The Fed Is Buying Again

The consensus story is a weak dollar. The data says something more interesting. The dollar is flat this year. What changed is that the Federal Reserve stopped shrinking its balance sheet and started buying again, and almost every risk asset bottomed within three weeks of each other.

Start with what is measurable

Most market commentary this year has leaned on a weakening dollar. It is worth checking that claim before building anything on top of it. The dollar index fell 9.5 percent across 2025. That was a genuine repricing. In 2026 it has done almost nothing: the index is up 1.1 percent year to date, and the Federal Reserve's own broad dollar measure is close to unchanged.

So if you are waiting for a falling dollar to explain what gold and digital assets did this summer, you will be waiting a long time. The dollar is not the story of 2026. Liquidity is.

The quiet turn: the Fed is a buyer again

Since February the Federal Reserve has run 38 outright purchase operations in the Treasury market, taking in roughly 219 billion dollars of securities. Its holdings of Treasury bills have gone from 195 billion dollars to 542 billion in twelve months, an increase of 177 percent. The total balance sheet is growing again for the first time in years.

Two points of precision, because they matter and because most coverage gets them wrong. First, this is bill buying, not long bond buying. The Fed is adding short paper, which is a reserve management operation, not a duration play. Second, quantitative easing is a specific policy with a specific purpose, and this is not it. Calling every balance sheet expansion money printing is lazy, and it makes you miss what is actually happening.

What is actually happening is simpler and more durable. The system needed reserves. The Fed is supplying them. Reserves are the raw material of risk taking, and risk assets noticed.

The debt arithmetic underneath

Total public debt now stands at 40.10 trillion dollars, 7.6 percent higher than a year ago. The average interest rate across all marketable Treasury debt is 3.443 percent and it is climbing, because debt issued in the cheap years keeps maturing and rolling into today's rates.

This is the part that does not resolve itself. Every month, more of the outstanding stock reprices upward. Interest becomes one of the largest single lines in the federal budget. That creates a permanent institutional preference for conditions where debt is easier to carry: lower real rates, controlled front end yields, and a tolerance for inflation running slightly warm rather than slightly cold.

You do not have to believe in any conspiracy to see the consequence. When the largest borrower in the world has a structural interest in cheaper money, the long run purchasing power of that money is the variable that absorbs the pressure. Gold has understood this for four thousand years. Assets with fixed supply schedules understood it more recently.

The tell: yields rose anyway

Here is the detail that separates a real thesis from a comfortable one. The Fed has been buying, and yields went up regardless. The two year yield is up about a quarter from where it started the year. The ten year sits near four and three quarters percent, well above its January level.

A central bank buying while yields rise is not a contradiction. It is a signal. It says the market is demanding more compensation to hold government paper, and that demand is coming from the fiscal side rather than the monetary side. That is a term premium widening, and historically it is one of the more reliable backdrops for hard assets, because it means investors are questioning the terms rather than the timing.

Everything bottomed in the same three weeks

Ethereum bottomed on 26 June at 1,566 dollars. Bitcoin and Chainlink both bottomed on 1 July, at 58,566 and 7.19 dollars. Gold bottomed on 17 July near 3,982 dollars.

Four assets with different holders, different narratives and different liquidity profiles do not bottom inside three weeks by coincidence. They bottom together when the thing they have in common changes, and the thing they have in common is the cost and availability of money.

Since those lows: Ethereum is up about 58 percent, Chainlink about 58 percent, Bitcoin about 34 percent, gold about 12 percent. That is what a liquidity turn looks like from the inside, and it is worth saying plainly that most people missed it because they were waiting for a headline instead of watching operations.

Gold: the correction was not the end of the trend

Gold ran to 5,496 dollars in late January. It then gave back a quarter of its value into the July low at 3,982 dollars, which is a violent move that convinced a lot of people the trend was over. It now trades near 4,447 dollars.

The structural case did not change during that correction. Central banks kept accumulating. The debt arithmetic above did not improve. What changed was positioning: a crowded trade cleared itself out, which is normally what a market does before it can continue rather than a sign that it cannot.

Levels that matter. Reclaiming the January high near 5,496 dollars requires roughly 24 percent from here, and that is the level that would confirm the primary trend intact. Holding above the July low near 3,982 dollars is what keeps the constructive case alive at all. If the fiscal picture stays as it is and the Fed keeps supplying reserves, a move through the January high opens the ground above it. If reserves are withdrawn again, the July low is where this gets retested.

Digital assets: still trading below the last cycle

It is important to be honest about where these assets actually are, because the recovery has been strong enough to make people forget. Bitcoin trades around 78,761 dollars, roughly 38 percent below its October 2025 high, and it would need to gain about 60 percent to reach that high again. Those are two different numbers and confusing them is how people end up with the wrong expectations. Ethereum near 2,470 dollars is still about half its record. Chainlink and Stellar remain far below the peaks they set in earlier cycles.

That is not a reason to dismiss them. It is the reason the arithmetic is interesting. An asset that has already corrected deeply, in a liquidity environment that is improving rather than deteriorating, has a different risk profile from one making new highs on enthusiasm.

The levels, asset by asset

What follows are levels, not promises. Each one is anchored to a price this market has actually traded at, which is the only honest way to talk about targets. The first number is the recovery level that would confirm the move. The second is the previous peak, which is where the argument would have to be reassessed rather than extended.

Prices in US dollars as at 1 September 2026. Percentages measure the distance from the current price to that level.

The September question: buying the long end while raising the price of money

Everything above describes what has already happened. The more interesting question is what happens next, and there is a specific combination being discussed for the autumn that deserves to be thought through properly rather than reacted to.

The combination is this. The Federal Reserve extends its purchases from bills into the long end of the curve, buying ten, twenty and thirty year Treasuries in the open market well before they mature. At the same time, or close to it, the policy rate goes up by at least a quarter of a point.

The first reaction of most readers will be that these two things contradict each other. Buying bonds is easing. Raising rates is tightening. You cannot do both. That reaction is wrong, and understanding why is the single most valuable thing in this article.

This combination has a name, and a history

Raising the short rate while capping the long rate is not a contradiction. It is a policy, and it has been run before. The Federal Reserve did it in the 1940s, holding long bond yields near two and a half percent while inflation ran into the high teens. Japan did a version of it for years under yield curve control. The technical name is financial repression, and the purpose is always the same: make the debt bearable by holding the cost of the long end below the rate of inflation, and let time and inflation erode the real value of what is owed.

Look at why it would be attractive right now. The thirty year Treasury yields 5.22 percent, up from 4.86 percent at the start of the year. The twenty year is at 5.21 percent. With 40.10 trillion dollars of debt outstanding and an average coupon already climbing, every additional basis point at the long end compounds into the budget for decades. A Treasury cannot refinance comfortably into a five and a quarter percent thirty year. So the long end becomes the thing you control, and the short end becomes the thing you use to look serious about inflation.

That is the regime. Hawkish at the front, quietly accommodative at the back, and negative real returns for anyone holding the long bond to maturity.

What the curve is already telling you

Before predicting a rate rise it is worth checking whether the market has already priced one. The two year Treasury yields 4.34 percent while the upper bound of the policy rate is 3.75 percent. The two year sits roughly 59 basis points above the policy ceiling.

That gap is the market's forecast. It is already pricing more than one quarter point of tightening. A twenty five basis point rise, on its own, would therefore not be a shock. It would be a confirmation. This matters enormously for how assets react, because markets do not respond to events, they respond to the difference between events and what was already expected.

The curve is also very flat. The ten year sits only 0.39 points above the two year, and the thirty year only 0.88 points above it. A flat curve with a central bank about to buy the long end is a curve that wants to invert further at the front and be pinned at the back.

The mechanism that actually drives the price of gold

Gold does not respond to interest rates. It responds to real interest rates, which is the nominal yield minus expected inflation. That single distinction explains most of the times people have been wrong about gold.

The ten year real yield is currently 2.42 percent. That is a genuinely restrictive number, and it is remarkable that gold trades near 4,447 dollars in spite of it. Gold has been rising in an environment that historically should have suppressed it, which tells you the demand is coming from somewhere other than the rate calculation: central bank accumulation, reserve diversification, and a growing preference for an asset that is nobody's liability.

Now apply the scenario. If the Federal Reserve buys the long end, the nominal long yield is capped or pushed down. If a quarter point rise fails to bring inflation down, or if inflation expectations rise because the market reads bond buying as monetisation, then the numerator falls while the subtraction grows. Real yields fall, possibly sharply, possibly through zero at the long end.

A collapse in long real yields from 2.42 percent toward zero is the most reliably bullish configuration for gold that exists in the historical record. Not a probable configuration. A reliable one, if it occurs.

Gold and silver: the scenario arithmetic

Gold trades near 4,447 dollars. The January high at 5,496 dollars is roughly 24 percent above here, and in this scenario that level is a waypoint rather than a ceiling, because the condition that produced the January spike would be present again and stronger. If long real yields fall by a full point from current levels, the historical relationship between gold and real rates argues for a move well beyond the January high rather than a stall at it.

Silver is the leveraged expression of the same trade and it behaves very differently. It trades near 67 dollars, putting the gold to silver ratio at roughly 66. In monetary metal rallies driven by liquidity rather than fear, that ratio compresses, because silver is a smaller market with an industrial demand floor underneath it and far less central bank supply overhead.

The arithmetic is worth stating plainly. If gold reclaims the January high near 5,496 dollars and the ratio merely holds at 66, silver reaches about 83 dollars. If the ratio compresses to sixty, which is unremarkable in a metals bull phase, silver reaches about 92 dollars. If it compresses to fifty, which has happened repeatedly in past cycles, the number is about 110 dollars.

That is the case for silver being the higher return and higher volatility instrument in this regime. It is also the case for sizing it smaller, because a ratio that compresses in a rally expands violently in a liquidation, and silver falls roughly twice as fast as gold when liquidity is withdrawn.

Digital assets: the sequencing is the whole answer

This is where most analysis goes wrong, because it treats the rate rise and the bond buying as one event. They are not. They hit at different speeds, and the order matters more than the direction.

The first move on a rate rise is almost mechanically negative for digital assets. Higher policy rates raise the discount rate applied to assets with no cash flow, the dollar typically firms on the announcement, and leveraged positions get liquidated into the volatility. Expect the initial reaction in Bitcoin to be a sharp move down rather than up, and expect the higher beta assets to fall considerably more. If that happens, it will be loud, it will feel like the thesis is broken, and it will be the least informative part of the entire sequence.

The second move is the one that matters. Long end purchases inject reserves that do not disappear, and they do so continuously rather than in one announcement. Liquidity is a flow, and flows beat headlines over any horizon longer than a few weeks. This is precisely what happened this summer: the bill purchases were running from February, and the market did not turn until late June. The mechanism led the price by four months.

So the honest prediction is a shape, not a number. A hawkish first reaction lasting days to a few weeks, followed by a liquidity driven recovery that is stronger than the initial drop, provided the purchases continue. The risk to that view is that the drop triggers enough forced deleveraging to overwhelm the flow, which is exactly what happened in March 2020 before the same flow then produced the largest rally of that cycle.

The levels I am watching in that scenario

For Bitcoin, the first test is the 2026 high near 96,899 dollars, which is about 23 percent above here. Reclaiming it puts the October 2025 record at 126,080 dollars back in play, a further 37 percent beyond that. In a sustained long end purchase programme, a new record during this cycle is a reasonable expectation rather than an aggressive one, because the supply schedule does not change while the quantity of money does.

Ethereum needs to reclaim 3,352 dollars first, roughly 36 percent from here. Its record near 4,946 dollars requires about 100 percent. Ethereum tends to outperform Bitcoin in the later stage of a liquidity expansion rather than the first stage, so early underperformance would be normal and not a signal.

Chainlink at 11.35 dollars has the widest gap between its 2026 high at 14.42 dollars and its record at 52.70 dollars, which is the profile of an asset with large upside and correspondingly large risk of never reaching it. XRP needs about 70 percent to reach its 2026 high of 2.3537764079779833 dollars. Stellar near 0.1776 dollars is the smallest and most speculative of the group and should be treated that way, whatever the percentage tables suggest.

One discipline worth repeating. The distance to a previous peak is not a forecast that the peak will be reached. It is a measurement of how much has already been lost. Confusing the two is the most common and most expensive error in this market.

What would make me wrong about September

If the Federal Reserve raises rates and does not begin meaningful long end purchases, this scenario collapses into ordinary tightening, and ordinary tightening is bad for every asset discussed here. That is the single largest risk to the argument, and as of today the evidence for the purchase programme is thin: exactly one outright coupon purchase has been executed so far, against fifty five bill purchases. The transition I have described is at its very beginning, if it is happening at all.

If inflation falls quickly, the entire repression logic loses its urgency, real yields rise, and gold gives back a large part of this year's gains. If the rate rise is larger than twenty five basis points, the initial deleveraging in digital assets could be severe enough to break the market structure rather than simply shake it.

And if the long end refuses to be controlled, meaning the Fed buys and the thirty year yield rises anyway, that would be the most important signal of all. It would mean the market is repricing sovereign credit rather than monetary policy, and in that case gold rises for a reason nobody should want to celebrate.

What would prove this wrong

A thesis without an invalidation level is marketing. Here is mine.

If the Federal Reserve stops its purchase operations and the balance sheet resumes shrinking, the mechanism described in this article is gone and everything above should be reconsidered from scratch. If inflation reaccelerates hard enough to force the front end sharply higher, the same applies: rising real rates are the historical enemy of both gold and long duration risk assets.

On the price side, the July lows are the line. Bitcoin below 58,566 dollars, Ethereum below 1,566 dollars, gold below 3,982 dollars: any of those would say the liquidity turn was a pause in a larger decline rather than the start of a recovery. I would rather state that in advance than explain it afterwards.

How I am thinking about it

The dollar is not collapsing and I am not going to pretend it is. What is happening is slower and, in my view, more consequential: a borrower with 40.10 trillion dollars of obligations and a rising average coupon needs a monetary environment that makes that burden manageable, and the central bank has quietly moved back to supplying rather than draining.

Assets with fixed or predictable supply are the natural expression of that view. Gold is the version with four thousand years of evidence behind it. Bitcoin is the version with a mathematically fixed cap. The rest of the assets discussed here are higher risk expressions of the same trade, and they should be sized accordingly, because a liquidity tide that lifts them will also drop them fastest if it goes out.

Positioning is a personal decision that depends on your circumstances, your time horizon and what you can afford to lose. Nothing here is a recommendation to buy or sell anything.