We are living through a market moment built on a dazzling idea: that artificial intelligence (AI) will remake business, lift productivity, and deliver a new era of profits. That idea is real. But what’s also real is that markets—like people—get caught up in stories, and stories can outrun substance. The recent giant bets by Michael Burry against two blue-chip AI names should be a loud wake-up call: when an investor who famously profited from a prior systemic excess smells a bubble, it’s worth asking why.
What happened (quick version)
In a regulatory filing for the quarter ending September 30, Scion Asset Management — Michael Burry’s fund — disclosed large put-option positions (bets that a stock will fall) on Palantir and NVIDIA: roughly $912 million of exposure to Palantir and about $187 million to Nvidia (total ≈ $1.1 billion). The news pushed short-term volatility in both stocks and refocused attention on whether AI enthusiasm has outpaced earnings and economics. 
Shorting, simply explained
Shorting is the mirror image of buying. Instead of paying now and hoping a price rises, you borrow (or buy an option) to sell at today’s price and profit only if the asset falls. It’s high-risk: if the price climbs, losses can be large and fast. Puts can limit some downside to the premium paid, but betting against high-momentum, widely held stocks is still a bold move. Burry’s put purchases are a way of saying: “I think these prices are wrong.” 
Why Burry’s move matters
Burry is not famous because he likes being contrarian; he’s famous because he anticipated the U.S. housing bubble in the mid-2000s and profited by buying credit-default swaps (CDSs) that paid off when mortgage securities collapsed. That episode shows an ability to see structural risk markets missed; it also explains why many investors read his moves as more than a stunt. But one caution: every market is different, and a correct past forecast is not a guarantee of future accuracy.
The anatomy of this AI-led market
Two forces are at work now:
- Concentration of capital. A handful of companies (the “Magnificent Seven”) now account for an outsized share of market value. Investors have poured cash into the winners and into firms that look like the winners. That concentration amplifies both gains and losses: if the leaders stumble, the index falls fast.
- Cross-investment and circular stories. Big tech firms are customers for one another’s AI services, invest in AI startups, buy chips from the same suppliers and sometimes own equity in fellow platform players. That entanglement makes growth look internally generated: company A buys from B, B’s revenue looks strong, investors cheer, and the cycle repeats. It’s healthy in moderation — dangerous when used to justify ever-higher valuations unmoored from net new revenue. (See corporate disclosures and recent market analysis showing how much of 2025 market gains were concentrated in a handful of AI-exposed names.)
Numbers matter: valuations vs economics
Nvidia, the chipmaker whose processors train many large AI models, has seen extraordinary multiple expansion because its products are indispensable to model builders. But GPUs and data-center infrastructure aren’t free. Training advanced models uses huge amounts of compute and energy, building the supply chain and the datacenters to host that compute costs billions. If revenue growth from AI services slows, or if margins compress because of rising costs (hardware, power, cooling), the math that justifies today’s valuations weakens. International Energy Agency (IEA) and technical studies suggest AI workloads could drive data-center electricity demand far higher by 2030 — a real cost that must be paid in cash and capital, not hype.
The danger: bubbles wipe value overnight
The dot-com comparison is not a poetic flourish — it’s a structural lesson. In 1999–2000, investors rewarded stories of internet scale even when profits were distant. When reality (cash flow) failed to follow expectations, prices collapsed and trillions of market value evaporated. Today’s AI story shares features with that era: rapid fund flows, speculative private investments, sky-high valuations in a concentrated set of firms, and a public narrative that assumes perpetual, frictionless expansion. One or two shocks — a disappointing earnings cycle, a regulatory clampdown on data or model use, supply constraints for chips, or a sudden spike in energy costs — could flip the narrative. When sentiment turns, paper value disappears quickly.
Real limits that investors are underweight
Energy and infrastructure: Data centers already use a meaningful slice of global electricity, and their consumption has been growing quickly. Large language model training is energy-intensive; projections show AI workloads could account for a material share of future data-center consumption. This is not an abstract environmental talking point — it affects operating costs and capital plans.
Hardware cycles: Advanced chips are expensive, supply is finite, and performance gains slow over time. Building capacity (and cooling) for training and inference is a multi-year, capital-intensive process.
Regulation and data limits: Privacy rules, export controls on AI hardware/software, and sectoral rules on how models may be used (health, law, finance) could constrain growth and raise compliance costs.
Read more: How an Israeli Strike in Doha, Forces Saudis to Bring Pakistan in to create a New Regional Order?
Concentration risk: With a few firms carrying disproportionate index weight, pension funds and passive investors may be exposed indirectly to concentrated bets made for reasons unrelated to their long-term liabilities.
So what should readers and policymakers take from this?
AI is not a fad. It will transform industries and create real value. But the market is not the same as the economy. When investors borrow the future to justify today’s valuations, they expose themselves to abrupt reversals. Burry’s puts are a hedge on that mismatch: he’s betting that at least some of today’s winners will disappoint relative to sky-high expectations. Whether he is right or wrong, his move is a useful reminder to separate engineering possibilities from profitable business models, and to price in the real costs of computing, energy, and regulations.
Policymakers should recognize the rapid growth of energy demand in the data-center sector and plan accordingly — the transition to cleaner power is essential but expensive. Investors should ask simple questions before piling into AI-themed names: Where will the revenue come from? How sustainable are margins after infrastructure and energy costs? What could go wrong next quarter that would make buyers panic?
Final thoughts
The AI boom is a historic technological shift, but history shows that revolutions often arrive with excesses. The dot-com boom birthed the modern internet — and left scars for investors who chased unicorns. The same could happen to AI: we can embrace the technology while avoiding a repeat of past market excess. Prudent capital, clear accounting of costs, better policy planning, and the humility to separate hype from economic reality will reduce the chance that today’s enthusiasm becomes tomorrow’s catastrophe.
Under the pen name Patience Quill, the author explores the intersection of global politics and economics, where national ambitions collide with financial realities.
The opinions presented here reflect the author’s personal analysis and experience, which may not fully align with the publication’s editorial outlook.