The Legendary Debate

Warren Buffett: leans back in his chair "Well, we've all looked at NVDA. Let me start with the big question—if we had to hold this business for 20 years, would we sleep well at night? Not looking at the stock price, just owning the enterprise itself."

Charlie Munger: "That depends entirely on whether you believe the current AI frenzy represents a permanent shift or another technology bubble. I've seen this movie before—the railroad boom, the dot-com era, the housing bubble. Human nature doesn't change. When you have everyone from taxi drivers to central bankers talking about AI, we're in the 'too much enthusiasm' phase of the cycle."

Dev Kantesaria: "Charlie, I understand your skepticism about cycles, but this isn't Pets.com. NVIDIA has created something remarkable—a complete ecosystem lock-in. Their software stack, CUDA, has become the lingua franca of AI development. It's not just chips; it's the entire computational architecture. This reminds me of ASML's monopoly in EUV lithography—once the ecosystem is built around your standard, customers can't leave even if they wanted to."

Pulak Prasad: "To answer Warren's question about 20 years: what fascinates me is NVIDIA's evolutionary adaptability. They've successfully navigated three major transitions—from gaming graphics to cryptocurrency mining to AI compute. This isn't a company that's resting on its laurels. They've demonstrated the survival fitness that separates long-term compounders from those that get disrupted."

David Tepper: "I'll tell you what keeps me up at night—the competition. AMD is coming hard with Instinct MI300X, and every cloud provider is designing their own AI chips. Google has TPU, Amazon has Trainium, Microsoft is working on Athena. When your biggest customers become your competitors, that's a structural problem that doesn't go away."

Robert Vinall: "David makes a valid point, but I'd push back on the competitive threat. The software moat is what matters here. CUDA has over 4 million developers trained on it. That's an ecosystem that took 15 years to build. It's not just about having a competitive chip—it's about having the entire software infrastructure. This is why Windows survived despite better operating systems existing."

Mohnish Pabrai: "All interesting points, but let me ask this: what happens when the AI bubble pops? Because bubbles always pop. In 2000, Cisco had 85% market share in routers, real products solving real problems, and it still took 15 years to get back to its peak. NVIDIA is trading at 40-50 times earnings at the peak of the cycle. That's not a 20-year compounder—that's a speculation."

Warren Buffett: "Mohnish raises the crucial question about cyclicality. Charlie, you've studied bubbles throughout history—what's different about this one?"

Charlie Munger: "What's different is the underlying technology is real and valuable, unlike the dot-com nonsense. But what's the same is the psychology. Every CEO now has to have an 'AI strategy,' every startup is an 'AI company,' and every investor is piling in. When I see that level of social proof and competitive frenzy, I remember what Keynes said: 'Worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally.' They're all buying NVIDIA because everyone else is buying NVIDIA."

Dev Kantesaria: "Charlie, I respect the psychological insight, but let's talk about the business fundamentals. The pricing power NVIDIA demonstrates is extraordinary. They're getting $30,000 for an H100 chip that probably costs $3,000 to make. That's 90% gross margins on incremental sales. When have you seen pricing power like that outside of monopolies or luxury goods?"

Pulak Prasad: "To Dev's point about pricing power—that's exactly what makes this interesting from an evolutionary perspective. Strong pricing power allows for massive R&D investment, which fuels further adaptation. NVIDIA is spending $8 billion annually on R&D. That's their mutation engine. They can afford to bet on multiple future technologies simultaneously."

David Tepper: "But here's the problem with that R&D argument: the pace of change in semiconductors is brutal. Remember Intel? They had 90% market share in CPUs, massive R&D budgets, and they still missed the mobile revolution. Now they're playing catch-up. What makes you think NVIDIA won't miss the next architectural shift?"

Robert Vinall: "David, that's exactly the right question. The answer lies in their platform approach versus Intel's component approach. Intel sold chips. NVIDIA sells systems—hardware, software, networking, everything integrated. That systems approach creates much higher switching costs. It's the difference between selling tires and selling entire car platforms."

Mohnish Pabrai: "All these arguments about moats and ecosystems—they all made sense for Cisco in 2000 too. But here's what nobody's mentioning: the customers are getting restless. The cloud providers don't want to be locked into NVIDIA's expensive ecosystem. They're actively working on alternatives. When your customers are actively trying to replace you, that's not a moat—that's a temporary advantage."

Warren Buffett: "Let's transition to the financials now. We've debated the qualitative story—what does the 10-year financial history tell us about this business quality?"

Dev Kantesaria: "The numbers tell an extraordinary story. NVIDIA's return on invested capital has gone from 15% a decade ago to over 90% today. That's the financial signature of a moat widening dramatically. They've achieved this while growing revenue from $4 billion to $60 billion. This isn't margin expansion through cost cutting—it's through pricing power and operating leverage."

Charlie Munger: "Dev, you're looking at the peak of the cycle. Look at 2019—ROIC was 20%, not 90%. What happens when AI spending normalizes? This 90% ROIC is like looking at a homebuilder's returns during a housing bubble. It's not sustainable."

Robert Vinall: "Charlie makes a fair point about cyclicality, but even if we normalize, the underlying trend is remarkable. If we take a 5-year average ROIC, it's still over 50%. That puts them in the top 0.1% of all businesses. The question isn't whether 90% is sustainable—it's whether the structural advantages support continued high returns."

Pulak Prasad: "The margin story is particularly telling. Gross margins have expanded from 55% to 70% over the decade, while operating margins went from 15% to over 50%. This isn't just cyclical—it's structural margin expansion as they've moved up the value chain from components to systems."

David Tepper: "But look at the capital allocation. They've spent $40 billion on stock buybacks over the past 5 years, mostly at much lower prices. That's great in hindsight, but what about now? They're buying back stock at 50 times earnings. That's value-destructive capital allocation at these prices."

Mohnish Pabrai: "David's exactly right. And look at the inventory swings—they got caught with massive inventory in the crypto crash, now they can't make enough. This is a brutally cyclical business disguised as a tech growth story. The volatility in working capital tells you everything about the underlying business stability."

Warren Buffett: "Now let's talk valuation. The stock sits at $187.67 currently. Who here would buy at this price?"

David Tepper: "I'm a buyer here. Not because it's cheap, but because the momentum is powerful. When you have this kind of institutional frenzy, it can go much higher than anyone expects. I learned during the financial crisis that sometimes you have to ride the wave even if you know it's a wave."

Mohnish Pabrai: "David, that's speculation, not investment. At 40 times earnings, you're paying for perfection. I need asymmetric upside—3:1 risk reward. Here, if AI spending slows even 20%, earnings could drop 50%. That's 1:1 risk reward at best. I'd avoid entirely."

Robert Vinall: "I appreciate the quality, but I can't get comfortable at this price. The crowd is too excited. I'd want to see it 30% lower, around $130, where you're paying 25 times normalized earnings rather than 40 times peak earnings."

Dev Kantesaria: "I'm with Robert. The business is extraordinary, but the price assumes continued 50% growth rates for years. At $150, I'd be interested. At current prices, I'm neutral."

Pulak Prasad: "Hold for me as well. I want to see how the next evolutionary adaptation plays out. The Blackwell transition is promising, but I need to see adoption rates before paying this multiple."

Charlie Munger: "Avoid. Entirely. This is peak enthusiasm. I've seen too many of these cycles. When it's obvious to everyone, it's usually wrong."

Warren Buffett: "I'm neutral as well. The business quality is exceptional—90% ROIC businesses don't come along often. But as Charlie taught me, the price you pay determines your return. At this price, the risk-reward doesn't favor us."

Warren Buffett: surveys the room "Let me try to synthesize where we've landed after this discussion. On the qualitative side, we broadly agree that NVIDIA has built something remarkable—a genuine ecosystem moat with CUDA, extraordinary pricing power, and demonstrated adaptability across multiple technology shifts. That's not nothing. In my experience, when customers are locked into your ecosystem and you can raise prices 50% without losing demand, you've got a special business.

But here's where we're divided: the sustainability of current earnings power. Charlie and Mohnish see classic cyclical excess—customers double-ordering, cloud providers building inventory, and the inevitable hangover that follows any technology mania. They point to Cisco's experience: great company, terrible investment if bought at the peak.

Dev, Pulak and I see it differently. NVIDIA's moat appears structural, not cyclical. The software ecosystem, the scale advantages in R&D, the systems approach—these aren't temporary advantages. Even if AI spending normalizes, they should maintain dominant market share and superior returns.

On valuation, we have a clear consensus: nobody finds current prices attractive. David is willing to ride the momentum, but the rest of us either want a substantial margin of safety or would avoid entirely. The key question is what normalized earnings look like. If this is truly a new paradigm, $10-15 EPS is possible in a few years. If it's cyclical, we could see $5-7.

I'm at hold—the business quality deserves respect, but the price demands perfection. We have four holds, two avoids, and one buy. The majority view is that this is a wonderful business at a questionable price. Reasonable investors can look at the same facts and reach different conclusions, which is exactly what makes markets interesting."