Latest / Stock Picking with Fexingo: Individual Equities, Research, and Building a Concentrated Portfolio / How CoreWeave Is Redefining AI Infrastructure Stock Picking
Transcript
- Lucas: So CoreWeave is up another four percent today, and it's been one of the most actively traded names in the market all week. For a stock that's been around for less than two years as a public company, it's already become a fixture in almost every ai focused ETF. Luna: Every time I see CoreWeave in a headline it's either 'biggest mover' or 'record high'. But I still don't have a clear picture of what the company actually does. Lucas: Right, so let's strip it down. CoreWeave is essentially a GPU cloud provider. They buy NVIDIA chips in bulk, build out massive data centers, and then rent that computing power to companies like Microsoft and Meta that need it to train and run AI models. They're not a software company, they're not a semiconductor company — they're the landlords of the AI boom. Luna: So when we talk about stock picking in this space, we're really talking about a company that's leveraged to the same AI capex cycle as NVIDIA, but with a very different risk profile. Lucas: Exactly. And that's what makes it interesting for a concentrated portfolio. If you already own NVIDIA, which is up two point two percent over the past five days, do you also want CoreWeave? That's the question I keep coming back to. And I think the answer depends on whether you're betting on the whole AI buildout or just the picks and shovels. Luna: Before we go deeper on that, I want to say — if today's episode helps you think about a decision differently, that's exactly what we're going for. And the way we keep these conversations ad-free is listener support. If you find value here, you can buy me a coffee at buy me a coffee dot com slash fexingo. Lucas: Yeah, that's the whole engine. It's not about perks or anything, it just keeps the lights on. Okay, back to CoreWeave. The thing that jumps out to me is the balance sheet. Luna: I've seen their debt numbers floating around. It's a lot. Lucas: It's a lot. They've financed most of their data center buildout with debt, which is typical for infrastructure plays, but the sheer scale here is unusual. As of their last quarterly filing, total debt was north of fifteen billion dollars, and they're still burning cash on an operating basis. That's not inherently a red flag — growth companies often spend ahead of revenue — but it does mean the key variable is whether those long-term contracts with hyperscalers actually hold up. Luna: And those contracts are with Microsoft and Meta, right? So they're essentially relying on two or three customers for the majority of their revenue. Lucas: That's the concentration risk in a nutshell. Microsoft alone accounts for something like sixty percent of their contracted revenue. Now, those are take or pay contracts, which means the customers have to pay even if they don't use the capacity. So in the near term, the revenue is fairly locked in. But the market is really pricing in the next wave of contracts, and that's where the uncertainty lies. Luna: So when we see the stock pop four percent today, is that just momentum, or is there a tangible catalyst? Lucas: Part of it is broader sentiment — the NASDAQ is up one percent this week, and AI names are leading the rebound. But there's also a specific catalyst: CoreWeave announced a new multi-year agreement with an unnamed Fortune 50 company to provide GPU capacity for inference workloads. That's a signal that they're expanding beyond their core training client base. Luna: That's a big deal because it starts to address the customer concentration worry. Lucas: Exactly. And it's why the stock is up thirty percent from its low in June. The market is starting to believe they can diversify. But here's my caution: the valuation is already reflecting a lot of that optimism. CoreWeave is trading at roughly twelve times forward revenue, which is higher than NVIDIA's multiple. And NVIDIA actually has the pricing power, the ecosystem, and the margins to justify a premium. CoreWeave is essentially a pass-through business with thinner margins and higher leverage. Luna: So for a concentrated portfolio, would you rather own NVIDIA or CoreWeave? Lucas: I'd rather own NVIDIA, honestly. But that's not to say CoreWeave doesn't have a place. If you're building a basket of AI infrastructure plays, CoreWeave gives you exposure to the data center buildout itself, which NVIDIA doesn't directly give you. NVIDIA sells the chips, but it doesn't operate the data centers. CoreWeave is more like a REIT for GPUs, but without the steady dividend. Luna: That's a useful framework. So when you're evaluating an infrastructure stock like this, what are the key metrics you look at beyond revenue growth? Lucas: First, contract duration and customer concentration. You want to see long-term agreements with a diversified base. Second, the cost of capital — how much debt are they carrying and at what rates? With rates where they are today, that's a significant drag. Third, utilization rates. You can have a lot of capacity, but if the GPUs are sitting idle, the economics fall apart. And finally, the path to free cash flow. CoreWeave is still negative on that front, and I'd want to see a credible timeline to positive cash flow within the next two years. Luna: And right now, what does that path look like for CoreWeave? Lucas: It's improving, but it's not there yet. They've been able to extend maturities on some of their debt, which pushes the refinancing wall out to 2028. But they're still paying high single-digit interest rates on a lot of that debt. If the revenue growth starts to slow, or if a big customer renegotiates, that leverage becomes a real problem. Luna: What about the recent headlines about the New York City Council probing prediction market platforms? I know that's not directly related, but it got me thinking about how much of this AI trade is driven by speculative flows rather than fundamentals. Lucas: That's a fair connection. The same way prediction markets can get ahead of themselves, the AI trade has a speculative element. But I'd argue CoreWeave's contracts are more tangible than a prediction market position. The real risk is that the entire AI capex cycle slows down. If the hyperscalers decide they've overbuilt and cut back on new orders, CoreWeave would feel that immediately. Luna: So in a sense, you're saying CoreWeave is a leveraged bet on the AI buildout continuing at the current pace. Lucas: That's exactly what it is. And that's why I think it's more appropriate as a satellite position rather than a core holding in a concentrated portfolio. The core should be companies with durable moats and strong balance sheets. NVIDIA, Microsoft, even Apple to a degree. CoreWeave is a higher-beta way to express the same thesis, but you have to size it accordingly. Luna: What about the ETF angle? You mentioned earlier that CoreWeave is in a lot of AI ETFs. How does that affect the stock's behavior? Lucas: That's a great point. When a stock is heavily owned by index funds and ETFs, it can lead to forced buying and selling regardless of fundamentals. CoreWeave is now a top ten holding in several popular AI ETFs, which means every time there's a rebalancing or inflows, there's going to be demand for the shares. That can amplify moves in both directions. Over the past week, CoreWeave has been more volatile than the NASDAQ, and that's partly a function of this ETF ownership. Luna: So you're essentially saying that the stock has a built-in bid, but also a built-in source of selling pressure when sentiment turns. Lucas: Right. And that's why I think it's important to separate the business from the stock. The business is straightforward: sell GPU capacity to the biggest AI consumers. The stock is a different animal, influenced by flows, sentiment, and leverage. For a stock picker, that means you have to have a clear thesis on both the operational side and the market structure side. Luna: What do you think the market is missing about CoreWeave right now? Lucas: I think the market is underestimating the competitive pressure from the hyperscalers themselves. Microsoft, Google, and Amazon are all building their own custom silicon. If they shift more of their AI workloads to their own chips, that could reduce their reliance on NVIDIA — and by extension, on CoreWeave's nvidia based infrastructure. That's a longer-term risk that I don't think is fully priced in. Luna: So it's not just about the next earnings report, it's about the multi-year trajectory of the AI stack. Lucas: Exactly. And that's why I like to look at the entire value chain. When you pick a stock like CoreWeave, you're also implicitly making a call on NVIDIA, on the hyperscalers' in-house chip efforts, and on the broader AI adoption curve. That's a lot of variables for a concentrated portfolio. It can be done, but you have to be aware of what you're actually owning. Luna: How would you play it if you wanted to add CoreWeave to a portfolio today? Lucas: I'd start with a small position, maybe one to two percent, and scale in over time. Set a mental stop based on the business story, not the stock price. If they continue to diversify customers and show progress towards positive free cash flow, I'd be willing to add. If they announce a troubled debt refinancing or lose a major contract, I'd reconsider the whole thesis. Luna: That's a disciplined approach. And it aligns with the broader theme of this show: building a concentrated portfolio around high-conviction ideas, but with clear risk management. Lucas: Exactly. And I think CoreWeave is a perfect case study for that because it forces you to think about what you're really buying. Are you buying a utility for the AI age, or are you buying a leveraged bet on the most speculative part of the tech trade? The answer probably lies somewhere in between, and where you land on that spectrum should determine the size of the position. Luna: Looking ahead, what's the one thing you'll be watching for when CoreWeave reports earnings? Lucas: I'll be watching the mix of revenue between training and inference. If inference is growing faster, that's a good sign because it means they're moving beyond the initial training surge and into ongoing production workloads. I'll also listen for any color on new contract signings, especially with customers outside the big three hyperscalers. And of course, the debt situation — any news on refinancing at better rates would be a positive catalyst. Luna: That gives us a clear checklist. Anything else that investors often overlook with this type of infrastructure stock? Lucas: The depreciation schedule. CoreWeave has a lot of capital expenditure, and they depreciate those GPUs over a certain number of years. But the actual useful life of these chips might be shorter, especially as new generations come out. That could lead to impairments down the road. It's an accounting detail, but it can have a big impact on reported earnings. Luna: So it's not just about the income statement, it's about the assumptions behind the numbers. Lucas: Right. And that's the kind of thing that separates a thoughtful stock picker from someone just chasing momentum. CoreWeave is a fascinating company, but it demands a lot of homework. If you're not willing to dig into the debt, the contracts, and the competitive dynamics, you're better off sticking with the simpler names in the AI trade. Luna: That's a solid takeaway. So, final thought — is CoreWeave a buy right now? Lucas: I'd say it's a 'hold and watch'. The momentum is real, and the business is improving, but the valuation and leverage give me pause. If the stock pulls back to a level that reflects more conservative growth assumptions, I'd be more interested. For now, I'd rather own NVIDIA or even Microsoft, which have more predictable earnings and stronger balance sheets. But CoreWeave is definitely a name to keep on the radar for 2027.