AI bubble talk is everywhere. Scroll X, read op-eds, listen to panels, and you hear the same refrain: too much hype, too much capital, too little revenue.
And yet, at the Cerebral Valley conference, something telling happened. When asked who thought we were in a bubble, most hands went up. Then host Eric Newcomer asked a second question: who plans to invest less in 2026? Not a single hand moved.
A seeming contradiction. But investors are not confused. They know many AI companies are priced at extreme valuations with little or no revenue. For them, the most costly mistake is not overpaying. It is sitting out and missing the upside. Even partial success against the opportunity they see could justify companies of enormous scale.
The numbers make this clear. Global spend on software today is roughly $300 billion per year. Public software companies account for about $2.2 trillion in market capitalization. These are large numbers, but they are still bounded.
Now compare that to labor. US employers spend on the order of $16 trillion per year on labor compensation. One country. One input. Software does not need to become a $16 trillion industry to be transformative. It only needs a credible path to sell into, and capture a small slice of, that labor spend. That is the opportunity investors are salivating over.
And that is because AI offers the possibility of changing what software means and what it can do.
For decades, software mostly digitized work rather than performing it. It turned analog filing cabinets into digital databases, but humans still did the job. CRMs, ERPs, payroll systems, and EHRs stored information more cleanly, but execution remained human. As Alex Rampell has put it, software watched work happen.
AI introduces a real break. Software can now act, decide, and execute. It shifts from being a system of record to a system of action. This is not incremental efficiency. It is a fundamental rewrite of what software is allowed to do.
AI creates a plausible path to replace slices of labor directly. When software can do the work itself, entire categories flip economically. Companies that move from selling software for the work to doing the work itself gain access to markets an order of magnitude larger. Markets that were never big enough to support venture-scale outcomes become legible for the first time.
To be clear, none of this means the world has already changed. It has not. Reliability is uneven. Trust is fragile. Many pilots will fail. Large portions of labor will never be replaced. The transition will be slow, messy, and asymmetric.
But investors are not betting on solved problems. They are betting on potential and direction. Even partial success opens markets far larger than SaaS ever reached. When the prize is labor rather than software, the upside overwhelms the risk.
That is why bubble talk can reach an all-time high while capital keeps pouring in. The checks are not chasing today’s revenue. They are chasing the possibility that software can finally do the work.