What the Treasury did, precisely

The mechanism matters here, so it is worth stating exactly rather than loosely.

The Treasury announced it would double the size of its liquidity support buyback operations for longer dated nominal coupon securities, taking each operation from $2 billion to at least $4 billion. The change applies to two sectors of the curve, 10 year to 20 year and 20 year to 30 year, and it runs from September 9 through November 4, 2026.

In plain terms: the government is going to buy back its own long dated debt at twice the previous scale, to take pressure off the far end of the curve. This is debt management, not monetary policy. No interest rate changed. The 30-year yield fell almost 10 basis points to around 5.187%, from a level that had been roughly its highest in 19 years.

Note the calendar detail, because it is a strange one. The operations do not start until September 9. Nothing has actually been bought yet. The entire move today came from the announcement of future buying, which tells you how thin the bid at the long end had become.

Three weeks on the long end, closed by the wrong institution

This is where the day becomes more than a price move, at least for anyone who has been reading along.

Date30-year yieldWhat we wroteBitcoin
Jul 295.21%, up 10bpA credibility trade, not a rate path tradeRose while equities fell 2.1%
Jul 31Near multi decade highsA hedge unwind, gold and Bitcoin sold togetherFell 2.92%, lost $62,500
Aug 175.29%, highest since 2007The co-movement has inverted, we cannot say which holdsGained 2.59%, cleared $64,000
Aug 195.187%, down 10bpThe Treasury did it, not the FedGained 6.07%, printed $69,500

The 30-year has now round tripped. It sits below the 5.21% level where our coverage of it began on July 29, and it got there without the Fed touching anything. On July 29 we described a market pricing a central bank that would be later rather than different, and said the bond market was the group that had to be convinced. It was not convinced by the Fed. It was handed a buyer.

There is a real analytical point in that, and it is not a comfortable one. For three weeks the honest reading of a 5.2% to 5.3% long bond was that inflation credibility was impaired and the Fed was tolerating it. Today's move says something narrower and more mechanical: a meaningful part of that yield was a liquidity and absorption problem at the far end of the curve, which a large enough buyer can relieve without any change in the inflation outlook. Both readings can be partly right. We had been weighting the first one heavily, and today is evidence that we were underweighting the second.

And it happened on FOMC minutes day. The minutes for the July 28 and 29 meeting are due at 18:00 UTC, which at the time of writing is still ahead of us. The largest crypto move in months arrived about three hours before the event the entire market had circled, and from a different building.

Monday's ETF number arrived, and it settles yesterday's question

Yesterday we set out one open item and promised to report it either way. Here it is.

US spot Bitcoin ETFs took in about $297.5 million on August 17. The breakdown was IBIT $160.2 million, FBTC $111.9 million, ARKB $14.2 million and MSBT $11.2 million. That single day recouped roughly 35.6% of the preceding five session outflow of about $385.2 million.

August 17 is precisely the day we reported a 50.4% taker buy share on Binance spot against a 2.59% price gain, and wrote that a neutral exchange split is not a quiet market but a market being priced somewhere our instrument cannot see. The number confirms it. Nearly three hundred million dollars of demand arrived through the fund wrapper on the day the exchange tape read a coin flip.

One caution that belongs in the same breath, because it is the same class of error we have been correcting all month. The figure was first reported as $137.3 million, before BlackRock's data was in. Anyone who read the early print got less than half the real flow, and would have concluded institutional demand was modest when it was the largest single day in the period we have been tracking. Partial data does not announce itself as partial.

Tuesday was a different story at roughly $19 million on preliminary figures, against a 0.30% price gain. So the sequence is $297.5 million, then $19 million, then today. Today's own figure will not publish for hours.

Today the exchange flow led, and that is the contrast worth keeping

Two big up days, three days apart, and our metric behaved completely differently in each. That contrast is more instructive than either day alone.

SessionPriceBinance buy shareWhere the demand was
Mon Aug 172.59% higher50.4%, a coin flip$297.5M of ETF creations
Wed Aug 196.09% higher, partial55.6%, and 61.8% in the key candleAggressive buyers on the exchange

Today the tape led. The 15:25 candle took $203.3 million at a 61.8% buy share and moved price 1.62%. The 15:20 candle before it ran 62.9%, the 15:00 hour aggregated $628.4 million at 58.3%, and the 20 hour figure is 55.2% on $1.32 billion. Ether cleared $2,000 at 15:05 on a 64.9% candle. These are the readings of a market reacting to news in public, at speed, by lifting offers.

So the metric is not broken and it was not broken on Monday either. It measures aggressive flow on one venue, and the two days differed in where the buyer was. Fund accumulation arrives quietly and off-exchange and leaves the split neutral. A genuine surprise arrives on-exchange and the split moves immediately. Knowing which regime you are in is most of the value.

Yesterday we wrote that when the split is neutral and the price is not, you should conclude you are looking at the wrong venue. Today is the control case for that claim: when the split moves hard with the price, the exchange is the right venue, and you can read it straight.

Everything went up, which is how you know it was rates

The breadth is the tell. This was not a crypto story.

Asset24h changeTaker buy share, 24h
Ethereum9.37% higher52.3%
XRP6.57% higher49.1%
Solana6.39% higher53.6%
Bitcoin6.07% higher55.6%
Tether Gold2.72% higher51.4%
Paxos Gold2.67% higher49.1%
BNB2.52% higher48.9%

Tokenised gold up more than two and a half percent alongside crypto is the part that rules out a crypto specific explanation. Reporting puts spot gold near $4,487 and silver near $65.60, both up more than 3.5%. When bullion and Bitcoin and altcoins all rally together on the same afternoon, the variable that changed is the discount rate, not anything about blockchains.

Total crypto market capitalisation is $2.412 trillion, up 5.11%, with Bitcoin dominance at 57.13% and Ether at 10.5%. Bitcoin now trades 2.75% above the July 21 high of $66,956, so this is not merely an August breakout. The whole July range is behind it, on a two month high.

Sentiment has finally caught up, fast. The Fear and Greed index reads 46, from 41 yesterday and 29 three days ago. That is 17 points in three sessions and the highest of this entire stretch. It is still, technically, labelled Fear.

The Ether over Bitcoin ratio now reads 0.030519, above 0.03 for the first time in the series we have been publishing. We are reporting the number and, for the fourth consecutive article, not calling it a trend. The last two times we held that line the series turned within a day, once against us and once for us.

What is scheduled, starting in about ninety minutes

Scheduled events and things we will measure, with no thresholds attached that claim to predict a direction.

  • FOMC minutes, today at 18:00 UTC. Still ahead of us at the time of writing. The written record of the 9 to 3 hold, and what Hammack, Kashkari and Logan put on paper after CPI at 3.4% headline and 2.5% core, PCE easing to 3.3%, GDP at 1.5% and a weak July jobs print. Note that the market has just repriced the long end without them.
  • Jackson Hole, August 21 and 22. Chair Warsh's first symposium in the job, now with a Treasury that has moved the curve for him two days beforehand.
  • Today's ETF flow, published tomorrow. Monday was $297.5 million and Tuesday about $19 million. Whether a 6% day pulls fund money in or lets holders sell into strength is the next real datapoint.
  • September 9, when the buybacks actually start. Today priced an announcement. The operations themselves begin on the 9th and run to November 4, and nothing has been purchased yet.
  • Whether the 30-year stays below 5.21%. That is the level our coverage started at on July 29. It is the cleanest single test of whether today was a repricing or a relief rally.
  • September 15, still doubled. The reported CLARITY procedural vote and day one of the Fed meeting that brings back the dot plot.

How to read this as a trader

Two things, and the first is the reason we have been writing about Treasury yields on a crypto site for three weeks.

The catalyst was not on the crypto calendar, and it was not even on the Fed's. Every preview published for today pointed at 18:00 UTC. The forecasts we quoted yesterday morning had Bitcoin eyeing $64,000 and Ether testing $1,900 ahead of the minutes; within hours Bitcoin printed $69,500 and Ether $2,115, and the minutes had not been released. The move came from a debt management announcement out of the Treasury, an operation that does not begin until September 9. If you were watching only crypto news and only the Fed, you saw a 6% candle with no explanation attached to it.

The generalisable version: for an asset whose main driver is the discount rate, the relevant calendar includes the fiscal authority, and auction and buyback schedules are part of it. That is unglamorous and it is where today's money was.

Second, hold your framework loosely enough to update it. We spent three weeks arguing a 5.2% long bond was an inflation credibility problem. Today's ten basis point drop on a buyback announcement is evidence that a real part of it was a liquidity and absorption problem instead, which is a much more fixable thing. That does not make the earlier reading worthless, and it does mean we were weighting it too heavily. Writing that down is cheaper than defending it. Our live dashboard separates aggressive buying from aggressive selling minute by minute across the most liquid Binance pairs, which is how the 15:25 candle was visible as a 61.8% buy print while it was happening rather than afterwards. For the arc behind this, see Monday's coin-flip breakout, the venue limitation in full, the Fed hold that started the 30-year story, and the day Bitcoin fell with gold at the same yields. If the buy and sell split is new to you, the volume guide covers the mechanics.