AivexaNewsSearch
AI news for builders and product teamsChecked every hour

It remains “very difficult [to find] actual, materially measurable financial impacts” from spending on AI, per new study

Collected Oct 14, 2026

A new study concludes that finding actual, materially measurable financial impacts from spending on AI remains very difficult, according to a report by Laura Bratton at The Information, cited in an emailed newsletter from Gary Marcus sent from New Delhi. The study itself was not directly linked in the newsletter.

The newsletter contrasts the minimal documentable return on investment with the trillions of dollars invested in AI, calling the juxtaposition perverse and often baffling. It recalls early 2023 expectations that ChatGPT would multiply every employee's productivity tenfold and radically lift the GDPs of entire countries, and states there is currently zero evidence of anything like that happening at scale.

Two further items were flagged in the same newsletter. Character.AI, described as an early generative AI darling, has filed for bankruptcy. CoreWeave bonds have reached 15%, which the newsletter reads as a sign that credit markets have had enough.

For developers and product teams, this is a reminder that adoption narratives and measured outcomes can diverge. The study's finding is about aggregate, measurable financial impact rather than about whether individual tools work; teams evaluating AI features may face harder questions about attributable returns. The Character.AI and CoreWeave items add a second signal: capital conditions for AI companies may be tightening. Treat both as reported indicators, with the study's methodology and scope not detailed here.

Why it matters: if the study holds, companies investing heavily in AI will be asked to show concrete financial returns, not just deployment. That may slow budgets for experimental projects and shift scrutiny toward measurability. Character.AI's filing and CoreWeave's bond yield, if confirmed, suggest financing conditions are becoming less forgiving for AI firms, an inference from the newsletter's framing rather than a stated conclusion.

Read at Gary Marcus

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Yet another study pointing in the same direction