As Anthropic’s IPO filing puts fresh numbers behind the AI industry’s growth story, a parallel debate has been building among economists and bank executives: not whether artificial intelligence will eventually pay off, but whether the money being poured into it now will be rewarded on anything close to the timeline investors are pricing in.
Goldman Sachs economist Jan Hatzius estimates the AI growth impulse, the boost the technology is giving to broader economic output, will peak in late 2026 or early 2027, a relatively near-term ceiling given how much capital has been committed to infrastructure meant to pay off over many years. Apollo’s Torsten Slok has made a related point about where the gains are actually showing up, arguing that productivity improvements remain concentrated in a handful of technology stocks rather than spreading broadly, and that the real verdict on the AI boom will be decided in the hundreds of companies outside the small group of mega-cap tech leaders, not within them.
The data backing that caution is specific. Only 6% of companies report AI delivering a significant bottom-line impact, defined as a 5% or greater contribution to earnings before interest and taxes, a figure that has stayed essentially flat year over year despite the scale of investment. Separately, just 2% of US households were paying for AI services as of April 2026, and a Yale CEO summit found 40% of chief executives acknowledging their AI investment has been inefficient so far.
Worker-level data complicates the picture further. Only 26% of workers report that AI meaningfully improves their productivity, while 44% of people who use AI tools daily say the technology actually adds friction to their work rather than removing it. Those numbers sit awkwardly next to the scale of infrastructure spending underway, raising the question of whether current investment levels are being matched by near-term productivity returns anywhere outside a narrow set of use cases.
Bank executives, for their part, remain publicly bullish on the longer arc even while acknowledging the near-term mismatch. Goldman Sachs CEO David Solomon expects capital investment in AI to continue regardless of short-term skepticism, and JPMorgan's Jamie Dimon has predicted AI adoption will move faster than prior technological transformations once it takes hold. BlackRock's Larry Fink has raised a different concern, flagging the risk of white-collar job displacement as adoption accelerates, a reminder that the debate isn't only about returns on capital but about how disruptive the transition could be once productivity gains do show up.
The throughline across these competing views is less a disagreement about AI's eventual value than a disagreement about sequencing: whether the infrastructure being built now will find paying use cases before the capital markets that financed it lose patience. With Anthropic's IPO filing putting real revenue and profit numbers in front of investors for the first time, that question is about to get a much more direct market test than analyst estimates alone can provide.
