Why AI Will Break the Back of Interest Rates, and How We Get Paid Either Way

Every ballplayer who ever hit .250 will tell you the box score does not always tell you what is happening in the game. You can watch a hitter go 0 for 4 and miss that he squared up three balls that found gloves.

The box score says failure. The at bats say something else entirely.

Wall Street is staring at the wrong box score right now. Every strategist on television wants to talk about inflation and the federal deficit, as if those two numbers are the only pitches being thrown. They are not wrong that these things matter.

They are wrong about which ones will decide the game.

I have spent 30 years buying assets that trade below what they are worth and waiting for the market to notice. Patience is the whole business. When I tell you that the next great move in long-term interest rates will be decided by something almost nobody is pricing in today, I am not chasing a headline.

I am reading the at bats, not the box score.

That something is artificial intelligence. Not whether it changes your life, but whether it changes the bond market. Those are very different questions, and only one of them pays you.

Torsten Slok, the chief economist over at Apollo, put out a piece recently that deserves more attention than it got. His argument is elegant in a way that most Wall Street research never bothers to be. He said that long-term Treasury rates could fall substantially within a year or two regardless of whether the AI boom succeeds or fails.

Read that twice. He did not say rates fall if AI wins. He did not say rates fall if AI loses. He said rates fall either way, through two completely different mechanisms that happen to point at the same destination.

That is the kind of asymmetry I have spent a career hunting for in community bank stocks trading below tangible book value. When you find a setup where multiple paths lead to the same profitable outcome, you do not need to be a genius about which path gets taken. You just need to own the right asset before the crowd figures out there was never really a wrong answer.

Picture the favorable branch first. AI delivers on the promise, and companies across the economy start squeezing more output from the same number of people. Banking, insurance, legal work, accounting, customer support, health care administration. All of it labor heavy. All of it ripe for a productivity jump that manufacturing already went through decades ago.

Here is the part that matters for bond investors. Productivity gains are deflationary in the best possible sense. A company that produces more with the same labor and capital does not need to raise prices as fast to protect its margins. Wage increases stop being inflationary because output per worker rises right alongside the paycheck.

Roll that dynamic across enough of the economy and the Federal Reserve gets what it has wanted since 2022 without having to strangle demand to get it. Inflation drifts toward target because supply expanded, not because the central bank crushed spending. Long-term Treasury yields fall because the term premium tied to inflation risk starts to shrink.

Call this the pitcher's strikeout. Clean, satisfying, no controversy. Corporate earnings keep growing. The economy hums along. Rates fall because the future genuinely got cheaper to produce.

Picture the other branch next, the one that keeps a lot of technology executives up at night even if they will never admit it in public. AI implementation turns out to be more expensive, more unreliable and more complicated than the marketing decks suggested. The tools remain useful. They just never become transformative enough to justify the trillions already committed to chips, data centers and electricity infrastructure.

Slok raises the possibility of the Nasdaq falling 50%. Sit with that number for a second. A decline of that size would count as a wealth destruction event, not a mere correction, rippling through household balance sheets, corporate confidence and every capital budget tied to the AI buildout.

Money does what money always does when it gets scared. It runs to safety. Treasury prices rise, yields fall, and the Federal Reserve likely responds to the fallout by cutting short-term rates if employment or credit conditions start cracking. Here the pitcher does not get his strikeout. The batter gets hit by the pitch and still reaches base, ugly and painful, but on the bases all the same.

Two roads. One destination. Long-term rates head lower whether the machines deliver or disappoint. The mechanism changes completely. The direction does not.

Slok framed his piece around a six-month window, and I understand why. Wall Street research has the attention span of a toddler at a birthday party, and nobody gets paid to publish a two-year thesis when the client wants an answer before lunch. But I have been doing this long enough to know that market narratives take longer to play out than the analysts promise and shorter than the skeptics hope.

My read is that this is not a six-month story. It is a two-year story, and probably closer to two years than one. Corporate capital spending cycles do not resolve in two quarters. Productivity data takes years to show up cleanly in the numbers, and even a genuine AI bust would take several quarters to fully drain out of corporate balance sheets and capital plans. Six months gets you the first read on the scoreboard. Two years gets you the final score.

The evidence that will decide this argument is not mysterious. Watch AI-related revenue growth at the hyperscalers. Watch corporate adoption rates outside the technology sector itself, because that is where the real productivity case gets made or broken. Watch data-center utilization, profit margins at companies actually deploying these tools, and whether customers using AI applications can point to real dollars earned rather than hypothetical efficiency.

Investors have financed this entire buildout on the promise of future returns. That patience has a shelf life. Somewhere in the next several quarters, the market stops asking whether AI is interesting and starts demanding to know whether it is profitable. That is the moment the narrative tips decisively toward one branch or the other, even though both branches still end with lower long-term rates.

Honesty requires me to tell you what this thesis does not promise. Slok never assigned probabilities to either scenario, and neither will I pretend to know the odds with any precision. Reality rarely hands you a clean binary.

AI could produce moderate productivity gains without ever generating the trillions in revenue currently baked into technology valuations.

The Nasdaq could correct 15% or 20% without collapsing by half. Continued AI infrastructure spending could remain mildly inflationary for years through higher demand for electricity, construction labor and semiconductors, even while the broader disinflation story plays out elsewhere in the economy.

Fiscal pressure does not vanish just because AI becomes the dominant headline. Heavy Treasury issuance and unresolved deficit concerns can keep pressing upward on term premiums even as inflation itself cools. A market narrative is not a law of physics, just a current, and currents get interrupted by other currents.

None of that changes my central point. The dominant force acting on long-term rates over the next two years is far more likely to be the AI cycle than the inflation and deficit chatter currently dominating financial television. Markets trade on anticipated conditions, not today's headlines. Anyone still positioning exclusively around this year's inflation print is preparing for a game that already ended.

If long-term rates fall, and I believe they will, the assets that benefit look very different depending on which branch of the tree you are standing under, with one enormous exception.

Long-duration Treasury bonds win in both scenarios. Falling inflation expectations pull yields down in the successful case. A flight to safety and expectations of Federal Reserve easing pull yields down in the disappointing case. High-quality corporate bonds ride the same wave, though credit spreads behave differently depending on which path unfolds. Equities split hard. Successful AI adoption could support earnings broadly, assuming valuations have not already front-run the good news. A technology bust would hit growth stocks and AI-adjacent infrastructure names the hardest, exactly the concentration risk sitting inside every major index fund right now.

Banks, business development companies and other lenders generally prefer the successful path, where gradually falling rates arrive alongside continued economic growth and stable credit quality. A technology-led downturn might lower funding costs for these same institutions, but slower loan growth, softer collateral values and rising defaults would likely swamp that benefit.

This is where 30 years of buying cheap, safe, cash-generating assets starts to matter more than any macro forecast. The Flagship Report Income Portfolio rests on an older idea than any bet on artificial intelligence. You get paid to own assets trading below their true worth while the market sorts out which narrative wins.

Every holding inside the portfolio was selected because it throws off real cash today, not cash promised three years from now if a chatbot finally learns to file insurance claims. That distinction matters enormously in a world where long-term rates are heading lower for reasons that have nothing to do with the quality of any individual business. When rates fall, income-generating assets that were already paying you handsomely become worth more, because the discount rate applied to those future cash flows shrinks. You do not need AI to succeed for that mechanism to work. You need rates to fall, and both of Slok's scenarios get you there.

The portfolio is currently positioned to deliver a cash dividend yield of 10%, paid to you in real dollars rather than promised in some future productivity miracle. Combine that yield with the price appreciation that tends to follow falling long-term rates across community bank preferreds, REITs trading at steep discounts to net asset value and high-quality income vehicles, and the total return math starts to look considerably better than what you are likely to get from an index dominated by a handful of richly valued technology names carrying binary outcome risk.

That last part is worth sitting with. A market-cap-weighted index today is effectively a concentrated bet on Slok's first scenario playing out cleanly, with very little room for error and almost no cushion if the second scenario shows up instead. The Flagship Income Portfolio does not require AI to win. Rates falling, cash flows arriving, and discounted assets closing the gap to fair value are enough. History suggests that setup wins far more often than it loses, and it does not require you to guess correctly about a technology cycle that even its biggest boosters cannot fully explain.

I grew up watching the Orioles lose plenty of games that looked winnable and win plenty that looked hopeless. What I took from those summers at the ballpark was not a lesson about talent, it was a lesson about positioning: the teams that lasted were built to survive a bad month, not just to capitalize on a good one.

That is the entire philosophy behind this portfolio. AI could still turn into the productivity miracle its evangelists promise, or into an expensive lesson in overreach that ends with a painful correction. I genuinely do not know which inning we are in, and anyone who tells you they know for certain is selling something other than honest research.

What I do know is that long-term interest rates are very likely headed lower over the next two years, for reasons that will look obvious in hindsight and feel underappreciated right now. A portfolio built to collect a 10% cash yield while waiting for that shift to happen does not need a crystal ball. It needs patience and a willingness to get paid while everyone else argues about which headline matters more.

Marcus Aurelius never had to think about semiconductor capital expenditure, but he understood the part of this that matters most. You do not control which scenario the market chooses. You control how well positioned you are when it does. Collect your dividends. Let the machines sort out the rest.