Author: Darko, IOSG

ETF inflows are often seen as a "thermometer" of institutional big money's confidence in Bitcoin. But week after week, it measures something else more: a hidden interest rate trade that is repeatedly turned on and off. This article explains how to distinguish it, how large this trade really is, and why it is quietly exiting.

TL;DR

1. Week by week, ETF flows are primarily driven by a hidden arbitrage trade, not conviction. Cash-and-carry arbitrage traders buy ETFs while shorting futures on the CME, hedging out price risk, but are indistinguishable from true bulls in the data. About half of the weekly flow volatility can be explained by new futures shorts added by hedge funds alone, with a correlation as high as 0.70.

2. Bitcoin's weekly price changes explain almost none of the flows. Using price returns to predict ETF flows yields results statistically indistinguishable from zero. Weekly funds are not chasing price performance but moving in lockstep with a hedged interest rate trade.

3. Arbitrage dominates the weekly "volatility," but has never been the main body of the "stock." Of the roughly $55 billion in cumulative ETF inflows, the current net arbitrage trade accounts for only about $1 billion; the rest is stable, directional buying, roughly $400 million per week, which, compounded over two years, constitutes almost the entire "mountain."

4. The correct statement is: ETF flows overestimate the "volatility" of conviction, not its "level." The weekly ups and downs are mostly "rented"—arbitrage capital comes and goes; while the assets that truly settle are mostly "owned."

5. This trade is exiting, and has been for two years. Leveraged fund short positions, once piled from about $3 billion at launch to roughly $14 billion by the end of 2024, have steadily fallen back to about $4.5 billion. Once the basis compresses to unprofitability, inflows and shorts will fade in tandem—do not misread the resulting outflows as the market passing judgment on Bitcoin.

I. The Number Everyone Watches

Every week, Bitcoin ETFs report how much money flowed in or out, and this number is often taken as a verdict. Large inflows mean institutions are piling in; outflows mean confidence is shaken. Flow data has quietly become the market's headline indicator for measuring conviction.

The problem is, not everyone buying the ETF is betting on Bitcoin. Some of the biggest buyers don't care which way the price goes—and once you account for them, the weekly flow numbers measure their activity more than anyone's belief. To understand why, you must first recognize a very different type of buyer.

A Buyer Indifferent to Price

There is a classic, boring trade called cash-and-carry arbitrage. A Bitcoin "future" is simply a contract to buy or sell Bitcoin at an agreed price on a future date, and most of the time, the futures price is slightly higher than the current spot price—for example, say Bitcoin is $100 now, but the contract expiring in three months sells for $103.

A trader can pocket this $3 spread without holding any view on price:

• Buy 1 Bitcoin today for $100 (often achieved by buying the ETF).

• Sell a futures contract at $103, promising delivery in three months.

See what happens at expiry. If Bitcoin surges to $120, the trader makes $20 on the coin but loses $17 on the contract—netting $3. If it crashes to $80, they lose $20 on the coin but make $23 on the contract—still netting $3. If it stays flat, it's still $3. In every scenario, the profit is the same. The directional risk is hedged out; traders call this "delta neutral." This $3 spread, expressed on an annualized basis, is the basis—essentially an interest rate the trader earns by parking capital in this trade; as long as it's higher than the risk-free return from putting money in U.S. Treasury bills (T-bills), the trade is worth doing.

Why This Pollutes the Headline Numbers

Here is the key. The first leg—buying 1 Bitcoin—is very commonly done by buying the ETF. Thus, a delta-neutral trader with no view on Bitcoin shows up in the data as an ETF inflow, superficially indistinguishable from a true believer.

When large amounts of cash-and-carry trades are put on, inflows look strong, and the "institutions are loading up" narrative naturally takes hold—even though these funds are hedged and will reverse the moment the trade is no longer profitable. In other words, the flow numbers don't just measure conviction; they measure the activity level of arbitrage desks. The question is how to separate the two—and how large each truly is.

How to Tell Them Apart

Cash-and-carry arbitrage traders leave a second footprint. For every $1 worth of Bitcoin they buy, they short $1 worth of futures on the CME (the regulated U.S. exchange where institutions trade Bitcoin futures). True believers leave only the first footprint; arbitrageurs leave both.

And the second footprint is public. U.S. derivatives regulators publish weekly reports disclosing the long and short positions of various trader categories on the CME. One category—leveraged funds, essentially hedge funds—is precisely where the cash-and-carry crowd gathers. So you can, week by week, line up the money flowing into ETFs against the new short positions established by these funds. If "demand" were truly conviction, the two shouldn't be highly correlated; if a large part of it is that hidden trade, they should move together.

II. What the Data Says: Week by Week, Flows Follow Futures, Not Price

The two move tightly together. In every week since the ETFs launched, weeks with more new futures shorts also saw more ETF inflows—almost one-for-one. About half of the entire week-to-week flow volatility can be explained by this one thing alone: how many new shorts the funds put on. The correlation is 0.70, the kind of strength you see between two things that are clearly related, not coincidental.

The point that should most sober up the believers: price itself explains almost nothing. Testing whether the week's Bitcoin return can predict ETF flows yields an answer statistically indistinguishable from zero. Weekly funds are not chasing performance; they move in lockstep with a hedged interest rate trade.

So, as a week-to-week signal, ETF "demand" is mainly arbitrage. The flow numbers are a poor thermometer for conviction because their ups and downs are the result of the basis trade being repeatedly turned on and off, not anyone changing their mind about Bitcoin.

But How Much of the Money is Actually This Trade?

It is precisely here that the simplistic narrative—"it's all fake"—falls apart, and the real story is more interesting. The basis trade dominates the weekly volatility, but has never been the main body of the capital.

Break down the weekly inflows into the part explainable by futures shorts (hedged) and the rest (directional), then cumulate them since launch. Of the roughly $55 billion in cumulative ETF inflows, the basis trade currently accounts for a net amount of only about $1 billion—the rest is steady, directional buying. This buying is roughly $400 million per week, week after week, regardless of the basis or price, and compounded over two years, it constitutes almost the entire mountain.

Looking at it in terms of asset share rather than flows paints the same picture: the hedged portion once approached 14% of ETF assets in 2024, and is now around 4%–5%. At its peak, it was a non-trivial minority; now it's just a sliver.

So, the more precise statement is: ETF flows overestimate the volatility of conviction, not its level. The weekly ups and downs are mostly "rented"—arbitrage capital comes and goes; but the assets that truly settle in are mostly "owned." This trade churns through the flow-of-funds data but has never been the bulk of the balance.

And this trade is now exiting

The hedge leg hasn't just been small all along—it has been shrinking for two years. Leveraged funds' short positions piled up from roughly $3 billion at launch to about $14 billion by the end of 2024, then steadily retreated to around $4.5 billion. This arbitrage trade has been unwinding across the entire period, not just recently.

This matters for interpreting the present. Heading into June, hedge positions roughly halved again—fund shorts shrank from about $6.4 billion to $4.3 billion—while ETFs saw daily outflows of $300 million to $500 million. Taken at face value, the numbers look like panic capitulation. But paired with futures data, it's simply the routine cleanup of a rate trade that is no longer profitable. Same outflow numbers, two entirely different stories.

When the basis compresses, demand fades

The cleanest corroboration is what happens when the trade stops being profitable. When that $3 spread narrows to near the level traders can earn risk-free, the trade is no longer worth doing. If a big chunk of weekly demand really is this trade, then weekly demand should weaken exactly when the basis compresses—and that is precisely what happens. Strip the trend from each series and observe the moments around a compression: ETF inflows fall below their normal rhythm, and funds simultaneously cover shorts, both moving in sync. Demand breathes with the trade.

True believers don't care about the futures basis. And this weekly "demand" clearly does.

Three: Who leads whom, and who is really operating

First, this correlation is contemporaneous—tightest within the same week, with no clear lead or lag; and what little directional evidence exists actually points in the opposite direction: ETF flows are driving the shorts, not the other way around. This fits the logic of a paired trade: buy the ETF first, with the futures hedge following, rather than shorting somehow "conjuring" inflows out of thin air. Second, the arbitrage crowd is not the only driver. Flows track leveraged fund shorts most closely, but they also resonate with directional institutional positioning—both types of buyers are active. The argument here is not that every single inflow is a hedge; rather, the hedge trade is the tightest, most reliable driver of week-to-week fluctuations.

Ethereum: The same trade, but the math barely works

Apply the same test to Ethereum ETFs, and the signature is still there, but weaker—the correlation with futures shorts is looser, and the steady directional bid underneath is nearly absent. The reason is clear. Holding spot Ether instead of futures means forgoing the staking yield Ether provides, roughly 3%–4% annually. After deducting that, the Ether basis is often negative—the arbitrage trade often can't even clear its hurdle rate. So Ethereum ETFs have neither strong conviction buying nor robust arbitrage positions to support them; they are simply smaller and noisier than their Bitcoin counterparts.

Four: How to read ETF flows from now on

The point is not a price call, but a method for interpreting flows. When the basis is rich, expect "institutional demand" to appear strong and to be largely hedged—don't mistake that strength for conviction. When the basis compresses, expect inflows and shorts to fade together—don't misread the resulting outflows as the market passing judgment on Bitcoin. Two numbers worth tracking: the annualized basis yield relative to the T-bill rate, and the weekly net short position of leveraged funds in the CME report. They will tell you how much of the next "demand" headline is real.

How we measured it

A few honest limitations. The basis is constructed using the front-month CME futures contract against spot, with the final days before each expiry removed (their extremely short time to expiry amplifies rounding errors into spurious spikes); a contract-by-contract series would sharpen the exact numbers but wouldn't change the conclusions. The relationship between flows and shorts is a strong co-movement, not proof that one causes the other—the key point is that they are two halves of the same trade. The futures short figure is an upper bound on the proportion of ETF buying that is hedged, because some shorts hedge coins held elsewhere.

None of this changes the core story. Week to week, Bitcoin ETF "demand" is primarily a hidden rate trade, not conviction—flows measure the intensity of arbitrage participation far more accurately than they measure belief. And that genuine buy-side demand is real, patient, and now constitutes the vast majority of what remains, because the "rented" portion has spent two years heading home.