Michael Burry AI bubble warning flags 87% of VC money in AI

Michael Burry AI bubble warnings are back in focus after the famed investor drew a sharp parallel between today’s artificial intelligence frenzy and the late-1990s Dot-com bubble. In a post on X, Burry argued that capital tied to AI now looks more concentrated than internet investing did at the height of the last major tech mania, reviving a debate that has been hanging over markets as investors watch Nvidia and the broader tech trade.

The timing matters. The warning landed during a broader tech stock pullback, with attention fixed on Nvidia as a barometer for AI demand. However, Burry’s point went beyond one stock. He framed the current boom as a systemwide concentration story, where venture funding, bond issuance, and infrastructure spending are all being pulled toward the same theme.

That is what gives the comparison its sting. Burry did not just say AI is expensive. Instead, he argued that the scale of exposure now resembles the 1999 TMT bubble, when enthusiasm for a transformational technology spread far beyond equities and into the credit system.

Why the Michael Burry AI bubble warning is getting attention

Michael Burry’s core argument is simple: too much money is chasing AI, and the buildup looks excessive even against the Dot-com era. He compared the current AI boom with the 1999 TMT bubble, pushing back on the idea that this cycle is too mature or too profitable to be judged by the same standards.

That comparison matters because it shifts the conversation from hype to structure. A bubble is not just about high stock prices. It is also about how many parts of the financial system become dependent on one narrative holding up.

Why Burry sees a Dot-com bubble comparison

Burry pointed to a setup where AI is pulling in capital across multiple layers of the market. In his framing, that undercuts the popular defense that the current wave is safer because many companies involved already have real businesses outside AI.

Instead, he suggested the breadth of AI-linked financing makes the Dot-com bubble comparison harder to dismiss. The resemblance, in his view, is not only about investor excitement. It is about how deeply the theme has spread into funding markets.

The funding and debt numbers behind the AI sector overconcentration

The most striking part of the Michael Burry AI bubble warning was the data behind it.

Citing Torsten Slok of Apollo, Burry said 87% of venture capital funding is now directed at AI. He contrasted that with 1999, when the share tied to internet companies was below 40%.

He also said 38% of high-yield bond issuance is linked to AI, while 49% of investment-grade debt issuance is tied to the sector.

Those figures form the backbone of his argument. They suggest AI is not just attracting equity enthusiasm. It is also absorbing an outsized share of corporate financing, including debt that reaches well beyond startup investors.

VC, high-yield, and investment-grade issuance

Taken together, the numbers point to an AI sector overconcentration problem rather than a normal growth cycle.

• 87% of venture capital funding is directed at AI, according to Slok data cited by Burry
• 38% of high-yield bond issuance is linked to AI
• 49% of investment-grade debt issuance is linked to AI

Why this matters: when a single theme dominates both venture funding and debt markets, the risks spread far beyond speculative traders. That can pull in institutions that usually sit farther from the edge of market excitement.

Why Burry says investors should worry

Burry argued that the current setup could put pension funds and other investors at risk. That is a key part of his warning. He is not describing a narrow corner of the market. He is describing exposure that may be embedded in vehicles widely seen as conservative or long-term.

He tied that concern to the Dot-com era’s credit damage. The article notes that by 2002, some $100 billion of investment-grade bonds issued during that earlier boom had turned into junk. In other words, assets that started with a safer label did not stay safe once the underlying story broke down.

That historical comparison gives the warning more force. If AI-related debt now makes up such a large share of issuance, the question is not only whether enthusiasm fades. It is whether the financing built around that enthusiasm remains sound if revenue expectations disappoint.

Pension funds and AI capex

Another pressure point in the article is the economics of the AI buildout itself. Confidence in the AI revolution, it says, has weakened as capital expenditure keeps rising while actual return on investment remains unclear.

That gap is central to the market debate. AI bulls have embraced giant spending on chips, data centers, and related infrastructure. But if reporting makes revenue hard to isolate, investors are left judging a spending wave without a clean measure of payoff.

The article also describes OpenAI and Anthropic as heavily subsidized and unprofitable, while noting that compute demand is concentrated in their hands. That creates an awkward loop: some of the biggest buyers supporting AI-related revenue are private companies that, in this framing, still rely on subsidies and have not reached profitability.

Why this matters: concentration can make growth look stronger than it is. If a large slice of demand comes from a small group of subsidized players, then headline AI revenue may not tell the full story about the durability of the boom.

The infrastructure argument is getting tougher

For a while, one of the strongest defenses of massive AI spending was the Dot-com analogy itself. The argument went like this: even if the first wave overspends, the infrastructure eventually becomes essential, just as internet-era buildouts later proved valuable.

Burry’s warning, as reflected in the article, pushes back on that. The piece argues that GPUs and CPUs may become obsolete faster than the fiber-optic infrastructure built during the Dot-com era. That difference matters because fast-depreciating hardware has a much shorter window to justify its cost.

If that view is right, the AI buildout has less room for error. Equipment bought in the mid-2020s may need to be used intensely and quickly, or it risks losing value within a few years. That is a much harsher economic test than infrastructure that can remain useful for far longer.

The same article notes that this rapid obsolescence does not apply in the same way to energy infrastructure for data centers. Still, the broader point remains: high spending alone does not guarantee lasting value, especially if completion timelines slip and the revenue side remains difficult to pin down.

What the Michael Burry AI bubble warning means beyond Nvidia

The current market mood has often been reduced to one question: can Nvidia keep delivering numbers strong enough to justify the AI trade? Burry’s argument is wider than that, and arguably more unsettling.

This is not just a Nvidia earnings warning story. It is a warning about how one theme can dominate venture capital, high-yield debt, investment-grade issuance, and institutional portfolios at the same time. Once that happens, the market is no longer betting only on innovation. It is betting on the funding structure behind that innovation holding together.

That is why the Michael Burry AI bubble comparison is drawing attention. It taps into a deeper concern now surfacing across markets: not whether AI is real, but whether the financial system is pricing its future too confidently, too broadly, and with too little margin for disappointment.