The Bessent Put

On Tuesday, the 30-year Treasury yield touched 5.33%, the highest in nineteen years. Within 24 hours, the US Treasury announced it was doubling its long-end buybacks: the liquidity support operations in the 10 to 30-year sectors go from $2bn to at least $4bn each, starting September 9th, with more to come at the November 4th refunding.

The 30-year fell to 5.19%. 2s30s flattened 7bps. Gold ended the day up $160 at nearly $4,500, Bitcoin rose 6%, ETH rose 9%, equities rallied, and the dollar fell 0.8% to below 99... a clean break of the technical structure it had spent all year building.

I can hear it already, “Fourteen billion of extra buybacks? Raoul, the market does that before lunch. You've lost it.” Yes, I see the size utterly clearly. And these are exactly the moments you look past the size, because the size was never the point...

For the first time, the fiscal authority, not the central bank, responded to a long-end sell-off within a day, with a purchase operation.

There is now a Bessent put under the long bond, struck somewhere around 5.3%. The market has been told the top of the range. That changes everything about how you trade bonds, and it ties together almost every thread we have been pulling on in GMI for two years.

What the market actually repriced

Start with what the sell-off was not...

The 30-year hit its 19-year high with the 10-year breakeven at 2.30% and the 5y5y forward at 2.31%. Inflation expectations barely moved; what blew out was real term premium, near its cycle highs on the Kim-Wright measure at 0.84%. That is a supply and demand problem, and the supply side has a new player in it...

The big five hyperscalers issued $159bn of bonds in the first five months of 2026, against an average of $28bn per year across 2020 to 2024. Amazon alone did a $54bn deal in March.

In five months, five companies borrowed more than they had in the previous five years combined.

Fuck me... the private sector is running its own duration cycle for the first time, with full-year estimates of $240bn against roughly $600bn of capex.

And note this is not accelerating government supply: Treasury’s long coupon sizes have been frozen for six quarters. The collision is exploding corporate duration against static government supply and a shrunken pool of buyers, with the pensions done de-risking and the foreign bid impaired by hedge costs.

The government is being crowded by the very buildout it needs to succeed, and this week it told us it will not allow that to set the price.

A mundane tool, an extraordinary signal

The mechanics are humble. The buybacks target off-the-run bonds clogging dealer balance sheets, dealers offer voluntarily, and the long coupon auctions were left unchanged. On paper, it’s dealer hygiene ahead of a heavy autumn calendar.

But the timing is the tell: a first-ever mid-quarter upsizing, the day after a 19-year yield high, with “at least” doing a lot of work in the language. The tool is mundane; the reaction function is not.

And there are receipts. Yellen was accused of “activist Treasury issuance” for tilting roughly $800bn towards bills, worth an estimated 25bps of 10-year suppression... about a full rate cut. Bessent campaigned against ATI and has now institutionalised it. When both parties run the same playbook, it is not a policy. It is a regime.

Felix Jauvin at Blockworks caught the tell two weeks ago when the QRA buyback language quietly changed. Marginal macro policy has moved from the Fed to the Treasury.

What the precedents teach

I wrote about this mechanism in GMI back in November 2022: keep the front end where policy needs it, issue at the front, buy back the long end. That was Operation Twist, 2011. It moved the 10-year about 15bps, flattened the curve, the stock market took off, and the dollar, note well, went up.

Then within a year the Fed decided twisting wasn’t enough and escalated to open-ended QE3. Hold that thought for November 4th.

Japan cuts both ways. A credible cap needs fewer purchases... BOJ buying fell from ¥80tn a year to under ¥20tn by 2018, because the threat did the work. That answers whether Treasury needs to scale this up: only when the bond market misbehaves.

But when inflation arrived in 2022, the visible cap became the cheapest short in macro, and the BOJ only held the line because it had a printing press. The Treasury does not... a capped defence with a capped budget is an invitation unless the Fed ultimately stands behind it. Hold that thought too — I’ll come back to it.

The precedent that worked is the 1940s. The Fed capped the long bond at 2.5%, real rates ran deeply negative, and debt to GDP went from 106% to 46% by 1960. Nobody defaulted and nobody austerity-ed their way out. They repressed their way out.

What made it work was the Fed's signature and negative real rates. Today we have neither... yet. Bills at 4%-plus against 3.4% CPI is repression-lite, the scaffolding going up rather than the machine running.

Eighty years apart, and it still rhymes.

Rebuilding the buyer

The buybacks weren’t even the first intervention this month...

On August 2nd, the US Treasury intervened in the currency market, jointly with Japan, to support the yen, and the BOJ was quietly given access to the Fed’s FIMA repo facility so Japan can borrow dollars against its Treasury holdings instead of selling them.

Why would the US Treasury care about the yen? Because Japan holds roughly $1.1tn of Treasuries, and a Japan defending its currency alone becomes a forced seller of the US long end. Same reaction function, seventeen days before the buybacks: protect the long end, whatever the instrument.

That was the supply side being defended. The demand side is the other half of the pincer, and Bessent has been building it for months.

In April he confirmed that Gulf and Asian allies had requested dollar swap lines, and his Tokyo and Seoul trips formalized the swap architecture under the Japanese and Korean banking systems, with the UAE plumbed in as a third leg.

Nobody is being asked to “please buy our bonds”. The conditions are being engineered under which buying them becomes rational again...

For Japan, run the maths. Hedging a Treasury from Tokyo costs the short-rate differential plus the basis: with Fed funds at 3.50% to 3.75% and the BOJ at 1%, roughly 2.9% all-in, so a 10-year Treasury at 4.65% yields about 1.75% hedged, below the 2.2% on a 10-year JGB at home. A $4tn pool of capital, held out of the world’s most important bond market by one and a bit percentage points.

Now run it forward: Warsh cuts 150bps, the BOJ adds 50, and the hedge cost falls towards 1.25% while the capped long end holds the yield up. The hedged pickup over JGBs swings from half a point negative to nearly a full point positive, and the structural Japanese bid returns.

The controlled steepener is not just a curve trade... the front end is the demand lever too.

And understand who the swap lines are really for, because Washington cannot hand Beijing a swap line directly... so it plumbs the neighbours.

The Japanese and Korean banks fund the Asian dollar credit stack, and the Chinese private sector sits at the end of that pipe with $1.5tn to $2tn of dollar debt, $400bn to $500bn of it rolling over the next year. When the pipe funds cheaply, China’s roll clears; when it tightens, China sells dollar assets, and 2015 showed us what that looks like.

So the swap architecture injects liquidity into the offshore dollar system AND keeps China’s rollover from becoming forced selling.

Then the loop closes: a stronger yuan is what Beijing wants anyway to fight deflation, a smooth roll means the PBOC stops burning reserves, and surplus reserves plus a weakening dollar puts China back at the long-end auctions. What started as US-led Eurodollar plumbing finishes as Chinese reserve buying of Treasuries... in my mind, the single biggest demand-side prize in the whole architecture.

The weaker dollar, then, is not a by-product. It is the demand-creation tool, because the marginal buyer’s constraint was never the dollar’s level; it was hedge cost and FX risk, and a managed decline against the right currencies collapses both.

The dollar falls so that the duration bid can return.

The machine at home

The domestic wiring went in first, in plain sight. The eSLR reform we flagged in the December GMI went live on April 1st, freeing an estimated $1tn-plus of bank balance sheet capacity for exactly one purpose: holding and intermediating Treasuries. The funding mix keeps shifting to bills, which banks buy because they match deposit growth, money market funds recycle, and stablecoins (a $308bn sector holding roughly $200bn of bills under the GENIUS Act) now underpin as the first new marginal buyer of government paper in decades.

When the banking system absorbs government paper, it is monetising the deficit. That is Treasury QE, as we have called it in GMI for two years. In plain English: America is refinancing its mortgage with its overdraft.

Then the volatility channel. As my good friend Michael Howell has long argued, collateral capacity moves inversely with bond vol: a falling MOVE means lower margin against Treasury collateral, more repo capacity, more liquidity. The MOVE is already in the mid-70s. Suppressing long-end vol isn’t a side effect of the buybacks; for the collateral machine, it IS the mechanism.

And the curve itself. Yield curve control was never about a flat curve; the BOJ pinned both ends in 2016 precisely to manufacture a steeper one.

This week’s move was a bull flattening, the cap being installed... with the long end now pinned, every basis point of curve shape from here has to come from the front end.

The steepener arrives via Warsh’s cuts or not at all, and a steeper curve is what re-engages the banks, whose lending creates money on top of all that Treasury collateral with several times the impact of QE. The AI capex cycle cannot be financed without it.

Which brings us to the biggest assumption in this piece: Warsh. He has held at every meeting and cut the Fed’s own bill purchases to zero last week.

But this is what we do in macro... we make assumptions and let the market score them. We assumed he wouldn’t deliver the hikes the market priced in July, and that has already unwound in our favor.

The next assumption is that he cuts in due course, justified by rising productivity, subdued ex-tariff inflation, and unemployment drifting up as boomer retirement drags participation lower.

The first test is Jackson Hole next week, his first as chair: watch for productivity being planted as the anchor of his reaction function.

The near-term catch

Before you extrapolate today into an instant liquidity boom: the RRP buffer that made 2023’s bill issuance an immaculate injection is empty, and the TGA is being rebuilt towards roughly $1tn into late October, a drain an order of magnitude larger than the buyback add. September and October can still deliver a liquidity air pocket. The machine has been built; it has not yet been switched on, and the switch is November 4th plus the first cut.

Do not confuse the wiring diagram with the current.

What it means

For financial conditions: the GMI FCI we have settled on runs two legs, rates and the dollar, with the dollar much the heavier... true in Julian’s original construct and in the new one we are testing in the Everything Code dashboards. Rates just got a ceiling, and the dollar is breaking down, so both legs are easing at once, right now, even while the plumbing stays tight into October.

Markets have been more willing to look through tightness in the old construct (the rates leg barely transmits when the biggest borrowers sit on cash), but this week eases both at once... for the first time in months, the signal is unambiguous.

And FCI is the first of the Everything Code Dominoes: it leads liquidity by around three months and the business cycle by around nine. Gold moves with financial conditions, which is why it was up $160 with the dollar down 0.8%... the relationship printing live on the tape.

Crypto expresses fully on the liquidity lag, so Bitcoin up 6% today is the market front-running a fourth-quarter expression, right where the November 4th switch sits.

For the dollar: Japan showed us the release valve. Cap nominal yields while inflation runs above 3% and the currency wears the adjustment; the yen lost a third of its value under YCC. The cleaner expression is gold against all fiat rather than DXY, and the debasement trade leads while the DXY leg can lag, as it did in 2011.

The bond market, meanwhile, has become a range trade: a state-defined ceiling near 5.3%, a supply-and-inflation floor around 4.5% until the foreign bid returns, and selling long-end vol works until a falsifier fires.

For the business cycle: a shortening WAM couples the interest bill directly to the policy rate, making every Fed decision a fiscal event. This is fiscal dominance built from the plumbing up. Volatility is not being removed from the system; it is being moved out of bond prices and into the budget, and the budget’s answer will always be liquidity.

Now step up to the Universal Code, because this is where the whole piece resolves...

The Universal Code says every system that survives does one thing: it routes energy to intelligence. Energy becomes compute, compute becomes intelligence, and capital flows to whatever performs that conversion most efficiently.

Everything above is the financial system clearing the path for that routing... the old debts roll at the front end, and the long end gets capped so that financing the past stops sucking in the world’s capital, freeing the bond market to finance the buildout: the $600bn of capex, the power, the grid, the data centres.

The buildout is the great game of nations now, and China is running the same play through different mechanisms. Their stronger yuan and our weaker dollar are the same move seen from opposite sides.

Debasement, as written in the Universal Code, is what the Second Law of Thermodynamics looks like in monetary systems... the unavoidable price of keeping the old system solvent while the new one gets built. That sets the routing rule for a portfolio: gold preserves you against the roughly 11% hurdle that debasement plus inflation sets, but the assets that compound past it are the ones doing the energy-to-intelligence conversion itself.

The missing pieces, found

Two dots close the loop... The first is the Fed’s signature, the thing the BOJ lesson says a cap cannot survive without. Look at the plumbing: with the RRP gone, every marginal bill bids directly for bank reserves, which sit only around $250bn above the system’s lowest comfortable level.

Fund at the front at this scale and you eventually get September 2019 again... repo stress that forces the Fed’s balance sheet back into growth whether the chairman likes it or not. The bill machine conscripts the Fed. The signature isn’t a hope; it’s an inevitability with a lag.

The second is the shock absorber to replace the spent RRP. TBAC has floated “TGA repos”: the Treasury deploying its excess cash directly into the repo market.

Conks, whose plumbing work on this is the best there is, has mapped the configurations, from the Treasury lending like a money fund all the way to anonymous repos where the government conceals its own hand.

It is the reverse repo facility rebuilt in mirror image, run from Bessent’s desk, and it reframes the TGA rebuild entirely... the $1tn is not just a drain, it’s the war chest the new facility would lend from.

If it lands, the Treasury will have taken the curve, the currency and the money market floor, leaving the Fed as settlement layer and lender of last resort.

And the deepest dot: the inflation circularity. Easier conditions eventually fire our inflation dominoes, and rising breakevens would put the cap itself to the test.

The only variable that closes that loop is productivity... the entire architecture is a bet that AI-driven productivity delivers nominal growth above funding costs without measured inflation.

This is where the Everything Code hands off to the Universal Code: the debt machine exists to fund the intelligence buildout, and the buildout is the only thing that can eventually retire the debt machine. Two frameworks, one loop, closing in real time.

The bottom line for investors

So does liquidity actually grow because of this? Yes, with a lag. Financial conditions are easing now, liquidity follows on the roughly three-month lead, and Total Liquidity turns up into year-end with the eSLR-unlocked banks doing the heavy lifting and, in time, a conscripted Fed behind them. The September–October air pocket is the entry window, not the exit.

Hyperscaler bonds: the crowding-out problem is being solved for them... the whole point of the architecture is that the buildout gets financed. Long-dated tech paper is the quiet beneficiary; spreads grind tighter, the new-issue concessions are the gift, and these borrowers are becoming the quasi-sovereigns of the intelligence age.

The dollar: a managed decline, not a crash. The break below 99 opens the path towards the mid-90s into year-end and lower into 2027, and every leg down eases financial conditions further while pulling the foreign duration bid back in.

Gold: coincident with FCI, so it stays the lead sled dog while both legs ease, and the March highs above $5,500 are the magnet. It preserves; it doesn’t compound.

Crypto: the highest-beta expression, on the liquidity lag, pointing at the fourth quarter. Today’s 6% in Bitcoin and 9% in ETH is the front-run... the real expression comes when liquidity turns after November 4th.

Equities: supported, with leadership from the energy-to-intelligence complex... power, grid, data centres, semis, quality tech. And watch the banks, because a controlled bull steepener is their oxygen returning.

And bonds you already know: a range trade with a state-defined ceiling. Trade it as one.

The watch list

This thesis is falsifiable, and I want it on record. Watch the buyback results from September 9th: full $4bn-plus fills, offers accepted at or through fair value, and a skew to the longest bonds is price-targeting visible in public data. Watch November 4th: a scale-up or a cut in long coupon sizes turns the metaphor into the mechanism; no change means this was dealer hygiene and I have overread it.

Watch Jackson Hole next week for the productivity anchor, and Warsh after that: a first cut arms the steepener, nothing by Q1 2027 disarms it. Watch 5y5y breakevens: above 2.6% and the “it’s not inflation” foundation cracks. And watch the 30-year: a weekly close through 5.45% with no Treasury escalation kills the put.

The bond vigilantes spent Tuesday toasting a 19-year high. “I told you so. The bond market always wins.” They think they are fighting the Fed’s successor...

They are actually fighting the sovereign’s own balance sheet, its currency desk, its regulator, its banking system and its funding market, all pointed at the same outcome. The 1940s showed how that ends, and this month the first girders went up in public.

Strip it all down to a single sentence: an ageing system is debasing its old promises to fund the construction of a new one. The Universal Code, written in Treasury press releases.

The size was small. The signal was the regime. November 4th will tell us how fast the rest of the building goes up.