The Storm Before the Calm
Four years of bond losses are a generational storm, produced by massive policy interference. A shift in policy and technological innovations is signaling that the calm is at hand.
Author’s Note: This article is more sophisticated and analytical than many prior ones. Some readers may find it boring, confusing, and possibly even annoying. So, below are a few bullet points meant to summarize it in a few sentences:
Summary bullet points:
Ø Federal Reserve Policy suppressed bond prices for four years. In short, the Fed acted like a bartender who kept saying “one more round” through guidance, rate cuts, and bill purchases. Big institutions used the free drinks to short long Treasuries and buy tech and AI stocks. Then the Iran oil spike showed up and started yelling “shots!” The party’s almost over, and while the bar closes, those present don’t have to go home, they can’t stay any longer. This spells the end of higher interest rates, and the beginning of rising bond prices.
Ø Spreads are falling despite the Fed’s efforts. The yield curve has been flattening, despite the rise in interest rates. In a way, the yield curve is a pool float slowly losing air. Banks earn their living on the gap between short and long rates, so a thin gap means the banks are like a bakery that can’t charge more than its flour costs. Lending dries up, and credit and inflation cool off. This leads to lower rates.
Ø The AI Bubble is deflationary. A stronger dollar spells bad news for tech stocks that, combined, are now valued at over 150% of the US economy. A stronger dollar weighs on profits and eventually drags prices lower. As AI stocks fall and many of these start-ups fail, credit deflation combined with technological innovation will lead to lower inflation and lower interest rates.
Ø The mortgage engine is a sleeping giant. Mortgage convexity is like a snowball that grows as it rolls downhill. When yields fall, homeowners refinance and mortgage bonds suddenly mature faster than planned. Holders then must buy long Treasuries to patch the hole, which pushes yields down further. This engine is currently asleep, but even a marginal drop in yields could awaken it. As interest rates fall, their fall is likely to speed up.
Ø All of this points to much, much lower interest rates over the next 1-2 years.
Introduction
The 10-year Treasury yield closed Wednesday at roughly 5.0%, about a hundred basis points higher than a year ago. The Fed held its policy rate at 3.75% to 4.00% after a September hike, the first since 2023, and two committee members dissented in favor of another. Brent crude spent September between $105 and $117. By any ordinary reading, this is a bond bear market in full cry.
My argument is that it is something more specific and more fragile: a bear market held in place by policy. For four years the Federal Reserve, and then the market's reaction to a war, supplied the conditions that kept long-term Treasury prices suppressed and the yield curve steep. The supports are now being withdrawn, the curve is (and has been) flattening, and this has massive implications for bank credit, disinflation and the dollar.
Four years of losses
The bond bear market of this cycle began as the result of an inflation shock, due in part to excessive COVID era spending, and subsequently due to the energy shock from the Russian-Ukrainian conflict. The Fed hiked their rate target from zero to over 5% by July 2023, the highest policy rate since early 2001, and long bonds, priced for permanently low inflation, took historic losses. However, long bonds didn’t rise as much as the Fed Funds Rate did, leaving the US Treasury yield curve uncomfortably “inverted”. In fact, inversion is never comfortable.
Inversion is really bad for traditional banks, who rely on a steep yield curve (and the spread they get between borrowing and lending) as a revenue stream. This led in part to the blow up of Silicon Valley Bank (and a few others) in the spring and summer of 2023. By the end of 2023, the financial system was in dire straits, leading the Federal Reserve to signal to the markets its intention to cut aggressively in 2024.
The part I think is misread by many is what happened after the hiking stopped. The Fed cut in 2024 and again in 2025, and the easing did not bring long yields down. Contrary to many people’s expectations, the aggressiveness of the cuts (the speed, which was faster than the market expected) actually pushed long-term interest rates higher. For the largest financial institutions, that wide and reliable spread was the opportunity.
Here is the mechanism as I read it. Easing cheapened short-term dollar funding. An institution could borrow at those falling short rates, “short” long-dated Treasuries against the position or run a steepener, and put the proceeds into assets that benefit from easy money: tech-related equities and the debt that financed the AI build out. This naturally produced a weaker dollar, which is just another side of the same trade. The DXY (the dollar index) fell roughly 9% in 2025, which made borrowing dollars to invest in risk assets cheaper still, because the funding currency was depreciating. A weaker dollar also produces stronger tech profits, as they source a lot of their revenues from outside the US. Each cut lowered the cost of the funding leg, and the short-Treasury leg paid when the curve steepened. The shorting was not a verdict on Treasuries. It was a funding leg, and it held long bond prices down while it pushed money into the Nasdaq and AI credit.
This is a market-structure argument, so forgive the technical jargon. But in plain English: the entire financial system is essentially one big trade. And it’s a trade that, at least for the past few years, has kept Main Street struggling under the weight of high long term interest rates, while Wall Street has feasted at the table of abundant, cheap, and easy credit. This explains why bonds (and much of the economy) have been in the doldrums, while the Nasdaq has reached epic heights.
The argument also explains why the structure of this trade, and the market under it, is so fragile. A position that depends on a steep curve, a falling dollar and a guided front end loses on all three when the curve flattens, the dollar rises and guidance disappears. That is the setup we now have, as the rest of this piece describes.
The policy that held bond prices down
There were three layers of support that all pushed in the same direction.
Forward guidance in 2023. By signaling that rates were at or near their peak and that cuts were coming, the Fed effectively pre-announced the end of tightening. Markets treat a promised easing cycle as a reason to own the front end and to demand more compensation for the long end, because easing into elevated inflation raises the odds that inflation stays higher for longer. Guidance compressed short-rate volatility and tilted the curve toward steepness. In other words, the Fed’s easing guidance was a gift to the banks, and the infantile AI sector.
Rate cuts in 2024 and 2025. Despite the initial relief, the system began to pull rates down in 2024 anyway. This led to the Fed, right before the presidential election, cutting while inflation was still above target and deficits were enormous. This was a steepener by construction. The front end fell with policy; the long end priced fiscal supply and inflation risk. The dollar fell with it: the dollar index lost roughly 9% in 2025, reaching its weakest levels in years.
The end of QT and the start of balance-sheet growth. QT officially ended on December 1, 2025, after shrinking the balance sheet by roughly $2.4 trillion. Eleven days later the Fed began what it calls reserve management purchases, buying short-dated Treasury bills at about $40 billion a month. The Fed is emphatic that this is plumbing, not stimulus, but like most bureaucrats, they are fond of making distinctions without a difference. The mechanics are essentially identical: the balance sheet expands, reserves rise, liquidity increases, and the Fed signaled it was willing to put the system's funding comfort ahead of balance-sheet discipline. To a market that reads regime change from the direction of the balance sheet, that was a form of QE, whatever the label.
Stack the three together and you get what the data showed: a steeper curve and a softer dollar. Cuts at the front end, more liquidity, and an open-ended message that - whatever the Fed’s “official” objectives may be – their primary concern is the financial markets. That is exactly the environment in which long-duration Treasuries are a poor place to hide, because every one of those actions makes shorting (i.e. pushing the price down) US Treasury bonds profitable. The result was over two years of heavy-handed Federal Reserve support for big banks (at the expense of Main Street), and a much steeper curve:
Despite all of this (and it really needs to be emphasized for our clients that have been with us for some time), the yield curve overwhelmed this policy interference too, and continued to flatten through February of this year. The financial system was simply too much for the massive policy support the Fed was providing, and interest rates once again began to fall.
Then came the energy
Earlier this year the war with Iran put an enormous amount of energy into the system, literally and figuratively. Oil price per barrel jumped from the mid $50s in December of 2025 to the low $80s within days of the early March escalation, and by September had run above $100, ending the third quarter up about 42%. Gasoline hit a record September average above $4.30, due to massive destruction to oil refineries in both the Middle East and in Russia, where Ukraine continues to punish (ironically) it’s allies in Europe.
For traders, an oil spike is a gift to the steepener trade. It feeds the inflation narrative at the long end while the central bank is expected to be slow to react at the front. Participants who had been pressing the steepener since the cutting cycle began had a fresh and powerful reason to add to it. The result was another leg of pressure on long bond prices, delivered by the market rather than by the Fed, but resting on the same foundation of policy support. Add to this the massive amount of debt issues by AI-related companies that are finding equity funding harder to come by, the duration market becomes (temporarily, anyway) saturated. The result was the steepener traders got part of what they wanted: higher interest rates.
The curve is flattening anyway
Here is the part that, to me, is the most important and the least discussed. Despite all of that support, the curve is not steepening. On October 23, the 10-year minus 2-year spread stood at about +31 basis points, down from about +53 this time last year. The 30-year minus 5-year spread was roughly +51 basis points. The steepener trade has been supported by the Fed, by the balance sheet and by a war, and it is still losing ground! In fact, the spreads have been falling for well over a year now, overwhelming the combined forces already described.
I want to be precise about what this is and is not. So far it is a bear flattener: the 2-year yield is up some 114 basis points over the past year, more than the 10-year (about 98) or the 30-year (about 76). The front end is rising faster than the back because the Fed hiked and the market is pricing more. That is not yet a bull flattener, and I will not pretend it is. But history has shown that bear flatteners always lead to bull flatteners (where all rates go down, but long rates go down faster). Always.
But a flattening curve does its work regardless of how it arrives. Banks borrow short and lend long. When the spread between what they pay and what they earn compresses, margins squeeze, lending standards tighten and credit growth slows, with a lag. That is how a flat curve turns into disinflation, a stronger dollar, and eventually much lower interest rates.
Warsh removes the put
In addition, the Fed that built the supports is no longer the Fed in charge. Kevin Warsh took over in May and has said repeatedly that he does not believe in forward guidance absent a crisis. He has shortened the post-meeting statement dramatically, declined to issue forward projections in the old style, and on Wednesday announced a reduction in rate-setting meetings moving forward. The plan is to hold seven rate-setting meetings next year, with one meeting devoted to strategy with neither a vote on the funds rate nor a policy statement, and six rate-setting meetings in 2028. Emergency meetings remain possible. But the direction is clear: less signaling, fewer scheduled decision points, and a chair who says plainly that he will not smooth the path for the market. Analysts quoted on Wednesday expect more volatility as a result.
In my framework this is one of the most important developments of the year. Forward guidance suppressed short-rate volatility, and suppressed volatility is what makes the carry trades and curve trades work. Remove it and the options-implied rate path becomes uncertain; term premium becomes harder to harvest; the structures that depended on a guided Fed need to be unwound or repriced. On the other side, long-term bond volatility falls as short-term treasury volatility rises. The lower volatility allows institutions to increase their allocation to bonds, thus adding to demand on the long end. The Fed is currently declining to hold the market's hand, allowing volatility back into short-term rates, and cautiously raising rates into an energy shock. All of these moves are designed to flatten the curve.
Why this is deflationary
Headline inflation has been running around 3.4% to 3.7% due to the energy shock, with core inflation expected to hold near 2.4% (a five-year low) in the most recent consensus previews. If the energy surge is the main driver, then the macro question is what happens when the energy bill bites into demand.
Tariffs. The 2025 tariffs act like a tax on the consumer and on margins. In the first Trump term, the pattern was a one-time price-level adjustment followed by weaker growth and demand, and the Fed cut three times in 2019 with trade tensions as part of the stated rationale. My expectation is that the same sequence repeats: a bump to goods prices that passes, then the drag on income and investment that remains.
Ø One caveat: during Trump’s first term, China largely ate the costs of the tariffs by keeping their currency pegged to the dollar. This time around, they decided to depreciate their currency as a way of exporting inflation (so to speak) into the US market. Their currency has fallen close to 10% against the dollar in the past 18 months. This could provide a little more support for US inflation in the short term, but not without imposing great cost to China’s own economy. It is not a sustainable strategy for them, and therefore it will not be sustained.
Demand destruction from oil. At $100-plus oil, consumers and industry cut consumption. High prices are the cure for high prices, and the cure acts with a lag. If the spike persists, I expect it to be the dominant disinflationary force of 2027, not an inflationary one.
Mortgage convexity. This is the engine I most want readers to understand. With the 10-year near 5%, mortgage rates are high and prepayments are low, so the mortgage market is currently a source of extension risk. But convexity is nonlinear. When yields begin to fall, prepayments accelerate, the duration of mortgage-backed securities shortens, and the holders of those securities must buy Treasuries to restore their hedges. That buying drives yields down further. It is a self-reinforcing bid for duration, and it turns on at precisely the point when the cycle changes. This, combined with Treasury Secretary Bessent’s recent decision to start utilizing the Treasury’s piggy bank to start buying long term US treasury debt, is a recipe for a demand squeeze in US bonds over the next year to two years.
If these channels do their work, the second half of this presidential term looks like the second half of the first: much lower interest rates, a strong dollar, and a market starved of duration supply relative to the demand from hedging flows.
The dollar and the Nasdaq
The dollar has already begun to turn. On December 19, 2025, the dollar index formed a golden cross (a technical term, and very positive), with the 50-day average crossing above the 200-day. Bank of America's technical team counted it as the 39th such signal since 1970 and noted that it was unusual because the 200-day average was still falling. The index is around 102 as of mid October, up roughly 7% since the recent low in January.
A stronger dollar is a problem for the AI trade. The Nasdaq-100 earns a large and growing share of its revenue outside the United States. A rising dollar translates those foreign earnings into fewer dollars, while higher energy costs weigh on overseas demand for devices, cloud services and the data center build-out itself.
There is a second, slower channel. The AI build has been financed with credit. If the equity bubble deflates, the funding dries up, and the credit that supported both the capital spending and the wealth effect goes with it. In addition, the technology is itself deflationary: when it works, it makes goods and services cheaper to produce and thus cheaper to buy. The same innovation that juiced these ridiculous valuations eventually pulls prices down across the economy. Bubble deflation and innovation deflation arrive together.
What would prove this wrong
A thesis worth publishing is one that tells you how it can fail. Here are the failure modes I take most seriously.
The Fed starts easing aggressively. The September hike increase was unanimous, the committee projected another before year-end, and futures point toward a December move. If the Fed keeps tightening into energy-driven inflation, the front end keeps rising, the flattening continues but likely turns into a bull flattener as economic growth and inflation slow down faster. However, if the Fed pivots (not my base case, considering Warsh’s sentiment) and begins cutting again aggressively, then it could delay the fall in interest rates for a short period of time. The market expects the Fed to hike or hold, and anything other than that would be additional interference. Something Warsh has seemed to swear away. Regardless, my thesis is about where the cycle ends up, not about next month's price. Rising yields and a flattening curve all lead to lower yields. But the timing may be different.
Oil may not break immediately. Demand destruction is a forecast based on relatively meager economic growth combined with higher energy costs. If economic output increases dramatically (which seems unlikely given the current infrastructure), energy stays high and headline inflation could remain elevated through 2027. This is about as likely as Wiley Cayote catching the Road Runner, but I suppose it’s possible.
Fiscal supply. Large deficits can keep a floor under inflation, and a flatter curve does not by itself cancel that. If the Democrats win Congress in the mid-terms, D.C. is likely to be in a form of deadlock for the next two years. Split power typically is the only time we see less spending in Washington, so I’m not expecting fiscal dominance to proliferate.
None of these are trivial. But the market's current bet, that long bonds need to be sold to protect against ever-higher inflation, is the same bet it made before the last two growth scares, and the policy infrastructure that made the bet profitable is being dismantled in real time.
The calm
The storm is the final stretch of a bear market in which the Fed, a war, and a steepening trade all pushed in the same direction. The calm, as it arrives, will come from things steepeners cannot ultimately fight forever: a flat curve, a tighter credit channel, a stronger dollar, and an economy where energy and tariffs have taxed demand until prices give way.
The curve is already telling us which way the pressure runs. The question is whether the policy that suppressed bond prices for four years can be withdrawn before the thing it was suppressing arrives. But that question is about timing, not destiny. Without the knowledge of Providence, I of course cannot guarantee a result. But in this case, our destiny appears to be written in, if not stone, at least hard pan.