The Fed’s Upcoming Productivity Bet

In 1995, Fed Chair Alan Greenspan made an unconventional bet that the U.S. was undergoing an era of faster productivity growth based on the adoption of technology. The decision enabled a significant shift in monetary policy that resulted in faster non-inflationary growth and increased prosperity. The adoption of easier monetary policy also fueled the Dot-Com Bubble.

The boom of the late 1990s earned Greenspan the moniker of the “maestro”. Some even speculated that the Fed may have abolished the ups and downs of the business cycle. However, Greenspan cautioned in an October 2002 speech that “long-term productivity optimism may currently seem a bit out of place”.
 

 

As Trump appoints a new Chair at the Fed and his allies tilt monetary policy to a more dovish direction, a new Trump-dominated Fed is likely to make a Greenspan-sized bet on AI productivity.

What does that mean for investors?
 

 

The Productivity Bull Case

I addressed the AI productivity question in a past publication (see AI Productivity and the Promised Land) but the issue is worth revisiting. Measuring productivity growth is always difficult, as the data tends to be noisy.

Here is the bull case for AI-driven productivity.

First, a St. Louis Fed study found that the adoption of generative AI outpaces the adoption rate of other major transformational technologies, the PC and the internet. The study “estimated that generative AI could plausibly grow labor productivity by between 0.1% and 0.9% at current levels of usage”.
 

 

FactSet reported mentions of AI are growing rapidly in the earnings calls of S&P 500 companies, though the degree of attention can be a two-edged sword. We can recall that during the Dot-Com Bubble, even mining companies felt compelled to discuss their “broadband strategy” in investor presentations.
 

 

Goldman Sachs found evidence that AI appears to be enhancing productivity by substituting capital for labour. Companies that mentioned AI in the workforce discussions were also cutting jobs.
 

 

Greg Ip at the WSJ called AI “The Most Joyless Tech Revolution Ever: AI Is Making Us Rich and Unhappy”. That’s because of the effect of AI on the employment market. During the Dot-Comb Bubble, “demand for digitally savvy workers was off the charts”. Fast forward to 2025, “the optimism is largely confined to AI architects and gimlet-eyed executives calculating how much AI can reduce head count while workers wonder whether they will be replaced by AI”.

A St. Louis Fed study confirmed these findings and concluded that “generative AI can target cognitive tasks performed by knowledge workers”.
 

 

A weak employment market? A transformational technology that promises productivity improvement? That’s the perfect combination for the doves at the Fed to argue for an easier monetary policy, and to make a Greenspan-sized productivity bet because r-star, or the natural rate of interest, has fallen.
 

 

The Productivity Bear Case

One of the arguments that the bears bring up is productivity gains from generative AI may be limited. Generative AI, or large language models (LLMs), gather large amounts of linguistic data by scraping the internet to find relationships between words, or “tokens”, to predict what output should follow. In other words, it’s a highly advanced spellcheck algorithm.

There is a school of thought among leading AI scientists that LLMs don’t think. They predict what might come next in a discussion. While that’s a useful tool, much like spreadsheets are useful tools, LLMs aren’t paradigm changing technologies. To be sure, AI advances have surpassed human abilities in closed-form games like chess, and some LLMs have been able to pass bar exams, but do you want one representing you in court?

A widely quoted study by the research organization METR tested software developers on 246 distinct tasks. Some developers were allowed the use of AI tools and others were not. Developers allowed the use of AI forecast shorter task completion times, but in actuality ended up with longer development times than the non-AI developers.
 

 

However, software productivity studies are very confusing. A different METR study found that the ability of generalist AI models to complete software development tasks at 50% reliability has doubled every seven months. The researchers concluded: “Extrapolating this trend predicts that, in under a decade, we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days or weeks.”
 

 

 

The Verdict

In the end, the jury is out on the bull and bear question. Much depends on the usefulness of AI across different applications and industries. The truth probably lies somewhere in between the two extremes. Technology adoption and penetration usually takes a lot longer than the consensus view. The effects of PC and internet adoption didn’t emerge until at least a decade after the initial introduction of the technology. The best-case analysis of productivity increase can be seen in the early 2000s, well after the internet came into common use (red line). Extrapolating the slope of the productivity gains from that episode to the PC and, today, AI (dotted red lines) seems ambitious.
 

 

From the market’s perspective, the preliminary verdict is that the Fed will undertake an inflationary outcome. Yield curve spreads are undergoing mild steepening reactions, which is discounting higher inflationary expectations in the future.
 

 

These results are consistent with our long-term observation of a relative breakout in gold against both the S&P 500 and 60/40 portfolio, which is a signal that the market is undergoing a shift from paper to hard asset leadership. In particular, gold’s relative breakout against the 60/40 portfolio is an indication of the reduced diversification effect of bonds against stocks. The addition of gold, in a measured way, to a balanced fund portfolio is likely to reduce overall portfolio volatility.
 

 

In conclusion, a Trump-dominated Fed is on the verge of a rate cutting cycle based on a probable Greenspan-style bet on AI-driven productivity. If AI does significantly boost productivity, the economy could be in for a period of non-inflationary growth and prosperity. The risk is a policy error, rising inflation and a falling USD. Much depends on the usefulness of AI across different applications and industries.

From the market’s perspective, the preliminary verdict is that the Fed will undertake an inflationary outcome. Yield curve spreads are undergoing mild steepening reactions, which is discounting higher inflationary expectations in the future. Regardless of the outcome of the productivity debate, we believe strength in the gold price is consistent with our long-term observation of a relative breakout in gold against both the S&P 500 and 60/40 portfolio, which is a signal that the market is undergoing a shift from paper to hard asset leadership.