly porcupine” line on Taiwan; source of the “Wall Street Weenies” framing.
  • Alex Karp — the “Karp on the mic” benchmark for inspiring public-company leadership.
  • Dylan Patel — SemiAnalysis; cited on Intel Foundry’s survival being critical to US dominance.
  • Jack Welch — the financialization / buyback-over-R&D playbook Kyle blames for Intel’s lost lead.

Companies

  • Contrary / Contrary Research — Kyle’s fund and its private-markets research database; publisher of the American-TSMC deep dive.
  • Intel / Intel Foundry — the piece’s central bet for sovereign leading-edge capacity.
  • TSMC — the incumbent to emulate, borrow from, or buy; $165B US investment; Arizona fabs.
  • Samsung — alternative fab expansion; Musk’s deal and floor-walking.
  • Nvidia — scale benchmark (Intel has more employees than TSMC and Nvidia combined).
  • Anduril — on-the-ground defense partner in Taiwan.
  • Palantir, Tesla, SpaceX — Karp’s / Musk’s companies as leadership + investor-base references.
  • Waymo — the co-CEO (policy + CTO) model Kyle floats for Intel Foundry.
  • SemiAnalysis, Bucco Capital — sources for the China 2027 timeline and the Tesla-investor-base point.

Concepts & writing