Henrik Zeberg Says Bond Yields Will Collapse: The Cycle Thesis, Examined

• The man, his incentives, and the venue that sells gold
• His system, stated at full strength: the Eiffel Tower
• The labour market, where the tape has already been kind
• The two masks he says are hiding the weakness
• The credit-spread test, and the sharpest thing in the interview
• The central bank, and where the market moved to him
• The destination trades: dollar, gold, equities, bitcoin
• Where his account converges with ours, and where it improves us
• Where we differ, and the part he does not solve
• The markers, and the verdict

Henrik Zeberg says bond yields are about to collapse, that the recession has already been signalled, and that the bitcoin top is behind us — so this episode states his cycle framework at full strength, reports where the record has already agreed with him, concedes where he improves our own work, and marks the claims in it that can never be wrong.

DeepDives · 2026 · An AI–human collaboration that weighs every source, mainstream or alternative, under strict rules of truth and logic.

§The man, his incentives, and the venue that sells gold

On the twenty-ninth of September, a Danish macro strategist sat down with the host of a programme presented by a gold dealer and made the most contrarian call in the market. Everybody else is arguing about how high bond yields can go; he says they are about to fall hard the other way, that the recession has already been signalled, and that one final melt-up in equities is weeks away. His name is Henrik Zeberg.

He runs the cycle research at Swissblock, a Swiss macro and digital-asset analytics firm, publishes the Zeberg Letter, a paid monthly, and is a Forbes columnist and a TED speaker, and his book, The Monetary House of Cards, introduces the business-cycle model the whole argument rests on.

Three days after he spoke, the most important number in his framework arrived, and it arrived on his side. On the second of October the American jobs report came in at twenty-nine thousand, against a consensus near eighty-four thousand; the unemployment rate ticked up to four point two per cent, and the two previous months were revised down by sixty thousand between them, taking July from a small gain to a loss of ten thousand jobs. So the examination is not whether a man calling a recession is a crank, but whether the framework behind the call is sound.

The incentives belong on the table before the argument does. A subscription product needs a call and a cycle product needs a clock, so the larger the top and the rounder the numbers on the destination, the more the letter is worth reading. That is a reason to test his numbers, not to dismiss them.

The venue matters too. The programme is presented by a gold dealer and the advertisement opens the show, so a guest who tells you to buy gold on the next dip is selling bitcoin content to an audience that came for gold. Our own series holds the same hard assets, and one of our recent episodes examined the opposite thesis, so nobody here is neutral.

§His system, stated at full strength: the Eiffel Tower

His system has one spine. Inflation, he says, does not lead; it lags, and what it lags is the labour market, because a demand shock takes time to travel through the supply chain, whereas a labour market turning over hits household cash flow immediately.

Put that ordering together with the ordinary observation that yields spike late in a cycle and you get his whole bond call. Capital demand from a boom pushes yields up into the late phase, exactly as it did in the third and fourth quarters of 2007, and exactly as it did going into 2001. Then employment rolls over, inflation follows it down, and yields fall with it. The spike is the left tower, the collapse the right, and the right is what matters. He calls the shape an Eiffel Tower, and expects the right side of it now, which is why he says the past year’s move in yields is the top of the move rather than a new repricing beginning.

§The labour market, where the tape has already been kind

The labour market is where he puts his weight, and here the tape has been kind, though not in every detail. His model, he says, produced a structural recession signal in July, one to three months ahead of the previous eight recessions without a false reading. That cannot be checked from outside his model, and he widens it himself in the same breath, saying one to three months can also mean four.

What can be checked is the payroll series, and there he is close to exactly right. The annual benchmark revision published on the twenty-eighth of August took the twelve months ending in March from two hundred and seventy-three thousand, about twenty-three thousand a month, down to one hundred and ninety-four thousand, or sixteen thousand a month, which is his number and the government’s number. He says the revision was seventy-six thousand, against a published seventy-nine thousand, and the twelve-month average is around thirty-one thousand. He says unemployment is four point one per cent; it was four point one when he spoke and four point two now.

The rest of his labour case holds up about as well, with one overstatement. He says long-term unemployment is higher than at any point outside a recession and still rising; the official series has one point nine million people out of work for twenty-seven weeks or more, twenty-seven per cent of everybody counted as unemployed. He says one point seven million people have left the labour force since November; the number reported is about one point three million, so he overstates by roughly a quarter, and the denominator he uses, a workforce of a hundred and seventy million, is exactly right.

Where he goes beyond what can be checked is the survey mechanism. He says the one hundred and sixty-two thousand August print came with the lowest first-print response rate ever recorded, which is why he refuses to believe it, and three days later that print was revised down to one hundred and thirty-three thousand. His instinct was right even though the percentage cannot be verified from outside.

§The two masks he says are hiding the weakness

Then come the two masks he says are hiding all of this, and this is the part of his case our own work has been circling without quite landing on it. The first mask is artificial intelligence. The build-out, he says, has added one to two percentage points to American growth this year by adding investment rather than jobs, and if you strip it out the economy is already contracting. That residual cannot be verified, but the weakness it covers can: July payrolls were revised to a loss of ten thousand, September came in at twenty-nine thousand, and the twelve-month average is thirty-one thousand.

The second figure in that argument is the one to put in front of any listener who thinks this is alarmism. He says the artificial-intelligence industry needs something like three trillion dollars of revenue a year to justify what has been built, against roughly one hundred and seventy-five billion being generated. The three trillion is not his invention; it is Sequoia’s David Cahn, from his July note, and the arithmetic is that about one and a half trillion of infrastructure spending requires about three trillion of revenue to pencil out. Cahn’s own estimate of the two leading model companies’ combined revenue is nearer eighty billion, so the gap is wider still.

The consumer is the third leg, and there he works with a comparison rather than a measurement. The saving rate, he says, is down to three per cent, where in the stagflationary years the household saver had nine to twenty per cent set aside; the official series puts it at three point zero per cent in July and four point one in August, and in the 1970s the rate ran between about eleven and thirteen per cent and peaked at seventeen point three in May 1975, so his twenty per cent is above the record. The household of the 1970s had several times today’s cushion, and a household with four per cent of income unspent cannot absorb a shock at the pump.

§The credit-spread test, and the sharpest thing in the interview

Now the single most interesting test in the interview, and one this series has not run. If there were a genuine crisis of confidence in government debt, he says, it would show up first and worst in the worst credit, because that is how a loss of confidence ranks a market: the weakest borrower pays the most. And the weakest borrowers are not paying the most. The American high-yield spread is about three point two per cent, near the tightest levels on record, and the record private credit defaults sit inside a market whose overall loss rate is still under one per cent of principal. If the bond-vigilante reading were right, high-yield spreads would be the first to blow out, and they are not.

Where he overstates his own case is on credit generally, treating private credit stress as something only starting. Fitch put the American private credit default rate at a record six point three per cent in August, with fourteen defaults against three in July, and borrowers under twenty-five million of earnings defaulting at twelve per cent. Houlihan Lokey has one in eight loans to ten-to-twenty-million-dollar borrowers marked below ninety cents on the dollar, against about one in a hundred in 2023. The stress is not starting; it is at a record.

§The central bank, and where the market moved to him

The central bank is where his prediction and the market have just agreed, and it is worth giving him the credit precisely. He says the hike was a mistake, calls the idea of another one in October ridiculous, and describes the committee as four hundred people with doctorates who misread the last two cycles. His date is wrong by four days: he says the move came on the twelfth of September, and the Federal Reserve raised its target range by a quarter point to three point seven five to four per cent on the sixteenth, the first increase since 2023.

The forecast is the interesting part. Going into his recording the market was pricing a second hike in October; by the first of October that had flipped to a hold, and after the second of October payrolls the hold probability jumped from thirty-six per cent to eighty-two per cent inside a week. So the market moved to him in five days, and he is on the record with the direction.

His historical comparison is checkable too, and it holds. Inflation peaked at five point five per cent in July 2008, and stood at three point nine per cent in January of that year while the Federal Reserve was already cutting.

§The destination trades: dollar, gold, equities, bitcoin

Then the destination trades, where a cycle framework either earns its keep or does not. On the dollar he wants a fall to ninety-three or ninety-four on the index before the crisis rally takes it to about one hundred and twenty; the index was ninety-nine and a half at the end of August and about one hundred and two in October, so it has gone the other way since he spoke. He pre-announces the branch that rescues him, a scenario where capital rotates into the United States and the dollar strengthens while risk assets rally on it, which he ranks as less likely rather than not possible.

On gold he is long-term bullish and tactically bearish, has been saying exactly that for a year, and names his own miss. He called this correction in October 2025 with gold around four thousand three hundred; the metal then ran about thirty per cent higher to a record near five thousand six hundred in late January before falling to about four thousand one hundred now. The destination was right and the sequence was wrong by a quarter and a third of the price. He names a floor around three thousand one hundred, and a long-run figure in the region of twenty-seven to twenty-nine thousand by 2030.

That record is worth anchoring, because it decides whether the correction he describes is real or imaginary. Gold closed at five thousand five hundred and eighty-nine dollars an ounce on the twenty-eighth and twenty-ninth of January, the highest price in its history, and traded as high as five thousand five hundred and ninety-five intraday; from there to four thousand one hundred and nineteen is a fall of about twenty-six per cent. Most of the correction he says is coming has already happened, and he said it would happen when the metal was a thousand dollars below its eventual peak.

His gold argument also contains the sharpest thing he says about the monetary system. He puts up the ratio of the stock index to the price of gold and points out that it is breaking down, and that it has broken down at three previous turning points: after 1929, after the dollar was taken off its gold link in 1971, and after the technology top in 2000. His reading is that the first and the third were speculative bubbles, that 1971 was a failure of confidence in the monetary system itself, and that this time both drivers are live at once.

The equity and crypto end of it, stated plainly, because these are the claims with clocks. He expects one final advance to new highs, and puts the Nasdaq in the region of thirty-seven to thirty-nine thousand and the S&P in the region of eight thousand five hundred to nine thousand. For scale, the Nasdaq closed at a record twenty-seven thousand six hundred and the S&P at a record seven thousand eight hundred on the sixth of October; his Nasdaq target is thirty-five to forty-two per cent above today’s record and his S&P target nine to fifteen.

Then the bust, which he names as deflationary rather than inflationary, with the housing market frozen at about four million resales a year against a norm near five and a quarter million, and private credit the fuse that finally reaches the equity market. The top, he says, could be in this quarter, with January the latest. On bitcoin his call is that the October 2025 record above one hundred and twenty-six thousand was the top, that the bear market has begun, and one more rally to about one hundred and fifteen thousand where it outperforms gold. Ethereum he has going to seven thousand two hundred to seven thousand four hundred.

§Where his account converges with ours, and where it improves us

Now the part that has to be said carefully, because a listener could be forgiven for hearing two accounts of the same patient and assuming one is dishonest. We have published that the American Treasury has crossed from selling its debt to managing the price of it: the buyback programme doubled on the nineteenth of August and runs to the fourth of November, the debt passed forty trillion dollars in the same week, and the thirty-year yield had been at levels not seen since 2007. He describes the same events and reads them as the ordinary spike at the top of a cycle.

And we agree on more than that. Our own debasement episode made the point that the artificial-intelligence build-out is a claim on the same long-dated capital market as the government’s own borrowing; our liquidity examination named private credit pain as the next thing on the dashboard, and he names the same sector as the fuse for the equity top. Our work has said for a year that the central bank fights the last war while the fiscal authority fights the current one. That is independent agreement by different methods, and it strengthens both accounts.

So here is what he improves. We have been writing that the labour market is the transmission belt; he states the stronger version, that inflation is a function of employment with a lag, which makes the late-cycle inflation prints close to useless as a warning, and given that core inflation has come down to three per cent while payroll growth has collapsed, he has the better account of this year than a framework that watches prices first. His separation of two kinds of crisis is cleaner too: we have folded the 1971 monetary-system break and the 1929 and 2000 speculative breaks into one story while he splits them.

§Where we differ, and the part he does not solve

Where we differ is the direction of the long end. Our series has published that once the state manages the price of its own debt, the long end stops being a market price and becomes an administered one, and that the visible consequence is a persistent erosion of its credibility rather than a rally. The tape has been running our way as much as his: the ten-year closed at five point two nine seven per cent on the thirtieth of September, breaking its 2007 peak and its 2007 closing high, a level last seen in May 2002, and the thirty-year reached five point six four per cent.

Give him his strongest reply, because it is a good one. A buyers’ strike and a loss of confidence look identical at the top of a cycle, and the difference between them is duration rather than shape, and our own account cannot distinguish them either. There is also genuine evidence for him in the same window: on the second of October, after a soft payrolls report, the two-year yield fell for its first weekly decline since the second week of August while the long end still rose.

The dollar is the harder problem for him. His central path has the index falling to ninety-three or ninety-four, and it is at about one hundred and two, having risen through the whole period in which he expected it to fall, and it is the leg his roadmap hangs from, because the gold takeoff, the equity top and the crypto rotation are all timed off the dollar turning.

Then the sharpest single point, and it is his rather than ours. If his ordering is right, that employment leads and prices lag, then the bond call is not a separate forecast at all; it is a corollary. Everything he says about yields follows from one proposition, that the labour market is the independent variable and inflation the dependent one, and that proposition is the one thing in this whole argument that can be adjudicated with a data series. The entire structure stands or falls on the payroll run-rate.

And the corollary of the corollary is the thing that makes him uncomfortable reading in a gold-selling venue. If employment leads prices down, then the debasement trade he has spent a year arguing for is not what is happening now. What is happening now, on his own account, is a disinflationary bust that will hurt gold, hurt bitcoin, and only then set up the monetary event that makes both of them.

There is one more asymmetry in his favour that he does not press hard enough. Almost everything the bears and the debasement camp said would happen at the long end of the curve has now happened: the ten-year broke its 2007 record, the thirty-year hit a twenty-four-year high, the curve steepened, the term premium came back, mortgage rates went to seven point three per cent, and the Treasury had to buy its own bonds. And the equity market made a record high on the sixth of October anyway, so if fiscal dominance meant asset prices had to break, the tape has had a year and has not.

The strongest case, at its best, is worth stating whole. He identified a slowdown at the top of 2024 and the slowdown came, at a pace nobody expected. He said the labour market was far weaker than the headline prints, and the benchmark revision proved it. He said the strong monthly number would be revised down, and three days later it was. He said the central bank was wrong to hike and would not hike again in October, and inside a week the market moved from pricing a hike to pricing a hold. Anyone who reaches for the word perma-bear has not read that record.

Then the falsification problem, which is the part he does not solve. His timing claims cannot fail: he gives a window of one to three months and widens it himself to four, and says the top could be in this quarter or in January. And one claim covers both outcomes, which is worse: on the dollar he says ninety-three or ninety-four, and then a stronger dollar with a risk-asset rally on capital rotation into the United States as the alternative. Set that beside a both-directions call on gold and an equity call satisfied by a melt-up and equally by a bust, and the composite forecast is true in nearly every world a listener could imagine.

§The markers, and the verdict

The verdict comes in four pieces. On the diagnosis he is right, and the second of October settled it: the labour market this series treats as the transmission belt really did stop carrying the economy, and a man calling a recession from a cycle model was reading it better than the market, which was still pricing a rate increase. On the method he has improved us, and his credit-spread test is a gauge we should have built ourselves.

On the destination trades he is wrong by sequence. The gold correction arrived a quarter later and a third higher than he said, and only after a record whose size he admits he missed. The dollar has gone the other way since he spoke. The equity top and the bitcoin bust are both still ahead of the tape, and a call whose two branches cover a melt-up and a bust is not yet a call at all. And on the bond market the verdict is unresolved, because his own signal has flickered once, with the two-year turning down on the second of October.

Which leaves the markers, and every one of them is dated. The first is the one he named, and it is the sharpest instrument in the interview: short-dated yields falling first and dragging the long end down behind them. Watch the two-year against the ten-year; if the two-year keeps falling while the ten-year holds above five and a quarter per cent, what is happening is a steeper curve on term premium, which belongs to our thesis rather than to his.

The second is the gauge he gave us. Watch the high-yield spread: if it stays inside four and a half per cent through the turn of the year, his claim that no confidence crisis has begun is confirmed, and if it breaks five, the bond story changes hands. The third is the labour series; the number to beat is his own falsifier, a run-rate accelerating to eighty, one hundred or one hundred and twenty thousand, which would be the soft landing, while anything near thirty thousand keeps his clock running. The fourth is housing: rising resales, rising permits, a market that unfreezes.

The fifth is the central bank, where the market now expects a hold on the twenty-eighth of October and a December increase is still well priced; if the committee raises rates in December into a twenty-nine-thousand payroll print, his charge that it is fighting the last war will have been demonstrated rather than merely argued. The sixth is the dollar, at ninety-three or ninety-four against about one hundred and two today, and the seventh is gold, at his floor around three thousand one hundred or at takeoff from here, which on his own schedule waits for the central bank to move.

And the eighth is the test that would prove him right rather than wrong, which matters more than all the others put together. If short-dated yields collapse, high-yield spreads stay tight, the dollar rolls over, and equities make one last high into the turn of the year before rolling, then a sequence that no other reading in this argument predicts will have happened exactly as he described it. Nothing else in the market would look like that combination, and if we see it, the honest thing will be to say so on air.

The incentives, named where they land, and then the close. Zeberg sells a monthly letter whose product is a cycle call, and a Swissblock service built on the same model, and he has a book making the same argument at book length, which is why the targets are round, the top the largest ever recorded, and the timing soft enough to survive a quarter of slippage. The venue points the same way: this is a programme presented by a gold dealer, and the guest’s most confident conclusion is that gold is the one asset that survives.

What survives after all of it is this. A cycle framework with one independent variable, employment, and one corollary, falling yields, got the labour market right while the market was still pricing a rate increase, and got it right on the government’s own revised numbers. It supplied the best test anyone has offered us for whether a debt crisis has begun, in the spread on the worst credit, and that test currently says no; and it has a destination side already wrong once on gold by a quarter and a third, currently wrong on the dollar, and protected on equities by a claim that is satisfied either way.

Which is roughly what you would expect from a model that is better at reading the present than at dating the future. The present is where the value is. On the twenty-ninth of September he said the labour market had broken and yields would follow it down; four days later the payrolls report agreed with him. If credit spreads stay tight while the front end of the curve gives way, he will have been early in the way that matters rather than early in the way that is an excuse.

This has been a deep dive into Henrik Zeberg’s case that bond yields are about to collapse, and whether the cycle framework behind it survives contact with this series.

Sources: this examination names its sources inline — the interview between Henrik Zeberg and the host of a programme presented by a gold dealer, with his commercial interests noted; the October employment report of the Bureau of Labor Statistics and its annual benchmark revision; the published work of Sequoia’s David Cahn on the revenue required by the artificial-intelligence buildout; the household saving-rate series and its historical range; the Fitch and Houlihan Lokey data on private credit defaults and markdowns; the Federal Reserve’s September decision and the market’s shifting odds for October and December; and the market tape for the dollar index, gold, bitcoin, the Nasdaq and the S&P. Every source is weighed with the same skepticism regardless of politics or business model. The audio version of this article, read by DeepDives, is available on Wavlake.