Central bankers are currently high on a specific brand of economic hopium. They are whispering about a "new era" where generative AI boosts productivity so drastically that the neutral rate of interest—that mythical $r^*$ where the economy neither accelerates nor drags—must naturally climb. The logic is lazy: AI makes us efficient, efficiency drives investment demand, and more demand for capital means higher rates.
It is a neat, academic fantasy. It is also fundamentally wrong.
I have spent two decades watching Silicon Valley "disrupt" industries only to leave them more deflated and centralized than before. If you think a spreadsheet-driven productivity bump is going to magically save the Fed from the gravity of low rates, you aren't paying attention to how software actually eats the world. AI isn't a bridge to higher rates; it is a high-speed elevator to a deflationary basement.
The Myth of the Capital Investment Boom
The "higher neutral rate" crowd relies on the idea that companies will engage in a massive, sustained capital expenditure (CapEx) spree to integrate AI. They point to the $100 billion data centers being planned by Microsoft and OpenAI as proof.
They are looking at the wrong side of the ledger.
In a traditional industrial revolution, like the steam engine or the internal combustion engine, you needed massive physical infrastructure. You needed steel, rubber, oil, and millions of laborers. That required heavy borrowing over decades, which pushed up the demand for capital and, consequently, the neutral rate.
AI is different. It is a capital-light, software-based force. Once the initial infrastructure (the "training" phase) is built, the "inference" phase—where the actual work happens—becomes exponentially cheaper. We are already seeing the cost of tokens drop by 90% year-over-year.
When a company replaces a 50-person customer service department with a single API call, they aren't taking out a $10 million loan to build a factory. They are cutting their budget. They are reducing their need for external financing. They are hoarding cash, not borrowing it. Look at Big Tech’s balance sheets; they are essentially giant hedge funds with a software business attached. They don't need the banks. If the biggest drivers of the economy don't need to borrow, the neutral rate cannot rise.
Efficiency is a Deflationary Weapon
The Fed officials talking about a productivity boost are ignoring the "Value Capture Paradox."
When an industry becomes hyper-efficient via software, the gains rarely stay with the workers or get reinvested into new, labor-heavy ventures. Instead, the gains are passed to the consumer in the form of lower prices or swallowed by the platform owner as profit.
Consider the "Staccato Effect" of AI implementation:
- The Displacement: 30% of administrative tasks are automated.
- The Margin Expansion: The company’s bottom line looks great for three quarters.
- The Price War: Competitors use the same AI to undercut prices.
- The Deflationary Spiral: The total dollar value of the sector shrinks.
If AI does its job, it makes things cheaper. If things get cheaper, inflation stays low. If inflation stays low and the world is awash in efficiency-driven excess capacity, there is zero pressure on the Fed to maintain high real rates. The "neutral" rate is forced down by the sheer weight of falling costs.
The Labor Force Fallacy
There is a common argument that AI will "augment" labor, making every worker a "super-worker," thereby increasing wages and consumption.
I’ve seen this play out in the SaaS (Software as a Service) world for fifteen years. Efficiency doesn't lead to higher wages for the many; it leads to "Winner-Take-Most" dynamics. When one person can do the work of ten, you don't pay that person ten times the salary. You pay them 1.5 times the salary and let the other nine go.
This creates a massive "Output Gap." The economy produces more goods and services with fewer people. This is a classic recipe for a lower neutral rate. For $r^*$ to rise, you need a shortage of capital and an abundance of profitable places to put it. AI creates an abundance of capital (through corporate savings) and a shortage of places to put it that aren't already dominated by an AI-first incumbent.
The Ghost of the 1990s
Economists are obsessed with the 1996-2004 period. They remember how the internet boosted productivity and allowed the Fed to keep rates higher than they were in the early 90s. They are trying to map that template onto 2024 and beyond.
But the internet required physical wires, towers, and hardware in every home. It created new markets (E-commerce, Social Media) that didn't exist before.
AI, in its current form, is largely an optimization engine for existing markets. It’s making coding faster, legal research quicker, and marketing copy more generic. It is an incremental improvement on the digital plumbing we already have. It doesn't require a new "New Deal" of infrastructure. It requires a subscription to a model hosted in a warehouse in Virginia.
Why the Fed is Setting a Trap for Itself
By talking up the neutral rate now, Fed officials are trying to justify keeping "higher for longer" as a policy. They want a reason to stay restrictive without admitting they are terrified of a 1970s-style inflation resurgence.
They are using AI as a convenient ghost to explain away why the economy hasn't crashed yet under 5% rates. "It must be the productivity!" they cry.
The reality is much simpler: the lag of monetary policy is real, and the massive fiscal spending of the last four years hasn't fully drained out of the system. We aren't seeing an "AI productivity miracle" yet; we are seeing the final fumes of a $5 trillion stimulus package.
If the Fed hikes—or refuses to cut—based on the idea that AI has permanently raised $r^*$, they will overtighten into a deflationary void. They will be fighting a fire with a firehose while the basement is already flooding.
The Real $r^*$ is Lower Than You Think
Imagine a scenario where AI reaches a point of "Autonomy."
- Company A needs to refresh its entire marketing strategy.
- Instead of hiring an agency for $500,000, it runs an internal agentic workflow for $50 in compute costs.
- That $499,950 disappears from the "Demand for Capital" side of the macro equation.
Where does that money go? It sits in a corporate treasury or goes into a stock buyback. Both actions put downward pressure on interest rates.
The "Neutral Rate" is a function of the balance between savings and investment. AI increases savings (profits) and decreases the cost of investment (software is cheaper than hardware).
$$r^* = \text{Balance of Savings and Investment}$$
In an AI-dominated world, the supply of savings goes up, and the cost of the most productive investments goes down. Basic economics tells you the price of money (interest rates) must fall.
Stop Listening to the Optimists
The industry insiders telling you that AI will lead to a 4% neutral rate are usually the ones trying to sell you a "Growth" fund. They need the narrative of a high-growth, high-rate environment to keep the "Roaring 20s" myth alive.
But if you look at the mechanics of the technology, it is a giant vacuum cleaner for costs. It sucks the "premium" out of labor and the "moat" out of many businesses. It is the most potent deflationary force ever invented.
The Fed won't find salvation in Silicon Valley. They will find a black hole that swallows their ability to maintain high rates. The "higher neutral rate" isn't a new reality; it’s a temporary hallucination caused by a misunderstanding of how software scales.
Prepare for a world where money remains cheap because the machines have made everything else even cheaper. The Fed is planning for a boom; they should be bracing for the Big Squeeze.
Go back and look at the "Long Transformation" of the 20th century. Every time we found a way to do more with less, the cost of capital eventually tanked. AI is the ultimate "more with less" machine. To suggest it will do anything other than crater interest rates in the long run is to ignore the history of every technology we’ve ever built.
Burn your 1990s playbook. The machines aren't here to help the Fed; they're here to make the Fed's current toolkit obsolete.
