The Temporary Monopoly: Michael Burry’s Short on Nvidia and the Unseen Cost of Centralized Compute

When the most famous contrarian in finance, Michael Burry, places a bet against the world’s most valuable chipmaker, it’s not just a trade—it’s a philosophical reckoning. Burry’s move to short Nvidia, while simultaneously buying call options as a hedge, signals a deep skepticism about the sustainability of the company’s dominance. At first glance, it seems like a classic value play: bet against a stock trading at 60 times earnings, with a market cap that has swelled past $2 trillion. But beneath the surface, this is a story about the fragility of technological monopolies, the hidden costs of AI infrastructure, and the quiet truth that code betrays when we do. Burry’s logic, as reported by BeInCrypto, is grounded in the belief that Nvidia’s “monopoly” is temporary. He acknowledges the company’s current pricing power and CUDA ecosystem lock-in, but argues that increasing competition—from AMD, Google’s TPU, and custom ASICs like Amazon Trainium and Microsoft Maia—will erode margins and revenue growth. Burry also points to the risk of a “capital expenditure bubble” in AI, where hyperscalers overspend on Nvidia hardware, only to see returns diminish. To hedge, he buys call options on Nvidia, a protective strategy that limits his downside if the stock rallies. This is not a full conviction short; it’s a calculated bet on mean reversion. As a decentralized protocol PM who has spent years navigating the intersection of code and capital, I see Burry’s thesis as both compelling and incomplete. He is right to question the durability of Nvidia’s dominance, but he may be underestimating the power of software ecosystems that have been built over decades. The CUDA platform, with its 4 million developers, is not just a technical moat—it’s a cultural one. Migrating from CUDA to ROCm or other frameworks is not a matter of a few months; it’s a generational shift. In my own experience auditing smart contract protocols, I’ve seen how even superior code cannot overcome the inertia of an established user base. The same principle applies here: Nvidia’s monopoly is not just in hardware, but in the collective habits of the AI developer community. Yet, the cracks are visible. The Blackwell architecture, while impressive, faces competition from specialized chips designed for inference, such as Groq and Cerebras, which offer better energy efficiency for specific workloads. The shift from “general-purpose” to “specialized” AI compute is real, and it represents a structural risk to Nvidia’s business model. Burnout is the tax on innovation, and Nvidia’s relentless annual product cycles are a form of innovation that may eventually exhaust both the company and its customers. The hyperscalers—Microsoft, Meta, Amazon, Google—are already designing their own chips, not just to save costs, but to gain independence from a single supplier. This is a classic “prisoner’s dilemma” in procurement: each company knows that relying on Nvidia is risky, but building a competitive chip takes years, and the coordination costs are high. From a valuation perspective, Burry’s short is not unreasonable. Nvidia’s price-to-earnings ratio of 60x in early 2025, while justified by 120% revenue growth, leaves little room for error. If growth slows to 50% or even 30%, the multiple compression could be severe. The company’s gross margin of 70% is a testament to its pricing power, but that margin is under threat as AMD’s MI300 series closes the performance gap and as custom ASICs begin to scale. The question is not whether Nvidia will lose market share, but how fast and how much. Burry’s central thesis—that capital expenditure in AI is entering a “bubble top”—is plausible, but it overlooks the fact that many enterprises are still in the early stages of AI adoption. The demand for compute may not be a bubble, but a permanent shift in infrastructure spending. What Burry’s trade misses, however, is the “system-level” moat that Nvidia has built. It’s not just about selling chips; it’s about the entire stack: DGX systems, NVLink interconnects, InfiniBand networking, and CUDA software. This integrated solution makes it costly for customers to switch, even if alternative chips are competitive. I have seen similar dynamics in blockchain infrastructure, where protocols like Ethereum maintain dominance not because of technical superiority, but because of the ecosystem of wallets, dApps, and developer tools. Code betrays when we do, and in this case, Nvidia’s code is a cage that also serves as a shield. But there is a deeper ethical dimension to this story. Nvidia is the “arms dealer” of the AI revolution, and its products are used in everything from autonomous weapons to surveillance systems. The company’s export controls on China have already created a geopolitical rift, accelerating the development of Chinese AI chips. This is not just a business risk; it’s a moral hazard. As an industry, we are building a future where compute is increasingly centralized, controlled by a single company that can decide who gets access to the most powerful tools. Decentralization, the core principle of blockchain, offers an alternative: a distributed network of compute providers that is resistant to censorship and single points of failure. The rise of decentralized physical infrastructure networks (DePIN) like Filecoin, Akash, and Render is a direct response to this centralization. They aim to democratize access to compute, but they are still orders of magnitude smaller than Nvidia’s ecosystem. The contrarian angle here is that Burry may be too early. Nvidia’s monopoly could last longer than the market expects, especially if the company successfully transitions to a software subscription model (CUDA-X) that locks in recurring revenue. The “tax” on innovation that Burry predicts might be deferred by years, as hyperscalers continue to invest in Nvidia’s ecosystem even as they develop internal alternatives. The true test will come in 2026-2027, when the Rubin architecture is due and when custom chips like Trainium and Maia are expected to reach scale. Until then, Burry’s short is a bet on the speed of disruption, not the direction. In the world of decentralized finance, we often talk about “permissionless” access and “trustless” systems. Nvidia’s dominance is a reminder that the physical layer of AI is still deeply permissioned, controlled by a single gatekeeper. The blockchain industry’s answer—distributed compute—is still in its infancy, but it represents a moral imperative. If we believe that AI should benefit all of humanity, we cannot rely on a single company to provide the infrastructure. The cost of centralization is not just financial; it’s existential. Burry’s trade is a warning, but it’s also an opportunity for the crypto community to accelerate the development of decentralized compute alternatives. As I reflect on this from my desk in Manila, I am reminded of the burnout that comes from building in a hype-driven industry. The constant pressure to ship, to scale, to compete—it takes a toll. Nvidia’s employees feel it too, even if the stock price doesn’t show it. The tax on innovation is real, and it is paid in human hours and mental health. Burry’s shortsightedness, perhaps, is that he sees only the numbers, not the people. The code betrays when we do, and the deepest betrayal is when we forget that technology is supposed to serve us, not the other way around. In the end, this article is not about predicting Nvidia’s stock price. It’s about understanding the trade-offs we make when we centralize power, whether in AI or in finance. Michael Burry’s bet is a reflection of a broader anxiety: that the current boom is unsustainable, that the emperor has no clothes. But the emperor has a very strong suit of armor, made of CUDA and contracts. The question is whether that armor will crack before the next wave of innovation arrives. For now, the prudent observer watches, hedges, and builds the decentralized alternatives that will one day make such monopolies obsolete.