Highlights

• Nvidia and Palantir's extreme valuations—trading at 24-46× and 400-480× earnings respectively—reflect inflated AI expectations that conflate infrastructure buildout with actual revenue generation and profitability.
• Michael Burry's thesis centers on the dangerous gap between unprecedented hyperscaler capex spending ($530B projected for 2026) and unproven AI monetization, compounded by tightening capital markets and extended accounting practices.
• While markets price a frictionless AI future, physical constraints—including critical mineral supply chains, grid capacity, Chinese competitive advantages in applications, and geopolitical risks—remain systematically underpriced.

---

Rare Earth Exchanges™ (REEx) exists to translate the world’s most strategic supply chains into investor-grade reality—combining accessibility, transparency, and insight. That mission is shaped in party by downstream demand tied to artificial intelligence. The same hyperscaler capital-expenditure surge powering GPU demand is also accelerating requirements for energy infrastructure, advanced materials, precision manufacturing, robotics, and defense-grade electronics—systems that ultimately depend on critical minerals and rare earth elements at multiple choke points.

Yet markets are beginning to behave as if this demand curve is both infinite and risk-free. Valuations across “AI-adjacent” winners are inflating far faster than underlying cash flows, while the physical side of the equation—permitting timelines, grid constraints, processing bottlenecks, geopolitics, and China’s embedded leverage—remains stubbornly real. In effect, investors are pricing a seamless AI future while discounting the hard constraints of the mineral-to-magnet-to-machine economy. That disconnect sits at the heart of today’s AI trade—and the growing risk embedded within it.

The AI EquitySurge: Extraordinary Gains, Fragile Assumptions

Few would dispute that NVIDIA and Palantir Technologies have enjoyed historic stock runs fueled by artificial-intelligence enthusiasm. Nvidia briefly crossed a $5 trillion market capitalization in 2025. Palantir surged past $400 billion. These figures are extraordinary—and increasingly difficult to reconcile with economic gravity.

Palantir is the clearer valuation outlier. With under $4 billion in annual revenue, the company trades at over 100× price-to-sales, 400× trailing earnings, and roughly 480× EV/EBITDA. These multiples exceed even the peak valuations of Cisco Systems or Amazon during the dot-com bubble. They imply decades of flawless execution, uninterrupted government spending, and minimal competitive erosion—conditions that history shows are rarely sustained.

Nvidia’s valuation is more internally coherent, but still fragile. The company is immensely profitable, generating nearly $100 billion in net income and more than $50 billion in free cash flow, with operating margins north of 60%. At roughly 46× trailing earnings and 24× forward earnings, Nvidia is expensive but defensible if current AI demand persists. The risk is not business quality—it is cycle risk. Nvidia’s results reflect an unprecedented hyperscaler capex surge. Any normalization in spending could compress margins and multiples quickly.

Michael Burry’s Core Warning: Infrastructure Is Not Demand

This is where Michael Burry’s critique becomes essential. The hedge fund investor, famous for identifying the U.S. housing bubble ahead of the 2008 crisis, Burry argues that today’s AI boom shows familiar signs of mania: assets treated as if they “can’t go down,” justified by narratives claiming valuation rules no longer apply.

Recent capital-spending data reinforces his concern. Big Tech spent approximately $61 billion on data centers last year alone. According to Goldman Sachs, global AI investment is projected to rise from roughly $400 billion to nearly $530 billion in 2026. Hyperscalers have increased AI capex by more than 60% in each of the past two years, with another ~30% increase planned.

The structural problem is funding. This expansion is unfolding in a tightening capital environment, not an era of zero-cost money. The U.S. Treasury alone must refinance roughly $8 trillion in debt, sending a wave of bond issuance into global markets just as corporate borrowing needs are peaking. Capital is no longer free, and AI infrastructure is among the most capital-intensive investments in modern history.

Burry’s central insight is that markets are conflating front-loaded infrastructure buildouts with durable end-user demand. Data centers are being built first; monetization is expected later—perhaps. If AI revenues fail to scale proportionally, today’s orders risk becoming tomorrow’s overcapacity.

Accounting practices may further cloud the picture. Extended depreciation schedules smooth near-term earnings, making returns appear stronger than the underlying economics justify. This is not fraud—but it is a distortion.

The Monetization Gap—and Why China Matters

A critical follow-on question is whether AI can be monetized fast enough, particularly in the United States. American technology leaders have built world-class frontier models, but many still lack scalable, cash-generating applications (in the pure AI space). Subscription chatbots have limits. The number of users willing to pay $20–$50 per month indefinitely is finite.

Investors, increasingly, are no longer rewarding capex alone; they want demonstrable returns.

China, by contrast, appears to be advancing more rapidly into the applications phase of AI adoption. Chinese AI firms benefit from:

• Cheaper capital
• Abundant and lower-cost power
• Lower operating costs
• Application-first system design

Beijing-backed AI investment is expected to exceed $70 billion this year, while power demand for Chinese data centers rose 25% last year. Export controls have forced Chinese developers to innovate around hardware constraints, resulting—according to multiple analysts—in more efficient models capable of running on less advanced chips. If sustained, this trend could pressure Nvidia’s long-term pricing power at the margin.

This matters because frontier models are increasingly commoditized. Applications capture value. China recognizes this dynamic—and may be moving faster in translating AI capability into economic output. But as we have also reported, the Chinese face their own over-production crises.

Palantir’s Structural Risks: Politics and Dilution

Palantir carries additional vulnerabilities that amplify valuation risk. A significant share of its revenue derives from U.S. government contracts—lucrative, but politically contingent. In Q1 2025 alone, Palantir booked $373 million from U.S. government clients, up more than 40% year-over-year. That growth depends as much on policy priorities as on market demand.

Burry has described this dependence bluntly as a form of “welfare.” Whether one agrees with the phrasing or not, the underlying risk is real: fiscal restraint rhetoric tends to resurface quickly when debt pressures intensify.

Insider enrichment compounds the concern. Palantir’s stock-based compensation exceeded $690 million in 2024, outpacing net income. Five billionaires emerged from a company generating less than $4 billion in annual revenue—an almost unprecedented ratio. Dilution rewards insiders while transferring risk to public shareholders, a pattern that historically appears closer to valuation peaks than beginnings.

Valuation Gravity Still Applies

Burry’s long-dated put options, expiring in 2027, are not calls for imminent collapse. They reflect patience. For a meaningful payoff, Nvidia might need to fall by ~40% and Palantir by ~70%. Those numbers sound extreme—until history intervenes. Cisco lost more than 80% of its value after 2000. Amazon fell nearly 90% before rebuilding.

To be clear: AI is very real, and its expansion is influencing critical minerals markets—from grid metals and energy systems to magnet demand and defense electronics. Nvidia is a genuine industrial champion. Palantir builds powerful software used by governments and enterprises. But great companies can still be poor investments at the wrong price.

A sober review of current financial data supports Burry’s core thesis: these stocks are priced for perfection in a world that rarely delivers it.

Burry is not betting against AI.

He is betting against unpriced risk—excess leverage, fragile monetization, geopolitical competition, and capital cycles that always turn.

History suggests that is rarely a foolish wager.

REEx Bottom Line

• AI infrastructure does not equal AI profits
• Capex not equal to monetization
• Narratives not equal to cash flows
• Valuation always matters—eventually

The emperor of AI is not naked.

But he is overdressed—and the market is starting to notice.

©!-- /wp:paragraph -->