AI stocks surge on expectations, but face ‘token mirage’ concerns

A trader's critique of the $8 trillion AI equity rally suggests that inflated usage metrics could be masking weaker demand than Wall Street wants to believe.

Add us on Google by Editorial Team Jun. 1, 2026

The AI trade has been the gift that keeps on giving. Stocks tied to artificial intelligence have powered the S&P 500 to all-time highs, minted trillion-dollar market caps like participation trophies, and convinced an entire generation of investors that we’re living through the next industrial revolution. But what if the metrics propping up this rally are, to put it politely, misleading?

That’s the argument trader Kevin Muir laid out in a June 1 Substack post that takes a scalpel to the $8 trillion rally in AI-related equities. His thesis: the growth in reported AI usage, often measured in tokens processed by large language models, could be artificially inflating perceived demand. The result is something he frames as a kind of measurement illusion, where rising token counts look like booming adoption but may not translate into actual productivity or durable revenue.

The token mirage, explained

Here’s the thing about tokens. In the context of large language models, a token is a chunk of text, roughly three-quarters of a word, that the model processes. When companies report surging token usage, it sounds impressive. More tokens equals more AI activity equals more demand for chips, cloud compute, and everything in between.

But Muir’s point is subtler than just calling it hype. Corporate AI policies are driving employees to interact with AI tools in ways that inflate these numbers without necessarily creating proportional value. In English: companies are using more AI, but “more” might mean more tokens burned, not more problems solved.

Nvidia and the valuation question

No discussion of the AI trade is complete without Nvidia, the company that essentially became the arms dealer of the AI gold rush. As of late May 2026, Nvidia trades at 33 times earnings with a market cap of $5.2 trillion. For context, that’s larger than the GDP of Japan.

Hyperscaler capital expenditure is projected to exceed $700 billion in fiscal year 2026. That’s an astonishing figure, representing a bet by the likes of Microsoft, Google, Amazon, and Meta that AI infrastructure will be the defining investment of the decade. Muir’s concern is that this capex cycle is being justified by token-based usage metrics that may overstate genuine demand.

Echoes of past hype cycles

Muir’s analysis draws parallels to previous technology hype cycles, where inflated metrics and speculative enthusiasm drove valuations far beyond what fundamentals could support. The dot-com era had its own version of the token mirage: eyeballs. Websites were valued based on page views and unique visitors, metrics that looked like growth but often masked the absence of a viable business model.

Muir isn’t alone in his skepticism. Michael Burry, the investor made famous by “The Big Short” for correctly calling the 2008 housing crisis, has taken bearish positions in semiconductor stocks linked to AI.

What this means for investors

The core risk here isn’t that AI is fake or that the technology doesn’t work. The risk is a valuation gap between what the market is pricing in and what companies will actually deliver in terms of productivity gains and revenue growth.

If hyperscalers are spending $700 billion-plus on AI infrastructure, the returns on that investment need to be proportionally massive to justify current stock prices. If those returns materialize slowly, or if token usage growth proves to be a poor proxy for actual business value, then the stocks trading at 30-plus times earnings are vulnerable to a repricing.

For crypto investors, this matters too, even though Muir’s analysis contains no references to crypto-native tokens or projects. The AI narrative has spilled across asset classes, lifting AI-adjacent crypto tokens and projects that promise to decentralize compute or tokenize AI services.

Investors would be wise to scrutinize the specific metrics companies use to report AI adoption. Token counts, API calls, and model queries are activity metrics, not value metrics. The difference between the two is where the mirage lives.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.