Directionally Right, Expensively Expressed

~13 min read. Part one measured what institutional money did to Bitcoin's volatility, and a follow-up set that against gold, the Nasdaq and the S&P. Part two explained why nobody is obliged to own it. This is what that leaves you holding.

Part two ended on an awkward distinction.

Bitcoin has a monetary premium — it produces nothing and has no industrial use, so the entire price accounts for monetary premium. What it lacks is gold's entrenchment, and the obligation that follows from it: central banks hold gold under reserve-management policy, so somebody is always required to buy. Bitcoin carries the premium but without a captive buyer, and so without a floor.

Which sets up this piece. Because in the third week of August 2026, Bitcoin did something that sits awkwardly with almost every explanation ever offered for it.

Long-dated US government bonds were selling off. The thirty-year yield had pushed above 5%. Federal debt had crossed $40 trillion. The Treasury then announced it would at least double its long-end bond buybacks — two weeks after the quarterly schedule had already been published, which is what made it a surprise rather than a formality.

Equities fell. Gold rose about 5%. Bitcoin rose 22.4% over the following week, its best since March 2024.

Set that against the standard explanations. If Bitcoin is a risk asset, it should have fallen with equities. If it is long-duration, higher yields should have hurt it. And if it is digital gold, it did not move like gold — it moved like gold with the volume turned up.

The inflation-hedge reading deserves a pause, because the long end was partly pricing exactly that: unease about the Fed's willingness to contain inflation expectations, sharpened by Warsh's July press conference. But an inflation hedge responds to inflation prints, and this move did not come on one. Alongside the credibility question, the market was pricing something narrower and more uncomfortable — the possibility that the sovereign has lost the appetite to let its own bond market clear.

That is the best backdrop Bitcoin has ever had, and there have been very few like it. April 2025 offered a version of the same thing — tariff-driven dollar weakness, a Treasury selloff, gold bid — and 2020 offered another. Three episodes in seventeen years is not much of a base to generalise from.

So: if Bitcoin cannot obtain a mandate — gold's captive demand from emerging-market central banks building reserves — what is it doing when it tracks the asset that has one?

1. Is Bitcoin digital gold?

Over the full history since 2010, the correlation between Bitcoin and gold is close to zero. That is the most important number in this debate and it rarely appears in it.

Something has shifted recently. The 90-day correlation reached roughly 0.50 in early September 2026, the highest since 2020, while the correlation with the Nasdaq 100 fell to around 0.30, a one-year low. Post-ETF academic work had found the gold relationship stabilising near zero while the equity correlation rose, so this runs against the prior regime.

Whether it is a regime break is another matter, and it would be premature to say so. Ninety days is one observation window, not a structural claim, and a correlation that took six years to appear can disappear in a quarter. What can be said is that the relationship has returned, and that it has not looked like this since 2020.

And the price action supports it more than most commentary has noticed, because both assets have been trading the same trade.

Peaks four months apart. Both bottomed in late June. Both rallied hard in August. This is not one asset working and another failing — it is two assets expressing the same view with different gearing.

Which is where the interesting number is. Bitcoin's drawdown is roughly 1.8 times gold's, and in the August rally it gained 22.4% against gold's 5%. Amplification of four to five times on the way up, roughly twice on the way down. That is the "amplified gold" claim, confirmed — and a claim most of its advocates have not thought through, because amplification is not a feature you get to enjoy in one direction.

But a leveraged proxy is not the same asset

Here the answer has to be precise, because "correlated with gold" and "is digital gold" are being treated as one claim and they are three claims apart.

Amplification is not substitution. A 1.8x drawdown ratio makes Bitcoin a geared expression of the same factor gold expresses — which is a different instrument, not a different asset class. That is the distinction: a leveraged gold ETF also tracks the debasement factor, more violently than bullion does, and nobody calls it digital gold. They call it a leveraged position on gold.

One regime is not an identity. Sixteen years of near-zero correlation, punctuated by two episodes — 2020 and now. The role in question is specific: the asset the world reaches for when it doubts the currency, held by states as reserves and by households as insurance. Gold has occupied it for millennia. Bitcoin is applying for the tenancy, not holding the lease.

And what makes gold gold is not a correlation at all. This is where part two comes back. Gold spent millennia acquiring its monetary standing, and central banks hold it under mandate because of that standing — the obligation is downstream of the status, not the source of it.

Even at a correlation of 1.0, Bitcoin would not be gold, because gold's position is institutional and Bitcoin's is statistical. Bitcoin's claim to the role rests on how it has lately behaved in price data. Gold's rests on reserve policy, accounting treatment and a thousand years of convention — none of which move when the price does. Correlation is something you measure; a reserve mandate is something you are subject to. The first can vanish in a quarter.

So the honest verdict is narrower than either camp wants. Bitcoin is not failing as a debasement asset — it is tracking the same trade as gold, more closely on a 90-day view than at any time since 2020, with roughly twice the drawdown and no compensating advantage in the recovery. Directionally right, expensively expressed, and structurally not the same thing — which is a verdict on the instrument, not on the thesis. If the debasement view is correct, Bitcoin has been expressing it faithfully. It has simply been charging more for the privilege than gold does.

Two implications for anyone sizing it. If Bitcoin is genuinely becoming the vehicle rather than the geared proxy, each successive debasement episode should see it capture more of the move relative to gold — a clean forward test worth tracking.

And the number to compare is never the correlation; it is the drawdown ratio. Correlation would have told you all through 2026 that the trade was working. The levels tell you what holding it that way cost: same view, same factor, twice the drawdown, which means less than half the position for the same risk budget and a far better chance of being stopped out before the thesis paid.

2. Correlation, volatility, and why beta is the number you want

Correlation is not a property of Bitcoin. People quote "0.4 to equities" as though it were a constant. It is a measurement of a past window, and what it measures is which shock dominated that window and who happened to be buying. The same asset printed roughly 0.74 against the S&P in March 2026 — with intraday co-movement so tight the fit reached 0.94 — close to zero in August, and around −0.30 in late 2025. Bitcoin did not change. The dominant macro factor did.

Two numbers, answering different questions. Correlation is scale-free and asks: do they move together, and how reliably? Beta is the regression slope, carries units, and asks: when equities move 1%, how much does Bitcoin move?

Beta = Correlation × (Bitcoin's volatility ÷ the other asset's volatility)

Beta has two moving parts, driven by different things. The macro regime sets the correlation — which shock dominates, and whether it hits both assets the same way. Flows and positioning set Bitcoin's volatility — leverage, depth of book, whether a marginal buyer exists. And as part two argued, whether a marginal buyer exists is a structural question, not a sentiment one.

Which produces results that look paradoxical until you write the identity down. High correlation, low beta: if Bitcoin's volatility compresses — locked float, deep options market, little leverage — it can track equities faithfully while moving less per unit of equity move. Reliable, and small. Low correlation, high beta: roughly where August 2026 sat, with correlation collapsing toward zero while realised volatility rose on thin liquidity and heavy short positioning.

You can decorrelate and become more dangerous at the same time.

That undercuts much of what is written about decoupling. Decoupling is reported as a maturity signal. Mechanically it is just as often a liquidity signal — and thin markets do not drift, they gap.

The practical rule: never use a single trailing correlation as a portfolio input here. Identify the regime, then use the correlation and the volatility conditional on it. When someone reports Bitcoin is now uncorrelated, ask what its volatility did over the same window.

3. A debasement trade — and the curve tells you which day it is

The most useful thing to know about Bitcoin and interest rates is that it depends on which end of the curve is moving, and why. Not the level. Not the direction. The driver.

A caveat about this table before you use it. It is built on seventeen years of price history, most of it from a market too small to have a macro regime at all, and several of these cells rest on one or two episodes. Regime 2 in particular has arguably two clean instances. The logic is sound and the examples are real, but this is a framework for organising what you observe, not a model with enough observations behind it to forecast from.

Regime 1 is where Bitcoin is hurt worst, for a specific reason. Equities also suffer when real rates rise — but if rates rise because growth is strong, earnings rise and cushion the blow. Bitcoin has no earnings channel. It takes the full discount-rate hit with nothing offsetting it.

Which is why the regime splits in two. When rates rise hard enough to break equity valuations, everything falls together and correlation goes positive: that is 2022. When rates rise into growth strong enough that equities absorb it, Bitcoin still takes the hit while equities hold up — correlation turns negative and Bitcoin underperforms anyway. Negative correlation is not always a compliment.

Regime 2 is the only one that has fully delivered on the thesis. Long yields rise not because the economy is strong but because investors doubt the sustainability of the borrowing and want more compensation to hold the paper, and the sovereign simultaneously signals it will lean against that — which, whatever the official framing, describes the August buyback announcement.

The clearest regime marker available is the sign of Bitcoin's response to a long-end selloff. In 2022 higher long yields hurt it. In August 2026 a bond selloff drove it higher. Same asset, opposite sign, because the driver changed. One indicator, no model required.

Two caveats. The August magnitude was mechanical — roughly $2.75 billion of crypto short positions were liquidated across the market, with Bitcoin accounting for around $1.7 billion, Bitcoin up 12% in 48 hours. The regime set the sign; the positioning set the size. And sudden decorrelations from equities during bond selloffs have historically been short-lived, marking local exhaustion more often than a structural shift. Regime 2 has been a visitor, not a resident.

Worth noting what did not drive it: there was no liquidity injection. The Fed's balance sheet was contracting, no reserve-management purchases were scheduled, and stablecoin supply was flat at around $302 billion. What the market bought was a signal about who blinks when the long end misbehaves — whether the sovereign tolerates higher yields and lets the market clear, or steps in to cap them. Buybacks suggested the second.

4. An asset with no maturity date

Bitcoin is often compared to a long-duration technology stock. The intuition is sound; the mechanics break in a way that explains the underperformance pattern.

A growth company has an identifiable numerator — a business, revenue, a plausible path to profit — so an implied duration can be computed, because cash flows arrive at times. Bitcoin has no cash flows. Gold at least has jewellery and industrial demand underneath; Bitcoin has nothing. No payments to weight, no dates to weight them by. Bitcoin's duration is not long. It is undefined. Nor is it rescued by saying the payoff is simply distant. An instrument that promises to pay nothing, forever, has infinite duration on paper — and duration has stopped meaning anything useful at that point, because there is no schedule left for a discount rate to act on.

What Bitcoin has is an observed sensitivity to real rates — measured by regressing returns on rate moves, not calculated from a payment schedule the way a bond's duration is — and that estimate is unstable, which is the finding rather than a measurement problem. Price sensitivity has two parts: the discount rate, which moves slowly, and the probability the market assigns to Bitcoin's eventual monetary role, which moves violently. Gold's equivalent term moves too — its volatility rose by two thirds this year and it fell 29% from its January high. But gold is repricing how much its monetary role is worth. Bitcoin is repricing whether it has one at all, and that is a far wider distribution to sit on.

One correction to the usual formulation. It is tempting to say Bitcoin underperforms equities outside the debasement trade, in both risk-on and risk-off. The risk-off half holds — Bitcoin is the highest-volatility line in any book, so it is cut first when risk budgets shrink. The risk-on half does not survive the history: in 2020–21, textbook risk-on, Bitcoin comfortably beat equities.

The accurate statement is narrower: the rate regime determines the correlation; the flow picture determines the relative performance. Bitcoin underperformed a rallying equity market in 2026 not because of duration but because its marginal buyer had gone — the largest corporate buyer holding cash, retail net selling, the carry trade no longer paying. A flow story in macro clothing, and conflating the two is how frameworks end up explaining last year and nothing else.

Where this leaves us

Three pieces, one argument.

Bitcoin has assembled the apparatus of a real asset class: custody, a wrapper, a legal classification, a derivatives complex. What it does not have is a rule obliging anyone to own it, and on the evidence of gold's fifty-two years it may never get one through the portfolio door.

Market structure has done real work: better custody, unlevered buyers, a functioning derivatives complex, and a return distribution that now looks more conventional than the S&P 500's. What market structure cannot do is supply a buyer who has no choice. That absence is why the sheer size of Bitcoin's moves has not budged in four years, why the correlation to everything is unstable, why the same asset can be 0.74 to equities in March and zero in August, and why decoupling episodes should be read as liquidity events rather than maturity milestones.

It is also, in the end, the answer to the digital gold question. Bitcoin genuinely trades the debasement factor, and on the most recent readings it tracks it more closely than at any time since 2020. But it trades it as a geared proxy — twice the drawdown, four to five times the rally — rather than as the asset itself. The difference between the two is not a correlation coefficient. It is a mandate.

None of which is bearish. Bitcoin's best regime — sovereign balance sheets under strain, long yields rising for the wrong reasons, governments visibly reluctant to let their bond markets clear — looks more probable over the next decade than at any point in the last four. If that is your view, this is the instrument that expresses it with the most convexity available.

Just be clear that convexity is what you are buying, and that you are paying for it in drawdown rather than in fees.