Why AI Hardware Stocks Are Splitting Into Two Markets

0:00Lucas: So it's June 12th, 2026, and AI hardware stocks are sending a really confusing signal. Nvidia is down a bit this week, AMD is up, but the real story is the equipment makers. Applied Materials up 15 percent in five days. Lam Research up almost 14 percent. ASML up 6.5 percent. That's a huge divergence. Luna: Yeah, it feels like the market is starting to distinguish between 'AI training' and 'AI manufacturing.' Nvidia is still the training king, but maybe the growth is shifting elsewhere. Lucas: Exactly. And I think there's a really specific structural reason for this. Which is — the chip industry is hitting a wall with current manufacturing techniques, and the only way to keep packing more transistors is to buy more advanced equipment. That benefits the tool suppliers directly. Luna: Right, and a couple of dollars a month is genuinely what keeps these shows going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. It makes a real difference. Lucas: Absolutely. And speaking of real differences — look at Intel. Up 13 percent this week. That's their highest level in a while, and it's tied directly to their foundry pivot. They're positioning themselves as a manufacturing partner, not just a chip designer. Luna: Which is basically what TSMC has been doing for years. But Intel is trying to do it with US government support and a different technology roadmap. Lucas: Right. And TSMC itself is basically flat this week, down less than one percent. So the market isn't rewarding the incumbent. It's rewarding the companies that are supplying the next generation of manufacturing tools. Applied Materials, Lam Research — they make the deposition and etching equipment that are critical for sub-3-nanometer nodes. Luna: So the thesis is: as chips get more complex, you need more expensive and more numerous tools per wafer. That drives revenue for the equipment guys. Lucas: Precisely. And ASML is the extreme example. They have a monopoly on extreme ultraviolet lithography, which is required for the most advanced nodes. So even if Nvidia's sales growth slows down temporarily, ASML's order book is still filling up because data center operators are building out capacity for inference workloads. Luna: And that's the other piece of the puzzle, right? We've been talking about training forever, but inference is becoming a massive compute load. Every time you query a large language model, you need a GPU or an ASIC to run it. Lucas: Right. And inference is less concentrated on Nvidia. You see companies like AMD, with their MI300 series, and even startups like Cerebras and Groq getting traction. So the equipment makers benefit regardless of who wins the chip design race. Luna: That's a really important point. You don't have to pick the winning chip designer. You can just bet on the manufacturing infrastructure. Lucas: Which is what the market is doing this week. Applied Materials reports strong bookings. Lam Research talks about multi-year growth cycles. And the SOXX index is up 4.5 percent in five days, even though Nvidia is down. Luna: So what's the risk here? If AI spending slows down, equipment orders would be the first to get cut, right? Lucas: That's the classic cyclical risk. But the counterargument is that we're still in the early stages of a buildout that could last another decade. Data centers are being constructed at a pace we've never seen. And each one needs racks of servers, which need chips, which need equipment. Luna: So it's a play on physical infrastructure rather than algorithmic innovation. Lucas: Exactly. And that's why Intel's foundry bet is interesting. They're trying to become a provider of that physical infrastructure. They have a lot of existing fabs, and they're building new ones in Arizona and Ohio. If they can execute, they capture a slice of that equipment spend too. Luna: But Intel has a long history of execution issues. The foundry business is capital-intensive and low-margin until you hit scale. Lucas: No question. But the market is giving them credit for the direction. Their stock popped 13 percent this week after they announced a partnership with a major cloud provider — I think it's Microsoft — to manufacture a custom AI chip. Luna: That's huge. A marquee customer validates the whole strategy. Lucas: It does. And it also shows that the big tech companies want alternatives to TSMC. Geopolitical risk is real. Everyone wants a second source. Luna: So if I'm an investor, how do I think about this? Do I buy the equipment stocks or the foundry stocks? Lucas: I think the simpler play is the equipment makers, because they have pricing power and technological moats. ASML is basically impossible to replicate. Applied Materials and Lam Research have decades of process knowledge. But the foundry play is higher risk, higher reward. Luna: And what about the companies that are actually designing the chips for inference? Like Broadcom, Marvell, those guys. Lucas: Broadcom is down 4 percent this week. Marvell is also down. So the market isn't rewarding custom ASIC designers right now. It's rewarding the people who build the tools to build the chips. Luna: Interesting. So there's a clear rotation happening. Lucas: Very clear. And I think it's sustainable because the equipment cycle is longer than the chip design cycle. A fab takes years to build and equip. Once the orders are placed, they're not easily canceled. Luna: Unless there's a recession. Then everything gets delayed. Lucas: Sure. But even in a downturn, the secular trend toward more compute is hard to ignore. AI models are getting bigger. Inference demand is exploding. And all of that requires physical chips. Luna: So the takeaway from this episode is: watch the equipment makers as a bellwether for the entire AI hardware ecosystem. Lucas: I think that's right. If Applied Materials and Lam Research keep raising guidance, it means the buildout is real. If they start missing, that's a red flag for everyone. Luna: Good stuff. Next week we'll dig into the software side — what's happening with AI model pricing and margins. Lucas: Looking forward to it. For now, if you enjoyed this, buy me a coffee dot com slash fexingo — it genuinely helps us keep the podcast ad-free and independent.