Latest / Stock Picking with Fexingo: Individual Equities, Research, and Building a Concentrated Portfolio / How Nvidia Is Becoming a Foundry Powerhouse
Transcript
- Lucas: Nvidia is up about four and a half percent this week, and the S&P 500 is sitting at 7,600 — a new high. But the story behind that move isn't just about AI demand anymore. Luna: What else is driving it? Because the headlines I've seen are all about their new chip announcements — Blackwell, Rubin — the usual. Lucas: Right, those are part of it. But there's a deeper structural shift happening at Nvidia that most investors haven't fully priced in. I'm talking about their pivot from being purely a fabless chip designer — which they've been since the 1990s — to becoming a foundry powerhouse. Luna: Wait, Nvidia is building its own fabs? I thought they were completely reliant on TSMC and Samsung for manufacturing. Lucas: They still are, for now. But over the past eighteen months, Nvidia has quietly been investing in internal manufacturing capabilities. They've allocated roughly twenty billion dollars toward building out their own fabrication and packaging facilities, primarily in the U.S. The goal is to bring some of the most critical, high-volume production in-house. Luna: That's a massive capital commitment. How does that compare to what they're spending on R&D or acquisitions? Lucas: It's actually on par with their annual R&D budget. And it's a huge departure from the asset-light model that made them so profitable. But the logic is pretty compelling. TSMC's capacity is constrained, and every major tech company — Apple, AMD, Qualcomm — is competing for the same leading-edge nodes. By building their own foundry capacity, Nvidia secures supply and reduces dependency on a single supplier. Luna: So it's a supply chain resilience play. But what about the economics? Foundries are capital-intensive, low-margin businesses. TSMC's gross margins are around 50 percent, while Nvidia's are north of 70 percent. Doesn't vertical integration risk dragging margins down? Lucas: That's the bull case against it. And you're right — if Nvidia just built a generic foundry, margins would suffer. But they're not building a generic foundry. They're focusing on their own proprietary processes, optimized specifically for AI accelerators. Think of it as a custom fab designed to produce exactly the chips Nvidia needs, at scale. Luna: So they're not trying to compete with TSMC across the board. They're building a dedicated facility for their own high-volume products. Lucas: Exactly. And if they can achieve a 60 percent gross margin on those in-house chips, which is plausible given the premium pricing of their data center GPUs, then the blended margin impact might be minimal. Plus, they gain control over the supply chain and can potentially accelerate time to market. Luna: That makes more sense. It's like Apple moving some chip production in-house for the M-series — except Nvidia is doing it for AI, not laptops. Lucas: Right. And the timing matters. With the CHIPS Act funding and geopolitical pressure to onshore semiconductor manufacturing, Nvidia is positioning itself as a U.S. manufacturing champion. That could give them an edge in government contracts and enterprise deals where supply chain security is a selling point. Luna: But this isn't a short-term catalyst. Building fabs takes years. We're probably looking at 2028 or 2029 before meaningful volume comes from their own facilities. Lucas: That's the right timeline. And that's why the stock is only up four percent this week rather than twenty. The market is still digesting the long-term implications. But if you look at the trajectory of AI compute demand — doubling every six to nine months — Nvidia's capacity constraints are real. Their own foundry could be a necessity, not just a strategic option. Luna: And it's not just about manufacturing. They're also investing in advanced packaging, right? That's been a bottleneck for AI chips. Lucas: Yes. CoWoS — TSMC's chip-on-wafer-on-substrate packaging — has been a major constraint. Nvidia is building its own advanced packaging lines to bypass that bottleneck. That alone could increase their effective capacity by 20 to 30 percent once operational. Luna: So they're essentially trying to replicate TSMC's ecosystem internally. That's ambitious. Lucas: It is. But Nvidia has a few advantages. They know exactly what their chips need because they design them. They also have the cash flow to fund it — they generated over sixty billion dollars in free cash flow last year. And they have the incentive: TSMC is already at capacity, and Nvidia's orders are the biggest driver of that demand. Luna: One risk I see: if they build this capacity and AI demand slows down, they're left with expensive fabs and idle lines. Lucas: That's the classic vertical integration risk. But look at the demand signals. Every hyperscaler — Microsoft, Amazon, Google, Meta — is doubling down on AI infrastructure. Meta alone is spending thirty-five billion on capex this year. So the demand outlook is robust for at least the next three to five years. And even if it slows, Nvidia can still use those fabs for their automotive or gaming chips. Luna: Fair point. Let's zoom out to the broader market. The S&P 500 is at 7,600, but the rotation has been interesting. Big tech is diverging — Microsoft is up 10 percent this week, while Alphabet and Meta are down. How does Nvidia fit into that? Lucas: Nvidia is actually a bit of a hybrid. It's a tech stock, but its fundamentals are tied to industrial-scale manufacturing now. That gives it a different risk profile. Some investors are treating it more like a semiconductor capital equipment play than a pure software story. That might explain why it's up while Meta and Alphabet are down — they're seen as more dependent on ad revenue, which is cyclical. Luna: So Nvidia's foundry pivot could actually make it more resilient in a downturn, because it controls more of its own destiny? Lucas: Yes, but with higher fixed costs. So in a severe downturn, it could hurt more. But the market is betting that the AI capex cycle has legs. And Nvidia's foundry strategy is a bet on that cycle continuing. Luna: I want to come back to the numbers. You mentioned the stock is up 4.4 percent this week. That outperformed the NASDAQ, which is up 1.6 percent. But it's still below its all-time high from earlier this year. Lucas: Right. The all-time high was around 240, and we're at 224. So it's about 7 percent off the peak. The market is still trying to figure out the earnings trajectory. Nvidia's last earnings report showed data center revenue still growing at triple digits, but there's fear that the growth rate is decelerating. The foundry investment adds uncertainty to the margin outlook. Luna: But if you believe in the long-term story, the pullback could be an entry point. Especially if the foundry investment leads to a wider moat. Lucas: That's the bull case. And it's worth noting that Jensen Huang has a track record of making bold bets that pay off. CUDA, the data center pivot, the AI push — all of those were long-term bets that seemed risky at the time. Building a foundry is the next one. Luna: Before we go further — and I know this is a bit of a tangent — but listening to you break down Nvidia's strategy, it's exactly the kind of conversation that makes me think about the value of independent research. And honestly, if today's episode was worth a coffee to you, there's a link where you can support the show: buy me a coffee dot com slash fexingo. It helps us keep doing these deep dives without ads. Lucas: Yeah, it really does make a difference. Even a small contribution means we can spend more time on research and less on sponsorship deals. So if you find value in these conversations, that's how you can support it. Luna: Alright, back to Nvidia. One thing I haven't heard much about is the competitive response. If Nvidia builds its own fabs, how do you think TSMC, Samsung, and Intel will react? Lucas: TSMC is already responding. They're building new fabs in Arizona and Japan, partly to keep Nvidia as a customer. But the reality is, Nvidia is TSMC's largest customer — they account for roughly 15 to 20 percent of TSMC's revenue. If Nvidia starts pulling volume in-house, that's a significant revenue loss for TSMC. So TSMC will likely compete aggressively on price and technology to retain Nvidia's business. Luna: So it could actually benefit Nvidia even if they never fully bring production in-house — just the threat of it gives them negotiating leverage with TSMC. Lucas: Exactly. That's a classic 'make or buy' strategy. By building the capability to manufacture internally, Nvidia gets better pricing and priority from TSMC. Even if they only shift 20 percent of volume in-house, the savings on the remaining 80 percent could be substantial. Luna: That's a smart way to think about it. So the stock's modest this-week move might be underestimating the strategic value. Lucas: I think so. The market tends to focus on near-term earnings beats and misses. But this foundry move is a multi-year thesis that could reshape Nvidia's competitive position. It's worth watching how they execute on it. Luna: Alright, let's keep an eye on their next quarterly report for capex guidance. Thanks, Lucas. Lucas: Thanks, Luna. And for the listener, if you want to dig deeper into Nvidia's financials, check out their investor relations page. The details on the foundry investment are in their 10-K. We'll be back tomorrow with another stock pick.