Last Thursday, we participated in our quarterly economic roundtable. Many topics were discussed, but the most interesting takeaway for me was the rapid improvement in mainstream artificial intelligence tools.
According to JP Morgan data covering tools such as Claude, ChatGPT, xAI, and others, AI functionality has increased dramatically over the last 24 months. In 2024, the average amount of human work that could be replicated almost instantaneously by AI was roughly five minutes. Less than two years later, that figure is approaching 70 minutes.
Think about that. Commonly available AI technology can now complete more than an hour of human labor almost instantly. Applied across sectors, companies, and geographies, the potential impact on productivity, margins, and economic growth could be substantial.
Today, the market remains focused on the “picks and shovels” of AI, including semiconductors, data centers, compute infrastructure, and power capacity. It is difficult to know how long the current wave of capital spending will continue, but recent results from the large chip makers suggest we may still be in the early innings, even if some expectations have become aggressive.
What may be more interesting is that the efficiency gains are no longer just theoretical. Oracle, which has a significant backlog tied to compute demand, has faced pressure as investors questioned how it would finance its buildout. By reducing its workforce by approximately 10%, Oracle showed that the efficiencies promised by technology are already influencing corporate decisions. The margin expansion story is not only about the future. In some cases, it is already beginning.
Some will reasonably question what AI means for the labor force. Certain jobs and tasks will be affected. But this is not the first time technology has changed the nature of work. The cotton gin, the Industrial Revolution, the calculator, the Xerox machine, the personal computer, and the internet all displaced certain types of labor, while also increasing productivity and freeing human capital for new uses.
Human beings, when given time and incentive, have an extraordinary record of adaptation and invention. We see no reason AI should be different. By allowing people to do more work faster than ever before, AI may ultimately magnify future innovation rather than diminish it.
The market will continue debating capital expenditures, valuations, infrastructure bottlenecks, and which companies capture the economics of this transition. Those questions matter.
But the larger point may be simpler: if widely available AI tools can already replicate more than an hour of human work almost instantly, the long-term productivity opportunity may still be underestimated.