LeCun Warns of a 'Big Bubble Explosion' in AI
Yann LeCun told CNBC that AI labs like OpenAI and Anthropic risk a 'big bubble explosion' because soaring prices still don't cover their operating costs.
Yann LeCun, Meta's former chief AI scientist and one of the field's most influential figures, told CNBC that AI labs such as OpenAI and Anthropic are heading for a 'big bubble explosion.' His argument is about economics, not capability: the prices these companies charge for AI keep climbing, yet their operating costs aren't falling fast enough to match, so the businesses lose money while investors effectively subsidize every query.
It's a pointed claim, and LeCun has skin in the game. His own startup, AMI Labs, raised $1 billion in March in what was described as Europe's largest-ever seed round, and it's pursuing a different bet entirely: 'world models' — systems built to understand physical reality — rather than the large language models that power today's leading chatbots. So his warning doubles as a pitch for the road he's chosen.
He isn't the only one nervous about the math
The cost concern isn't fringe. OpenAI's own CEO, Sam Altman, recently called AI costs 'a huge issue' for businesses, and the broader industry has been raising prices, shifting to usage-based billing and burning enormous sums to fund data centers. LeCun also took aim at a competitor, calling Elon Musk's xAI 'a kind of failure,' pointing to founder departures and trouble recruiting top talent, and saying he doesn't expect it to keep pace with OpenAI or Anthropic. Whether or not you buy the 'bubble' framing, the underlying tension — revenue that doesn't yet cover the cost of serving AI at scale — is real and widely acknowledged.
Why this matters for you
Subsidized pricing is great while it lasts: a lot of what you use today is cheaper than it actually costs to run. The flip side is that if the economics don't improve, some of that gets corrected later through higher prices, tighter usage limits, or tools that quietly disappear. None of this means the technology is going away — even bubbles that burst tend to leave the useful infrastructure behind. But it's a reason not to wire a critical workflow to a single cheap plan that may not stay cheap. My take: treat today's prices as promotional, and keep your setup portable enough to switch.
Don't panic, but don't assume today's pricing is permanent either. If you depend on an AI tool for real work, know what you'd do if its price doubled or its limits tightened — ideally keep your prompts and workflows portable across more than one provider. Bubble or not, the cheapest plan today is the one most likely to change; building in an exit keeps you in control.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles