A prominent voice in the AI community is warning that market enthusiasm may be outrunning technological reality, warning that today’s valuations reflect faith more than fundamentals.
In a new BNN Bloomberg interview, Gary Marcus, a cognitive scientist and long-time critic of pure-scaling AI approaches, says investors have entered belief-driven territory at a time when companies are still far from a real breakthrough.
Marcus points to steep capital commitments, hardware intensity and persistent reliability issues as signs that the industry is leaning on expectations that may take years or decades to play out.
“It is a bubble. The question is, when will the bubble pop? The valuations just don’t make sense in terms of the income that these companies are making and the enormous amounts of infrastructure that they require. It’s just not economically sensible. It’s all driven by hype and a kind of belief in a false god of artificial general intelligence that could do anything, but is actually pretty far from what we know how to make right now.”
Marcus notes that the mismatch between economic output and capital inflows is widening, citing the scale of investment required to build frontier AI systems and the lack of clarity on long-term monetization paths.
“You have companies like OpenAI losing a few billion dollars every quarter, and its latest valuation is half a trillion dollars, and it’s made commitments for over a trillion dollars. It just doesn’t really make sense.”
He says the challenges are not only financial. Marcus has long contended that current large language models face structural technical constraints that cannot be resolved simply by scaling compute or data. He compares present-day optimism about AI to early promises around fully autonomous driving, noting that timelines repeatedly slid as real-world complexity mounted and commercial deployment lagged headline expectations.
“The things that we have now are these black boxes that nobody knows how to debug. Nobody really knows how to interpret. And everybody’s just kind of hoping for the best. It’s a little bit like Elon Musk keeps promising that driverless cars will take us from New York to Los Angeles next year. and it never materializes.”
He adds that while he believes artificial general intelligence will eventually arrive, the horizon for trustworthy systems capable of operating autonomously across domains remains much further out than today’s market signals suggest.
“Artificial general intelligence that you could trust to do anything you want is probably at least 10 years away. It might be 20 or 30 years away.”
Disclaimer: Opinions expressed at CapitalAI Daily are not investment advice. Investors should do their own due diligence before making any decisions involving securities, cryptocurrencies, or digital assets. Your transfers and trades are at your own risk, and any losses you may incur are your responsibility. CapitalAI Daily does not recommend the buying or selling of any assets, nor is CapitalAI Daily an investment advisor. See our Editorial Standards and Terms of Use.