Something has gone wrong with the people running America's most valuable technology companies. Call it AI psychosis: an affliction characterized by grandiose proclamations, timeline hallucinations, and an apparent inability to distinguish between product roadmaps and prophecy.
The symptoms have become impossible to ignore. In recent months, we've witnessed CEOs predict artificial general intelligence by next Tuesday, promise that their chatbots will cure cancer and solve climate change simultaneously, and suggest—with apparent sincerity—that their language models possess consciousness. The rhetoric has escalated from optimistic to messianic to clinically concerning.
The incentive structure is the illness
The condition isn't mysterious once you examine the underlying pathology. AI companies are valued not on revenue or profit but on narrative momentum. A CEO who says "we're building useful productivity tools" gets a polite valuation. A CEO who says "we're creating superintelligence that will reshape human civilization" gets billions in fresh capital and a CNBC appearance.
This creates a ratchet effect. Each funding round requires grander claims than the last. Competitors must match the hyperbole or appear insufficiently ambitious. The result is an arms race of absurdity, where executives compete to make the most outlandish predictions about technology that, in its current form, still struggles with basic arithmetic and confidently fabricates legal citations.
The market rewards the madness
Investors bear significant responsibility. The same venture capitalists who would laugh a founder out of the room for overpromising on a SaaS product will write nine-figure checks based on fever dreams about machine consciousness. They've created a market where sobriety is punished and mania is rewarded.
Public markets have followed suit. Companies with no clear path to profitability command valuations that assume they'll capture most of global GDP. Analysts who express skepticism are dismissed as lacking vision. The entire ecosystem has developed a tolerance for claims that would, in any other industry, trigger SEC inquiries.
The real victims aren't in the C-suite
The consequences extend beyond embarrassing interviews. Employees are burning out chasing impossible timelines set by executives who've lost touch with engineering reality. Customers are deploying half-baked AI systems based on capabilities that were promised but never delivered. Policymakers are crafting regulations based on science fiction scenarios while ignoring present harms.
Meanwhile, researchers doing careful, incremental work—the kind that actually advances the field—find themselves overshadowed by carnival barkers with better PR teams.
Our take
The cure for AI psychosis is boringly simple: accountability. Start asking CEOs to reconcile last year's predictions with this year's reality. Demand that investors explain their valuation models in terms that don't require believing in magic. Treat AI like every other technology—capable of genuine utility, subject to genuine limitations, worthy of genuine scrutiny rather than religious devotion. The technology is interesting enough without the mythology. The executives might benefit from remembering that.
