Why AI Might Drive Interest Rates Down: A Contrarian Perspective

Thesis: Advanced AI could push interest rates significantly lower over the next decade, not higher. This contrarian view runs against market consensus and recent economic models, but my reasoning is straightforward: AI is likely to create mass unemployment before it delivers mass prosperity, forcing governments to engage in financial repression (keeping interest rates below inflation) to manage an otherwise impossible economic transition.

In other words, as AI automation displaces millions of workers and pressures government finances, policymakers will have strong incentives to push interest rates down in real terms – even if economic theory suggests rates “should” rise. Below, I’ll lay out why I expect lower rates, discuss the standard argument for higher rates (and why I think it’s wrong for this scenario), and sketch how a regime of financial repression could unfold.

Why I Expect Lower Rates in an AI-Driven Economy

The unemployment shock will come before the productivity boom. AI can eliminate jobs instantly, but it takes time for the productivity gains to filter through the economy. For example, a truck driver could lose their job the same year self-driving trucks become viable, whereas the total productivity boost from autonomous logistics might take years to materialize. In the interim, millions of unemployed or underemployed workers would cut their spending, leading to a collapse in consumer demand. With consumers buying less, businesses face gluts and will cut back investment despite the new technological opportunities. In short, a sudden AI-driven unemployment wave is deeply deflationary: less spending and investing mean downward pressure on prices and economic output. If loss of jobs happen abruptly, the shock to demand would overwhelm any short-term “investment boom” from AI tech itself. This deflationary backdrop puts downward pressure on interest rates, since in a weak economy with high joblessness, there’s little appetite to borrow for expansion and strong political pressure to make credit cheap.
Government debt math leaves no room for high rates. The U.S. government is entering this AI era with debt around 120% of GDP, the highest level in history outside of World War II. Servicing that debt already costs on the order of $900 billion per year (and projected to hit$ 1 trillion by 2025–2026). Every percentage point increase in average interest rates would add hundreds of billions more to annual interest costs. Simple arithmetic: 1% of $30+ trillion in debt is about$ 300 billion in added interest. Sustained 5% real interest rates on U.S. debt would eventually mean ~$1.5–1.8 trillion in interest expenses – an untenable burden that would dwarf most other federal spending. At a 10% rate, debt dynamics spiral into an immediate crisis of sovereign default risk. In short, the government literally cannot afford to let rates rise much without defaulting or entering a debt death spiral. Policymakers know this, and as debt costs climb, the incentive to suppress interest rates will become overwhelming. No matter what textbook models say, a democracy with such debt levels will find a way to keep rates low rather than commit political suicide via austerity or default.
Politics will demand intervention (cheap money) in an AI jobs crisis. Imagine unemployment surging to 15% or 20% because of rapid AI automation. In a democracy, that kind of mass joblessness creates irresistible pressure for government action. Voters would demand relief: think Universal Basic Income (UBI), expanded unemployment benefits, public jobs programs, maybe even shorter workweeks or job guarantees. These social programs cost trillions, and funding them during a recessionary, high-unemployment period is tricky. The path of least resistance will be financing massive support programs through debt and money creation, while at the same time preventing borrowing costs from exploding. Additionally, there will be a strong political urge to “do something” about the situation: we might see proposals for AI taxes (to redistribute from tech firms to displaced workers), strict job protection laws, or other interventions to slow down automation. All of this points toward a much bigger role for government in the economy. To enable that without breaking the bank, central banks and regulators would likely work to cap interest rates and generate some inflation. Financial repression is essentially a silent way to tax the private sector: by keeping interest rates below inflation, the government can pay for social support and gradually erode its debt, at the cost of savers. In a severe AI-driven downturn, no elected officials or central bankers will be willing to tell tens of millions of angry citizens that “markets need 30% interest rates” – they will instead find ways to keep rates low, damn the torpedoes.

Taken together, these points suggest that interest rates will be forced down (in real terms) as the AI revolution unfolds. High unemployment and massive debt create a scenario where the central bank and government must prioritize low rates, even if that means tolerating higher inflation. This is essentially financial repression, which I’ll detail later. First, though, let’s address the opposing viewpoint: many economists argue that transformational AI should cause interest rates to rise, not fall. Why do they say that, and why do I still think they’re wrong?

The Conventional View: Why Many Expect Higher Rates

It’s important to understand the mainstream argument that AI will push interest rates up. The best exposition of this view comes from a recent paper by Chow, Halperin, and Mazlish (2025), who make three main points:

Consumption smoothing and explosive growth: If AI is about to deliver, say, 30% annual economic growth in a few years, standard economics says people should start borrowing from the future aggressively. In technical terms, the Euler equation implies interest rates rise with expected growth. Intuitively, if we expect to be much richer next year thanks to AI, we’d want to spend more now and pay it back later. Households would borrow and investors would demand higher returns for lending to these future-rich consumers. Similarly, even if one fears AI could cause an existential catastrophe (unaligned AI scenarios), the logic is to spend now while you can, which also drives up interest rates as everyone tries to borrow against a future that might not exist. In short, “consumption smoothing” behavior in face of transformative AI would, according to these models, cause a spike in real interest rates.

Historical precedent – productivity booms raise rates: Every major technological or productivity revolution in the past has eventually been accompanied by higher real interest rates. For example, during the Industrial Revolution, as investment opportunities exploded, the demand for capital pushed interest rates upward (at least in many analyses). After World War II, the massive post-war economic boom saw interest rates rise from the ultra-low wartime levels. More recently, the late 1990s tech boom coincided with real U.S. interest rates above 4%. The logic is that when the economy offers abundant high-return projects (like railroads, factories, or Internet startups), capital is in high demand and lenders can charge more. Many assume the AI revolution will be similar: a wave of highly productive investments that will compete for funding, driving rates up.

Market efficiency and investor behavior: If AI truly transforms the economy, markets should start pricing this in well in advance. Rational investors won’t willingly hold assets with negative real yields if they see that AI-related ventures can earn, say, 30% returns. Capital should flow to the highest-return opportunities. In a free market, this means interest rates on safe assets would rise until they at least roughly compensate for the growth and opportunities elsewhere. In essence, if the future looks incredibly bright (or incredibly risky), investors will adjust: either scenario leads to higher yields on bonds because people either want to borrow now (driving rates up) or they demand a premium to lend (also driving rates up). Under this view, sustained negative real rates would imply that investors are asleep at the wheel – something proponents of market efficiency find unlikely.

At first glance, these arguments are compelling and grounded in solid economic reasoning. Indeed, if we had functioning markets with no frictions, low government debt, and a gradual transition to AI, I would agree that interest rates should rise in anticipation of huge growth. However, I believe this time is different, for several key reasons. The conventional view assumes an idealized world that doesn’t account for the political and social realities we’ll face. Let’s go through why the textbook logic may break down in the scenario of rapid, tumultuous AI-driven change.

Why This Time Might Be Different (Rebuttal)

Political constraints will override the Euler equation. The elegant theory that “rates should rise to ~30% if growth will be 30%” comes with a gigantic caveat: that outcome cannot happen if it would bankrupt the government and half the private sector. The standard models assume things like a rational, omnipotent central bank, no limit on government debt, and smooth market adjustments. In reality, a spike of interest rates to double-digits would trigger an immediate fiscal crisis for high-debt countries (as discussed earlier) and likely a wave of defaults for businesses and households with debt. No democratic government will sit idly and let market interest rates reach levels that guarantee its own insolvency. Instead, policymakers will invoke emergency powers – explicit or implicit – to prevent such a scenario. In other words, when the Euler equation says “rates should be 30%,” politicians and central bankers will say “Nope, we’re capping them at 3%.” Economic theory goes out the window when it collides with political survival. In the past, when governments have faced similar constraints, they intervened rather than “letting markets work” (for example, central banks in the 1940s and 1950s kept rates artificially low despite growth, under political direction).
Unlike past booms, we’re starting in a fragile position. Historical growth spurts (Industrial Revolution, post-WWII, etc.) occurred under very different initial conditions. Governments then had relatively low debt loads (the UK’s debt was on the order of 30% of GDP at the start of the Industrial Revolution, nothing like 120% today), populations were young and growing, and there was no expansive social safety net that everyone expected. Also, past innovations rolled out over decades, giving society time to adapt. By contrast, we enter the AI era with huge debt, aging populations, and an electorate used to the idea that government should cushion major economic blows. There is far less tolerance now for “letting the market run its course” if that means depression-level unemployment or mass impoverishment. Additionally, we’re no longer on a gold standard; governments can print money at will. In the 19th century or under Bretton Woods, policy had stricter constraints and inflation was eventually kept in check by convertibility limits. Today, with fiat currency, there is more flexibility to finance deficits via the printing press – which again points to inflationary debt erosion rather than high real rates. Another difference is global coordination: previous tech booms were mostly national, but a disruptive AI transition will be a worldwide phenomenon. Major economies will likely face similar pressures simultaneously, making it easier (or more necessary) for them to all resort to financial repression without one country’s bond market standing out as an outlier. All these initial conditions tilt toward a response of suppressed interest rates and managed outcomes, rather than a laissez-faire surge in rates that “should” happen in theory.
Markets won’t be free to find equilibrium. In a textbook world, if investors demand higher returns, interest rates rise until equilibrium. But in the scenario we’re considering, authorities simply won’t allow markets to reach that equilibrium if it’s deemed socially destructive. We have numerous tools in the modern playbook to prevent markets from clearing at a high interest rate. For example, central banks can implement yield curve control (pegging certain bond yields at a target and buying unlimited bonds to enforce it). Financial regulators can impose rules on banks and institutional investors to make sure there’s always a bid for government bonds (more on this shortly). In extreme cases, governments can even outlaw certain capital flows or slap on controls to keep money from fleeing to higher-yield opportunities abroad. In short, the free movement of capital and market pricing of bonds can and will be constrained. Historical reference: after World War II, the U.S. and UK kept nominal rates low and achieved negative real interest rates for years through a combination of regulations and inflation. They didn’t call it repression; it was framed as “prudential regulation” and post-war stability measures. We are likely to see a similar story play out in an AI-disruption scenario. When the choice is (a) let markets set interest rates freely and risk sovereign default + 30% unemployment + social unrest, or (b) intervene heavily to cap rates and tolerate some inflation, any sensible government will choose (b) without hesitation.

In summary, the standard argument for higher rates assumes away the very forces that will be dominant: political intervention and crisis-driven constraints. This time is different because AI’s impact could be extraordinarily fast and disruptive, colliding with a leverage-heavy, politically charged economic backdrop. The result I foresee is financial repression – essentially an environment where nominal interest rates are held below inflation for a protracted period. Let’s talk about how exactly governments can achieve that.

The Financial Repression Playbook: How to Keep Rates Low

Financial repression is not a speculative concept; it’s a well-documented strategy governments have used in the past when debt was high. The basic idea is to steer funds from the private sector to the government by engineering a situation where savers get negative real returns. In practice, this means forcing interest rates below the inflation rate, so that each year debtors (like the government) effectively pay back their loans in cheaper currency while creditors (savers) lose purchasing power. How can a government pull this off? There’s a toolkit, often implemented in stages:

In combination, these steps amount to financial repression: a controlled environment where nominal interest rates are kept below inflation through policy and regulation. For example, imagine inflation is running at 6% but the central bank, via yield curve control, pegs government bond yields at 3%. Savers in bonds (or bank accounts with capped rates) lose ~3% of their purchasing power per year. Over a decade, that’s roughly a 26% loss in real value – effectively, over a quarter of their wealth transferred away, largely to the benefit of debtors (the government and whoever borrowed cheap money). This is essentially a slow-motion partial default on government debt: the debt gets paid back in currency that’s worth much less. Historically, the U.S. used this strategy after WWII (debt-to-GDP was reduced from ~120% in 1946 to ~50% by 1960 in no small part due to growth and negative real rates), and many other indebted countries have done the same. I argue that in an AI-induced crisis of unemployment and debt, a similar pattern is very likely. It’s the only way I can think of to square the impossible equation of massive social spending needs + high debt + avoiding default. You confiscate some wealth silently via negative real rates.

So, I’m betting that interest rates (especially real interest rates) are going down, not up, in the age of AI. In a sense, the market’s signals will be overridden. I’m not saying AI is bad for the economy long-term; actually, eventually AI could make us vastly wealthier. But before that wealth arrives, we’ll go through a dangerous transition, and in that transition, normal market logic gets suspended. By the time AI-driven prosperity is fully realized (if it is), we may already have a very different financial system in place.

In conclusion, those preparing for an AI-driven future should take into account not just the direct technological effects, but the political-economic reactions. The safest prediction in politics is that leaders will do whatever it takes to avoid immediate crisis. High interest rates in a high-debt, high-unemployment scenario spell crisis, so they simply won’t be allowed. The exact mechanisms and timeline can be debated, but the direction seems clear. Interest rates (in real terms) are likely headed down – potentially way down – and staying there, regardless of what standard economic models say they “should” do. Investors and policymakers alike would do well to consider that when planning for the next decade. The age of AI may also be the age of financial repression.