Ninety-eight point eight one seven.
That is the whole of it. The dollar index rose three hundredths of one percent on the ninth of September and closed at 98.817. The dispatch carrying this information runs four lines. It names no counterparty currency, quotes no official, gestures at no policy meeting, and declines, notably, to name a year. It is a tick, transcribed, and it has been handed to us as news.
Sit with it anyway. Not because it matters, but because of how little it takes.
Arithmetic is the only honest place to begin. Three hundredths of one percent of 98.817 is 0.0296 index points. Call it three cents on a hundred-dollar bill, printed to the third decimal because the third decimal is what the instrument publishes. Translate that into the underlying crosses and the movement becomes even smaller than it looks. The euro carries the largest weight in the basket by a wide margin, so a 0.03% move in the index implies roughly a 0.05% move in EUR/USD, which at a rate near 1.10 is something on the order of half a pip. During a London afternoon, the spread between the best bid and the best offer on that pair is frequently wider than the entire move. The index did not rise so much as it failed to fall.
So the first honest observation is that this is not information. It is a reading of the noise floor, published as though the noise floor were a signal, and consumed as though the consumption of it were a form of diligence.
But the noise floor is not nothing. A microphone left running in an empty room still records the shape of the room. What it captures is not music, but it is reverb, and reverb is a property of architecture. A single unremarkable print on the dollar index is a pulse taken at one point in a system that most people who trade crypto have never actually looked at: the offshore dollar market, the repo market that sets the funding price of everything, and the increasingly peculiar question of what a stablecoin is when you stop calling it a stablecoin and start calling it what it is, which is a dollar claim issued by a company rather than a treasury.
I want to spend this essay on that architecture. There is a number involved, and the number is almost meaningless, and that is precisely the point. The absence of signal is the signal. Echoes of early hype in the quiet of current data — and quiet, in my experience, is the only condition under which structure is legible.
II. What the Instrument Measures, and What It Refuses To
The dollar index is old by financial standards. It was constructed in 1973, in the wreckage of the Bretton Woods arrangement, as a way of pricing the dollar against a fixed basket of peers. For its first three years it was a geometric weighted average; in 1976 it was redefined as a fixed-weight index, and it has only been structurally revised once since, when the introduction of the euro in 1999 collapsed the legacy European currencies into a single line item. Everyone reading this has been alive for exactly one composition change of the world's most-quoted currency benchmark.
The weights are worth stating plainly, because they explain most of the index's behavior and almost none of its reputation. The euro accounts for 57.6% of the basket. The yen, 13.6%. Sterling, 11.9%. The Canadian dollar, 9.1%. The Swedish krona, 4.2%. The Swiss franc, 3.6%. Add them and you have the whole dollar, as far as this instrument is concerned. There is no Chinese renminbi in the basket. There is no Korean won, no Singapore dollar, no Brazilian real, no Indian rupee, no Mexican peso. The dollar index is a measure of the dollar against a club of rich-country currencies, most of which are themselves satellites of the dollar system, and it is quoted as though it were a measure of dollar strength in the world.
That gap between what the index is and what it is used for is not a technicality. It is a full accounting of why the number is nearly worthless as a macro signal and nearly indispensable as a plumbing diagram.
When the index rises three hundredths of a percent, what has actually happened is that a weighted mix of six exchange rates, dominated by a single currency pair, drifted by an amount roughly equal to the bid-ask spread. There is no mechanism by which a move of that size reflects a change in the Federal Reserve's reaction function, or the Treasury's issuance calendar, or the disposition of foreign official reserves. It reflects the atomization of order flow across venues, the timing of a snapshot, and the ordinary asymmetry of a market where liquidity is provided by a handful of firms who reprice continuously.
I have spent a decade and a half watching these instruments, first as an undergraduate pulling apart token sale documents that promised monetary policy, later auditing the curves that DeFi uses in place of one. The consistent lesson is that a number's meaning is a property of its construction, not its publication. A 0.03% print on a fixed-weight index of six developed-market currencies is a measurement of developed-market currency noise. Every headline that reads it as a verdict on anything else is performing an act of interpretive inflation.
And yet. And yet the index is used, constantly and by serious people, as the denominator of global liquidity. Bitcoin's correlation to it in 2022 was among the tightest relationships that asset has ever exhibited to any external variable. So the instrument is simultaneously meaningless as a measurement and load-bearing as a convention. That is not a contradiction. It is a description of how macro trading works: participants agree on a proxy and then behave as though the proxy were the thing, at which point it becomes the thing. Reflexivity is not a flaw in the system. It is the operating principle.
Which brings us to the plumbing.
III. The Offshore Dollar, Which Is Where Crypto Actually Lives
Here is a fact that does not appear in the four-line dispatch, and that governs far more of crypto's daily behavior than any print on the index ever will. The majority of dollar-denominated credit in the world is not issued by the United States. It is issued outside it.
The Bank for International Settlements has spent years trying to put a defensible lower bound on the size of this offshore dollar system, and the number keeps coming out somewhere in the neighborhood of thirteen trillion dollars in off-balance-sheet dollar obligations — obligations created through FX swaps and forwards, instruments that do not appear on any bank's balance sheet as a loan but function, economically, as one. The daily turnover of foreign exchange markets, as of the last triennial survey, sits in the vicinity of seven and a half trillion dollars a day, the overwhelming majority of which is not trade finance, not tourism, and not investment. It is funding. It is the machinery by which institutions borrow one currency to lend another, roll a position, hedge a book, and satisfy a regulatory ratio at four in the afternoon.
This is the market that actually sets the price of a dollar. Not the index. The index is a thermometer taped to the outside of the building. The repo market, the FX swap market, the cross-currency basis — these are the heating plant.
Anyone who has watched crypto's liquidity cycles closely has watched this market, whether they knew it or not. The mechanism is not mysterious once you name it. When the cross-currency basis widens — when it becomes expensive for a non-US institution to borrow dollars via swap — the cost of carrying a dollar-funded position rises. When it widens sharply, as it did in March 2020 and again in the autumn of 2022, dollar-funded positions get unwound whether or not their holders want them unwound. Assets with the most leverage and the least natural buyer base unwind first. That basket, for the past several years, has reliably included crypto.
So when a reader sees 98.817 and thinks "dollar strong, risk assets pressured," they are approximately correct and structurally illiterate. The index did not cause anything. The index is one visible output of a funding system whose internal pressures are mostly invisible, and whose invisibility is not an accident. The eurodollar market is deliberately under-measured. It was built to be under-measured. Its defining innovation was the separation of a dollar claim from the jurisdiction that issues dollars, and every subsequent layer — swaps, forwards, the basis — has extended that separation deeper.
Crypto, from the beginning, has been a resident of this architecture. Not a challenger to it. A resident. The stablecoin gave the offshore dollar a retail face and a public ledger, and the public ledger is what makes the residency visible, which is why it took a decade for anyone to notice.
IV. The Stablecoin Is a Eurodollar With a Block Explorer
Strip the branding. A large dollar-denominated stablecoin is a claim on a company, redeemable at par, backed by a portfolio of short-duration instruments and secured lending, issued outside the banking perimeter, and used primarily for settlement between parties who do not wish to touch a bank. That is not an analogy for a eurodollar. It is a eurodollar, with better reporting latency.
The distinction that matters is not legal form but balance sheet substance. The largest issuers now hold portfolios of Treasury bills measured in the tens of billions of dollars. At points in the past several years, the Treasury holdings of a single stablecoin issuer have been large enough to place it among the more significant holders of short-dated US government paper in the world — comparable to a mid-sized sovereign, larger than most money market funds. This is a remarkable structural fact, and it is under-discussed because it is unglamorous. A private company, incorporated in a jurisdiction chosen for its regulatory hospitality, has become a marginal buyer of the instrument that funds the United States government, and it acquired that role by issuing a token that a retail trader in Manila or Lagos or Buenos Aires holds as a savings vehicle.
The consequences ripple in two directions, and both of them touch the index.
In the first direction, the stablecoin's reserve portfolio is a mechanical short-duration Treasury position, which means its yield is a function of the front end of the US curve, which means stablecoin issuers are, functionally, a floating-rate carry trade with a token wrapper. When the front end moves, the economics of issuing stablecoins move. When the economics move, the incentive to mint or redeem moves. When minting moves, dollar demand from the crypto complex moves. And dollar demand from the crypto complex is a component — small, but not zero, and growing — of the marginal bid for dollars in the offshore market.
In the second direction, the stablecoin is a transmission channel in the opposite polarity. A redemption is a sale of T-bills and a destruction of a dollar claim. A wave of redemptions is a wave of bill sales into a market that may or may not want them. This is the scenario that keeps central bank researchers awake, and it is the reason the regulation of stablecoins has moved from an afterthought to a front-burner item in every major financial center in the space of about three years.
I worked on the margins of one of those efforts. In Hong Kong, in 2024, I contributed to parts of the territory's digital currency work, and the thing that struck me most was not the technology. It was the aesthetics. A central bank digital currency, at least as it was being prototyped, is designed to be rigid. The interfaces are calm. The rules are legible. The color palette is institutional. Everything about it communicates that the issuer would like the system to be boring, and would like very much for you to think of it as plumbing rather than as money.
Compare that to the way a stablecoin presents itself: loud, procedurally opaque, marketed with the visual grammar of a consumer app, governed by a foundation whose decision-making is somewhere between corporate and tribal. One is a Swiss watch. The other is a reef. And the reef, by every measure that matters — volume, users, geographic reach, hours of operation — has been winning.
V. The Rate That Is Not a Price
There is a place where these two aesthetics collide, and it is the lending market.
I have written before about the interest rate models used by the large decentralized lending protocols, and I will be brief here, because the point is structural rather than polemical. In a traditional money market, the rate is discovered. Dealers quote, borrowers hit, lenders lift, and a level emerges from the interaction of order flow with a central bank's policy band. The overnight rate in a developed market is one of the most heavily arbitraged numbers on earth, because an enormous quantity of leverage is priced off it and a basis point of error is worth a great deal of money.
In a decentralized lending pool, the rate is computed. There is a formula. It has a base component, a shallow slope that applies below some target utilization, and a steep slope that applies above it. The target — the kink — is a parameter. The slopes are parameters. The base is a parameter. All of them were chosen, at some point, by a small number of people, discussed in a forum, and ratified by token holders who mostly did not read the proposal. The result is a rate curve that rises smoothly and then violently, and that has no more relationship to the marginal cost of dollar funding in the offshore market than a thermostat has to the weather.
This is not a criticism of anyone's competence. It is an observation about what the thing is. A decentralized lending rate is an administered rate wearing the costume of a market rate. It moves when governance moves it. It does not move when the cross-currency basis widens, because the pool cannot see the cross-currency basis. It does not move when SOFR ticks, because the pool has no position in the repo market. It moves when utilization changes, and utilization changes when depositors withdraw and borrowers repay, and both of those are lagged responses to price moves that happened somewhere else.
Which produces an interesting empirical artifact. In moments of genuine dollar funding stress — and I have watched a few — the DeFi rate curve reacts late, reacts partially, and frequently reacts in the wrong direction first. Deposits flee toward the perceived safety of the largest stablecoin; utilization spikes; the rate rockets to the steep slope; borrowers who cannot refinance elsewhere pay it; and then, a day or two later, the whole thing settles back because the underlying funding pressure was transient and the pool never had a real transmission channel to it in the first place.
I once spent the better part of a week auditing a stablecoin invariant — this was in the DeFi summer of 2020, when the curve designs were new and the elegance of them was genuinely seductive. What I found was not a bug. It was a property. In equilibrium, the invariant is beautiful; it prices near-parity swaps with almost no slippage, which is a small miracle of mathematics. In stress, the same invariant becomes a mechanism for absorbing the asset that is failing. When one leg of a stable pair drifts, the pool's formula does not defend the peg. It accepts the depegging asset, in increasing quantity, at a rate that gets better for the arbitrageur the worse things get. The design is not broken in a crash. The design is working exactly as constructed, and the construction is a machine for being the last holder.
That, I think, is the correct posture toward a great deal of DeFi infrastructure in a bull market. Everything is elegant until it is load-bearing. The euphoria of a rising tape is very good at hiding the difference, because in a rising tape nothing is load-bearing. Every curve is beautiful when no one is standing on it.
So: two systems, one called money and one called crypto, both claiming to price the same thing — the time value of a dollar — and neither of them actually pricing it. One administers its rate through a committee and a policy band. The other administers its rate through a forum post and a slope parameter. In between sits the actual market, in the tens of trillions per day, invisible to both, and moving on its own logic. Echoes of early hype in the quiet of current data.
VI. The Sequencer, Which Is One Machine
Now descend a level, from the price of money to the ordering of transactions, because the same pattern repeats and it repeats with a fidelity that should embarrass anyone still writing roadmaps.
An optimistic rollup, as deployed, has a component called a sequencer. Its job is to receive transactions, order them, execute them, and post the resulting state commitment to a parent chain. It is, in practice, a single server. It is operated by the foundation or the company that built the rollup. It has custody of the mempool. It decides what goes in what order. It is the block builder and the clock.
The consequences are not theoretical. A single sequencer is a single point of censorship — it can decline to include a transaction, and the user's only recourse is to force inclusion through the parent chain, which is expensive and slow and which almost no one has ever done. It is a single point of liveness failure — when it goes down, the rollup stops, and the definition of "down" includes a bad deploy, a cloud region outage, a spike in demand, or a misconfigured fee. And it is a single point of value extraction: the sequencer sees every transaction before it is ordered, which means it sees every arbitrage before it is ordered, which means the largest and most reliable revenue stream in the rollup economy flows to whoever holds the key.
Every serious rollup team has published a plan to decentralize this. Shared sequencer networks. Leader election among a permissioned set. A marketplace for ordering rights. Encrypted mempools. Threshold decryption. Based sequencing, in which the parent chain's proposer does the ordering. The names change every six months; the diagrams are consistently excellent.
I have been reading these documents for about two years now, and the thing I keep noticing is the gap between the diagram and the deployment. Not a skeptical gap. A measured one. Count the production systems running with a genuinely decentralized sequencer, at meaningful scale, with meaningful value at stake, and the number is very close to zero. Count the systems that have announced one, and the number is most of them. In between the announcement and the count sits a category of document that is not quite a lie and not quite a plan — it is a shape drawn on a slide, and it has been redrawn enough times that it has acquired the patina of a commitment.
Why does this matter for a macro essay about the dollar? Because it determines what crypto is. If the execution layer of the crypto economy is a handful of single servers operated by a handful of companies, then the crypto economy is not a decentralized alternative to the dollar system. It is a hierarchical system with its own chokepoints, its own rent extraction, and its own de facto central banks — and it is denominated in dollars, settles in stablecoins issued by companies, reads its prices from oracles operated by committees, and reaches its users through RPC providers that are, in the overwhelming majority of cases, three or four firms with AWS accounts.
Strip away the vocabulary and you find a structure that looks remarkably like the one it claims to replace: a small number of gatekeepers, a settlement asset issued by a private entity, and a retail population that holds claims it does not fully understand against institutions it cannot audit. That is not a criticism of the intent. It is a description of the shape. And shape is what survives a cycle.
VII. Two Cities, One Ledger
There is a second architecture worth describing, and I have a closer view of it than most, because I have spent the last two years inside it.
Hong Kong's digital currency program has moved in a cadence that is legible if you plot it against the calendar. The e-HKD pilot ran its first phase in 2023 with a set of selected firms, and its second phase in 2024 with a larger and more institutionally serious cohort. Project mBridge, the multi-central-bank cross-border bridge with the Hong Kong Monetary Authority, the Bank of Thailand, the People's Bank of China, and the Central Bank of the UAE, moved from prototype to a minimum viable product during the same window, and then the BIS Innovation Hub stepped back from it, which is a development most English-language coverage handled with a single sentence and which deserves considerably more. Project Ensemble, on tokenized deposits and settlement, was announced in 2024 with a list of participating banks that reads like the membership roll of a private club. Project Aurum, the retail CBDC prototype built with the BIS, had already laid the groundwork.
On the other side of the water, Singapore's Monetary Authority has run Project Ubin, Project Orchid, Project Guardian, a payment services licensing regime, and a stablecoin framework, each of them announced with the timing and polish of a product launch. Singapore is very good at this. It has been very good at this for a decade.
The interesting thing is not that both cities are working on the same problems. Everyone is working on the same problems. The interesting thing is the sequencing — which announcements follow which, and how closely. When one hub publishes a pilot cohort, the other publishes a framework. When one publishes a framework, the other publishes a cohort. The lag is measurable in months, not years, and it has been measurable in months for the entire duration of the current cycle.
Read that as competition and the story is simple and slightly cynical: two financial centers, each with a legitimate need to remain relevant as settlement migrates, each treating digital-asset regulation as a tool of jurisdiction-level positioning. Read it as coordination and the story is noble and slightly naive: two centers trying, in good faith, to build the standards that a fragmenting financial system needs. I have read enough of both sets of documents to say that the truth is duller and more interesting than either. The technical content of the two programs is substantially similar. The rhetoric is almost interchangeable. The differentiating variable is timing, and timing is a competitive instrument.
One thing the documents do not say, and that the dates imply: the licensing regimes are not primarily about welcoming innovation. They are about the moment when the alternative financial center to the north became unacceptable to a certain class of institution, and about which city gets to be the legal wrapper for the capital that consequently needs somewhere to sit. Regulation is a product. It has a launch window. The window opened, and both cities shipped.
For crypto, the practical consequence is a slow bifurcation. Capital that wants Asia exposure through a compliant wrapper will increasingly have two doors, and the doors will look nearly identical, and the terms behind them will be negotiated bilaterally and disclosed nowhere. That is not a prediction. That is already the state of the market, and the pleasant abstractions of "regulatory clarity" are the marketing copy for it.
VIII. The Contrarian Turn
Now the part where I disagree with nearly everyone.
The consensus position, stated in a thousand research notes and a hundred thousand posts, is that crypto has grown up. It is no longer an idiosyncratic asset with its own weather. It is a macro asset, trading on liquidity conditions, correlated to the dollar, legible to the same analysts who cover the ten-year yield. The 2022 correlation between Bitcoin and the dollar index — at points around negative eight tenths — is the standard exhibit.
I think the consensus has the direction of causation almost exactly backwards, and I think the error is a microstructure error that a macro lens cannot see.
Begin with the transmission channel. If crypto were a genuine macro asset, it would respond to macro information — employment prints, CPI, central bank guidance — through the same channels that equities do: discount rates, growth expectations, risk premia. It mostly does not. What it responds to, with a fidelity that is almost mechanical, is the availability and cost of collateral. And the availability and cost of collateral in the crypto system is determined by four things: the issuance and redemption behavior of the major stablecoin issuers, the funding rate on perpetual futures, the haircut applied to crypto collateral in prime brokerage arrangements, and the willingness of the largest market makers to hold inventory.
Each of those is a dollar-funded activity. Each of them is downstream of the offshore dollar market. And the dollar index is not the offshore dollar market. The dollar index is a fixed-weight average of six developed-market exchange rates, published continuously, and its correlation with crypto is a correlation between a visible proxy for dollar conditions and an asset whose actual dollar conditions are set in venues the index does not observe.
What does that produce? Two series that move together because they are both responding to the same underlying variable — the state of global collateral — with different lags and different noise profiles. That is not a causal relationship in the ordinary macro sense. It is a co-movement of proxies, and the reason it is so tight in stress is that in stress the underlying variable dominates everything else.
The practical consequences of getting this wrong are not trivial. If you believe crypto is a macro asset, you position it using macro inputs, and you expect the relationship to hold through cycles. If you believe crypto is a collateral asset with a macro-looking correlation, you position it using the state of the collateral market, and you expect the correlation to break precisely when the collateral market's structure changes — which is to say, when the stablecoin issuers, the prime brokers, and the market makers change their behavior, which is precisely what bull markets cause them to do.
Which brings me to the present. This is a bull market. The tape is up. The structures that were quietly load-bearing in 2022 have been quietly restructured in the interim, and the restructuring is mostly invisible because nothing has tested it. Newly funded protocols with nine-figure treasuries are announcing sequencer decentralization roadmaps whose timelines extend past the point at which their tokens unlock. Stablecoin issuers are holding portfolios whose composition is disclosed quarterly and audited by firms whose engagement terms are not. Lending pools are quoting rates derived from parameters ratified by governance votes in which participation was measured in single-digit percentages.
None of that is a prediction of catastrophe. It is an inventory. And the reason an inventory is worth taking in a bull market rather than a bear market is that in a bear market the inventory takes itself. Structures degrade silently, and the degradation is only visible in the aftermath, and the aftermath is a bad time to be learning what you own. The cracks were always in the street. The rain is what makes them visible, and it is not raining.
So here is the contrarian claim, stated as plainly as I can manage. Crypto's correlation to the dollar index is real, measurable, and structurally uninformative. It is a shadow cast by a funding market that most participants have never seen and that no index measures. Every model that treats the dollar index as an input to a crypto price is a model with a missing variable, and the missing variable is the one that actually moves. Echoes of early hype in the quiet of current data — the hype was that crypto had become macro. The quiet is that it became collateral.
IX. What To Watch When Nothing Is Happening
I want to end where I started, with the number, because the number is not going to help anyone and it is worth being honest about that.
A 0.03% move in a fixed-weight index of six currencies, closing at 98.817 in a year that the brief does not name, is not a reason to do anything. It is not a signal. It is not a warning. It is a reading taken at a moment when the meter happened to be pointing slightly up, published because the business of publishing is continuous and the appetite for readings is infinite.
What it is good for is orientation. The 98 handle is not a level that means anything on its own. It is a place the index has visited during different regimes, and the same number means different things depending on which regime you are standing in — which is the real lesson of a datapoint that arrives without a year attached. A level without a context is an object without a shadow. It has no depth. You cannot tell how far away it is.
So if there is a watchlist, let it be a watchlist of context rather than levels. Watch the cross-currency basis, because it is the price of the dollar that crypto actually pays. Watch the stablecoin issuers' reserve disclosures, because they are the closest thing the sector has to a central bank balance sheet and they arrive quarterly and nobody reads them. Watch the sequencer deployments, or rather the absence of them, because the distance between a rollup's decentralization diagram and its production architecture is the single clearest measure of how much of this industry is still slideware. Watch the licensing calendars in Hong Kong and Singapore, because the months between their announcements tell you what the announcements are for. And watch the lending rate curves, not for the level, but for the moment they fail to move when the world around them does — because that failure is the fingerprint of an administered rate, and administered rates are the oldest illusion in finance.
Ninety-eight point eight one seven. Tomorrow it will be something else, by an amount that is also approximately zero, and it will be reported with the same calm finality, and a certain number of people will read it and feel informed.
The question worth carrying forward is not whether the dollar is strong. It is whether we have ever actually been watching the dollar, or whether we have been watching a shadow on a wall and calling it the light source — and what becomes legible on the wall when the tape finally stops moving and the room goes still.