July 15, 2026
Bubble, trouble, toil and ROI
Speculative Growth and the AI "Bubble" [pdf]
Even if the AI hype pops, commenters say the money spree might still change everything
TLDR: The paper argues that even if AI stocks are overpriced today, the spending boom could still leave behind real machines, tools, and growth that matter later. Commenters split hard between “that happened before and built useful stuff” and “nice theory, but does AI actually earn its keep?”
An MIT economist just dropped a spicy idea into the already overheated AI debate: what if an AI bubble can still do something useful even after the stock-market glitter fades? In plain English, the paper argues that sky-high AI company values can trigger a building frenzy—more chips, more data centers, more tools—and even if investors later sober up, the stuff built during the mania could leave the economy permanently stronger. Translation: the hype might be fake, but the concrete is real.
And the comment section absolutely ran with that. One reader instantly turned the whole paper into the internet’s favorite format, posting a clean-room “tl;dr”: yes, AI company futures may be shaky, but the broader economy may be less doomed than the loudest pessimists claim. Another commenter brought receipts, saying Stripe Press’s Boom basically vibes with this thesis: wild investment manias can leave behind useful infrastructure. But not everyone was buying the elegant theory. One skeptic side-eyed the whole thing with, “This has strong signs of LLM writing,” which is the most 2026 insult imaginable for an AI paper.
Then came the real class-war sting. A commenter winced at the paper’s blunt setup—workers just work, capital owners own everything—calling it “Ouch” and reminding everyone that workers used to own more through pensions. Others questioned the giant assumption underneath all of this: does AI even produce enough real value to justify the spending spree? So the drama line is set: visionary explanation of the boom, or fancy academic cope for a bubble nobody wants to name out loud?
Key Points
- •The paper argues that AI-related valuations should not be viewed only as either justified by fundamentals or as a bubble; temporary overvaluation can still leave a lasting real economic effect.
- •Caballero’s model says AI capital expands productive capacity and shifts income toward high-saving capital owners, which can lower long-run interest rates as capital accumulates.
- •This saving and investment feedback can create multiple steady states, including a low-capital state and a self-sustaining high-capital state.
- •A move from the low-capital state to the high-capital state requires a temporary belief-supported rise in valuations that accelerates investment before learning removes the pricing wedge.
- •If the correction comes after enough capital is installed, workers can end up with higher wages and the economy can remain at a higher-capital equilibrium; if it comes too early, the transition fails.