Over the past few months, markets have experienced an almost unconditional affair with anything related to Artificial Intelligence. Large and medium-sized companies have jumped on an unprecedented wave of investment, pushing valuations, fueling expectations, and establishing a new buzzword vocabulary: accelerated data centers, language models, agents, sovereign computing. In this climate of euphoria, NVIDIA has become the emblem of the era: a chip and software platform provider that has grown at a dizzying pace, now competing for the throne of global market capitalization.
In the same environment, Michael Burry— the investor who warned of the 2008 mortgage collapse— has decided to go against the grain. His fund, Scion Asset Management, has declared short positions through put options on NVIDIA and Palantir. This is no minor gesture or attention-seeking stunt: it’s a direct challenge to the narrative supporting the AI rally. The inevitable question fuels headlines and hallway conversations: Is there a bubble about to burst?
The move: puts on AI icons and defensive calls
Regulatory filings reveal that Scion has taken positions in put options equivalent to millions of shares of NVIDIA and Palantir. These contracts gain value if stock prices fall; they do not entail a classic short sale but clearly express a bearish outlook or hedging needs. Additionally, Burry has included call options on Halliburton and Pfizer—two companies with very different profiles—energy and pharmaceutical—and distanced from the AI sector frenzy. This deliberate contrast reduces exposure to “narrative risk” and diversifies potential macro catalysts (crude oil prices, healthcare cycle, public and private investment outside the tech domain).
It’s worth noting a nuance often lost in market noise: 13F filings reflect a delayed snapshot, without details on strike prices or expiration dates, and allow for complex strategies (spreads, hedges, pairs). In other words, Burry is signaling, not necessarily betting on an imminent collapse. But the signal, coming from him, carries weight.
A tense tableau driven by expectations… and capital
Burry’s thesis hinges on a perception increasingly heard: investor enthusiasm may be outpacing the actual economic reality. The valuations of AI leaders discount years of perfect growth, margins sustained in the stratosphere, and endless investment appetite from tech giants, governments, and regulated sectors. Capital expenditure on AI infrastructure has become a key indicator: commitments for 2025 and 2026 already point to an aggregate figure many analysts place well above $300 billion. The bullish view argues we’re witnessing a new industrial layer—an “operating system” for the economy—and that each euro invested today will be monetized exponentially tomorrow. The skeptical view fears “circular capex”: giants buying technology from each other, financed by markets demanding growth, in a cycle that could slow down if returns take longer than expected.
This is where Burry sharply criticizes the narrative. His implicit message isn’t to deny the technological revolution—doing so would be absurd—but to question the exuberance in multiples and the underestimation of execution risk. Discomforting questions emerge: Will all AI deployments generate positive cash flow within the timeframes reflected in stock prices? How many pilot projects will surpass the real return threshold beyond demos? What happens if installed capacity exceeds usable demand for a couple of years?
NVIDIA—undisputed leader or under constant scrutiny?
NVIDIA has established a dominance difficult to challenge: GPUs for training and inference, interconnection networks, software (CUDA, cuDNN, Triton, frameworks), and a developer community that amplifies network effects. Demand, for now, exceeds supply despite ramped-up advanced packaging. The company insists there’s no bubble: what’s being built is foundational infrastructure for the coming decade, and the shift from generalist computing to accelerated is not a fad but a paradigm change. In this view, chips aren’t an end but part of a complete stack that will continue attracting investment as AI shows practical benefits in productivity, research, industry, or services.
Critics counter that technical leadership alone doesn’t shield against cyclicality. They note that part of recent growth stems from unprecedented concentration of spending among a few with massive balance sheets. If these portfolios slow down—due to prudence, interest rate changes, regulation, or saturation—the impact on orders and guidance could be abrupt. In this hypothetical scenario, a widely owned company by institutions and retail investors becomes more vulnerable to shocks in expectations.
Palantir—between mission and multiples
Another focus of Burry is Palantir, currently experiencing a strong period: significant defense traction, strategic contracts, a narrative of data platform for enterprises and governments, and a discourse blending mission—“software for a cause”—with commercial acceleration. CEO Alex Karp has responded vehemently to shorts, defending the business outlook and tangible utility of his products. But the debate over multiples persists: how sustainable is a valuation that discounts near-exponential growth in the medium term? The tech history is filled with excellent companies that, for years, were mediocre investments because they paid too early for distant horizons.
The ecosystem casts a wary eye: AMD, suppliers, and “second derivatives”
Burry’s move isn’t happening in a vacuum. The entire value chain watches closely: direct competitors like AMD, manufacturers, assemblers, packaging providers, high-speed networks, integrators, and of course end clients—from hyperscalers to banks, pharma, automotive, or healthcare. Any hiccup in lead order rates trickles through by capillaries: budgets are re-evaluated, schedules shift, projects move from “go” to “wait”.
But upside secondary effects could emerge if the cycle normalizes: lower component price tensions, better supply chain planning, greater focus on energy efficiency and true TCO, and a healthier ecosystem shifting from “buy what’s available” to “buy what makes sense”. In such a scenario, providers with specific value propositions—ranging from alternative GPUs to optimized software or niche infrastructure—could gain market share without systemic shocks.
Bubble warning or prudence reminder?
Burry’s stance weighs heavily on interpretation. Many see his move as a “bubble top” signal for AI trading. Others view it as what it truly is: a well-articulated contrarian opinion worth considering to avoid herd thinking. In any case, his thesis prompts investors to undertake the stress tests every responsible investor should do amid euphoria: What if monetization timelines extend by 12 or 24 months? What are the assumptions on real utilization of accelerated data centers? How much demand depends on four or five buyers with their own agendas? How sensitive are margins to shifts in the training-inference mix?
On the corporate side, the blind spot often lies in linearly extrapolating successful pilots into massive deployments. On the market side, the risk is confusing narrative with actual cash flow. The debate isn’t whether AI matters— it does—but what fair valuation each part of the chain deserves at each cycle stage.
The “green side” response
NVIDIA’s stance remains consistent: there’s no bubble, just a technological transition. The figures for committed demand and guidance from major buyers support this interpretation—so far. Moreover, the company has reiterated that supply will remain strained for several quarters due to manufacturing and packaging limitations, even with capacity expansion investments. Meanwhile, proprietary software and ecosystem tools bolster a competitive moat that makes it more than just chips at a price per FLOP.
Yet, all theories—no matter how robust—must coexist with result discipline. In a market that has rewarded perfection, any nuance—a more cautious guidance, delayed deliveries, or shifted spending—can trigger a multiple correction. That’s the space where Burry’s approach finds opportunity.
What’s next?
The immediate uncertainties are twofold: first, whether Scion’s positions stay aligned in size and direction; history shows some managers swiftly adjust exposure with derivatives. Second, whether the “AI consensus”—from hyperscalers to governments—confirms with actual cash flows and contracts what the narrative promises. Parallel to this, monitoring regulator sentiment, interest rate trends, and hardware cycles is essential, as each new generation compels reinvestment and reassessment of returns.
Whatever the outcome, Burry’s move plays a valuable role: popping the complacency bubble, if any. In the best case, the sector continues to prove utility and monetization, and markets recalibrate expectations smoothly. In the worst, excess investment may seek an exit through common bubble channels: forced sales, multiple compressions, fleeing into perceived safe assets. No one can honestly say today which path will dominate. But viewing both possibilities remains healthy—and professional.
FAQs
What does it mean that Michael Burry is buying “puts” on NVIDIA or Palantir?
A put option is a contract that gains value if the underlying stock drops below a strike price before expiration. It doesn’t necessarily involve short selling or betting on an immediate crash. It can be a hedging strategy, a bearish directional position, or part of a more complex trade.
How much can a 13F reveal about a fund like Scion’s investments?
13Fs show long positions at quarter-end and, for derivatives, the contract types and approximate notional sizes. They do not include detailed strike prices, precise expiration dates, or whether options are part of spreads or hedges. Plus, they are reported with a delay—several weeks—so positions may have changed.
If a “AI bubble” bursts, which sectors would be most affected?
Sectors most exposed would include those directly linked to accelerated data center investment—chip makers, component suppliers (GPUs, HBM memories, interconnects), advanced packaging, integration providers, and some software firms tied to large-scale deployments. Companies whose valuations depend on unproven monetization prospects might also suffer. Conversely, defensive sectors (healthcare, consumer staples, energy) tend to better withstand volatility.
How might a severe correction in NVIDIA impact other players like AMD?
A sharp decline in NVIDIA due to expectation revision might cascade to peers through sector multiple compression. But effects won’t be uniform: each company has its roadmaps, partners, product mix, and margin levers. During normalization cycles, some alternative suppliers might even gain market share as buyers diversify to reduce dependence.
Sources
- Barron’s: breakdown of Scion Asset Management’s puts on NVIDIA and Palantir and portfolio weight (Barron’s)
- Business Insider: coverage of Burry’s short positions and mention of calls on Halliburton and Pfizer; Alex Karp’s response (markets.businessinsider.com)
- Yahoo Finance / WhaleWisdom / Fintel: confirmation of 13F filings with puts on NVDA and PLTR and calls on PFE and HAL (Yahoo Finance UK)
- Bloomberg / MarketWatch / Reuters: Jensen Huang’s comments on “no AI bubble” and sustained demand; context of industry capex (Bloomberg)
- Business Insider: increase in big tech capex for AI infrastructure in 2025–2026 (Business Insider)