Creations, redemptions, and reading the data

Every evening, a handful of numbers tells the market what institutions did with Bitcoin that day, and in June 2026 those numbers set a record no one wanted. Here is the machinery behind ETF inflows and outflows: who actually buys the coins, what cash versus in-kind settlement changes, and how to read flow data without being fooled by headlines.

In June 2026, more than $4 billion left the United States spot Bitcoin ETFs in a single month, the worst stretch since the products launched, and the number led every market story for weeks. Most of the people trading on that headline could not describe what an outflow physically is: who sold what, to whom, whether any actual Bitcoin hit the market, or why the answer changed in 2025.

That gap matters because ETF flow data has become the single most-watched institutional signal in crypto. It updates every trading evening, it is free, and it describes the behavior of the largest regulated pool of Bitcoin demand ever assembled, funds that collectively hold well over a million coins. Reading it correctly is a genuine edge. Reading it the way headlines present it is a reliable way to buy tops and sell bottoms.

This guide covers the full machine: how ETF shares come into and out of existence, the role of authorized participants, what changed when the SEC allowed in-kind settlement, how flow numbers are actually produced and reported, and a practical framework for separating signal from noise, tested against the most brutal flow month on record.

What a spot Bitcoin ETF actually is

A spot Bitcoin ETF is a fund that holds real Bitcoin with a custodian and issues shares that trade on a stock exchange. Each share is a claim on a sliver of the fund’s coins, so the share price tracks Bitcoin’s price, and investors get exposure through an ordinary brokerage account with no wallets, keys, or crypto exchanges involved.

The critical design detail is that the number of shares is not fixed. When demand for shares outruns supply, new shares are created and new Bitcoin enters the fund. When supply outruns demand, shares are destroyed and Bitcoin leaves. That elastic supply is what keeps the share price glued to the value of the underlying coins, and the expansion and contraction of the share base is precisely what flow data measures. An inflow means the fund grew and acquired coins; an outflow means it shrank and shed them.

What flow data does not measure is trading. Millions of ETF shares change hands daily between investors on the exchange without the fund growing or shrinking at all, the same way a stock trades without the company issuing or buying back shares. Volume is churn; flows are net expansion or contraction. Confusing the two is the first beginner mistake, because a heavy-volume day with zero net flow means positioning battles, not institutional entry or exit.

Authorized participants: the machine’s operators

Ordinary investors cannot create or redeem ETF shares. That privilege belongs to authorized participants, or APs: large trading firms and banks, names like Jane Street, Virtu, and JPMorgan Securities, that contract with each fund to keep its share price aligned with its net asset value.

The mechanism is arbitrage. If buying pressure pushes an ETF’s share price above the value of the Bitcoin behind it, an AP steps in: it delivers Bitcoin or its cash equivalent to the fund, receives newly created shares in large blocks called creation baskets, and sells those shares into the demand, pocketing the premium. That creation is an inflow. If selling pressure pushes the share price below the value of the coins, the AP runs the machine backward: it buys discounted shares on the exchange, returns them to the fund, receives Bitcoin or cash worth slightly more than it paid, and profits from the gap. That redemption is an outflow.

Two properties of this design are worth internalizing. First, flows are demand-driven but AP-executed: the headline number reflects investor behavior, but the actual Bitcoin transactions are performed by a handful of professional firms optimizing execution across venues and time. Second, the arbitrage is what makes the ETF trustworthy: the reason the share price cannot drift far from the coin price is that a room full of well-capitalized firms is paid to punish any drift. When you read that a fund took in $500 million, the precise meaning is that APs created $500 million of new shares because investor demand made it profitable to do so, and the fund’s custodian now holds correspondingly more Bitcoin.

The concentration of the AP layer is itself a quiet risk factor that flow readers should keep in mind. Each fund contracts with a short list of firms, issuers maintain backups so a single AP’s outage does not strand the arbitrage, and custodial redundancy has grown for the same reason. The system has held through every stress so far, including the record 2026 redemption waves, but the mechanism that keeps ETF prices honest runs through fewer counterparties than most investors assume, and a disruption there would show up as premiums and discounts no flow table would explain.

Cash versus in-kind: the 2025 change that rewired the plumbing

When the SEC approved the spot ETFs in January 2024, it attached an unusual restriction: all creations and redemptions had to settle in cash. APs could not touch Bitcoin directly; they delivered dollars, and the fund itself bought the coins, or received shares back and sold coins to raise the payout. The stated concerns involved anti-money-laundering controls around APs handling crypto.

Cash settlement made every flow a forced market transaction. An inflow meant the fund had to go buy Bitcoin, immediately, at market; an outflow meant it had to sell the same way. That gave early flow data an unusually direct link to spot price pressure, and it imposed real costs: every buy-sell leg carried slippage and spread, and forced sales inside the fund generated taxable events that could hit shareholders who never sold a share.

In July 2025 the SEC reversed course and approved in-kind creations and redemptions, aligning crypto ETFs with how commodity funds have always worked. APs can now deliver actual Bitcoin to create shares and receive actual Bitcoin when redeeming. The change cut costs, tightened tracking, and improved tax efficiency, since an in-kind transfer of appreciated coins is not a taxable event for the fund.

It also subtly changed what flow numbers mean. Under cash settlement, an outflow implied near-immediate spot selling by the fund. Under in-kind settlement, an outflow can mean an AP received coins and warehoused, hedged, or sold them on its own schedule. The headline number still measures fund expansion and contraction accurately, but the mapping from flows to same-day spot pressure loosened, one of several reasons sophisticated desks treat flow data as positioning information rather than a mechanical price input.

The life cycle of a single inflow

The machinery is easiest to hold onto as a story, so follow one hypothetical $100 million inflow from decision to data point.

A pension consultant approves a 1% Bitcoin allocation, and on Tuesday morning the fund’s advisor buys $100 million of ETF shares through an ordinary equity order. That buying pressure nudges the share price a few basis points above the value of the Bitcoin behind each share. An authorized participant’s monitoring systems flag the premium within seconds, and the firm begins selling shares to the advisor’s order flow while simultaneously buying Bitcoin, on spot venues, over the counter, or from its own inventory, to stay hedged.

By the afternoon the AP has accumulated a full creation basket’s worth of exposure. It delivers the Bitcoin to the fund under the in-kind process, receives a block of newly created shares at net asset value, and uses them to flatten the short position it built selling to the advisor. The fund’s custodian logs the new coins; the AP books a few basis points of arbitrage profit; the advisor’s client owns its allocation at a fair price.

That evening the issuer reports the day’s net creation, the aggregators compile it, and a $100 million inflow prints on every dashboard. On-chain watchers see the custodial wallets grow by the corresponding coins, typically the next business day as settlement completes. Note what the headline number compresses: the buyer’s identity, the AP’s execution path, the split between spot and OTC sourcing, and the timing of every leg. The flow figure is true, and it is the last line of a longer story, which is exactly the right level of trust to place in it.

Run the film backward and you have a redemption: advisor sells, share price dips below asset value, AP buys the discount, returns shares, receives coins, and disposes of them on its own schedule. June 2026 was that film playing daily for a month.

A short history of the flow era

Flow data has only existed as a market force since January 2024, and its brief history splits into three chapters that explain how the market learned to read it.

The first chapter was the launch year, when inflows compounded so relentlessly that the funds swallowed multiples of newly mined supply and flow headlines became the daily proof of the institutional adoption thesis. The market’s first lesson was that flows could drive price: creation-driven buying was mechanical, visible, and large, and front-running the evening print became a trade of its own.

The second chapter taught the harder lesson that flows follow macro. The February 2025 outflow record of $3.56 billion arrived on rate fears, not on any Bitcoin-specific news, and recoveries tracked the Federal Reserve’s tone more than crypto’s. Sophisticated readers stopped treating flows as a Bitcoin verdict and started treating them as a risk-appetite gauge wearing a Bitcoin costume.

The third chapter is 2026, the stress test: a 13-day, $4.37 billion outflow streak in May, the $4 billion June, cumulative yearly flows turning negative, and then the discovery that none of it marked the death of demand, only its migration to channels the dataset cannot see. The chapter’s lesson is the one this guide is built around: flow data describes one pipe in a building full of plumbing, indispensable for that pipe and silent about the rest.

How the numbers are produced and where to find them

Each issuer reports its fund’s daily net flow after the United States close, and aggregators compile the fund-by-fund figures into the totals that make headlines. Farside Investors publishes the raw daily table going back to launch; SoSoValue and similar dashboards visualize the same data; Bloomberg terminals carry basket-level detail. On-chain analysts add a verification layer, since the funds hold real coins in identifiable custodial wallets whose balances can be watched directly, though the on-chain settlement typically lags the reported flow by a business day.

Three reporting conventions trip up new readers. Flows are reported in dollars, so the same dollar figure represents different amounts of Bitcoin at different prices. Net flow is the published number, meaning a day with $400 million in and $100 million out prints as $300 million, obscuring the gross activity. And the category total can hide sharp divergence between funds, which turns out to be one of the most informative details available.

Reading flows like a professional

The June 2026 record month is the best masterclass the dataset has produced, and the professional framework comes down to five rules it validated.

Rule one: streaks beat prints. A single day’s flow is noise; the market’s memorable moves came from runs, the 13-day, $4.37 billion outflow streak in May, then June’s follow-through to a $4.06 billion month that turned 2026 cumulative flows negative for the first time, context that framed Bitcoin’s slide below $59,000. Systematic desks generally require three to five consecutive sessions of confirming flows before treating a shift as regime change.

Rule two: breadth beats totals. When the streak finally broke on July 2 with a $221 million net inflow, the fund-level detail undercut the headline: one large fund absorbed $166 million while the biggest fund still bled $40 million. A genuine turn shows several funds green at once, led by the largest; a single fund catching the total is a bounce, not a regime.

Rule three: divergence from price is the highest-value signal. Flows agreeing with price mostly confirms consensus. The June setup, record outflows while price found a floor near $58,000, was informative precisely because someone was absorbing everything the ETFs shed, and the on-chain data identified whales accumulating more than 270,000 BTC over the same stretch. Flow data tells you what one cohort did; its disagreement with price tells you another cohort exists.

Rule four: cross-check the demand gauges. ETF flows describe the wrapped, regulated channel only. Pairing them with the Coinbase Premium Index, which tracks whether United States buyers are out-bidding the global market, catches cases where the channels disagree, and pairing both with on-chain cohort data completes the picture. June’s negative flows plus negative premium plus heavy whale accumulation was a three-gauge story no single dataset could tell.

Rule five: respect the macro context. Flow-driven selling in 2026 tracked rate expectations, a hot 4.2% inflation print, and even competition for risk capital from the SpaceX listing, whose index inclusion turned passive flows into their own crypto story. Mandated money de-risks mechanically when real rates rise; reading its flows as a pure Bitcoin verdict misreads how the sellers themselves think.

What flows can and cannot tell you

Kept within its range, flow data answers one question with unmatched clarity: is the regulated American wrapper for Bitcoin expanding or contracting, and how fast. That is a real and valuable question, because the cohort using that wrapper is enormous, slow-moving, and influential.

The dataset cannot see over-the-counter accumulation, offshore demand, corporate treasuries, or long-term holders, the cohorts that, in June, took the other side of the trade at record scale. It reports yesterday’s positioning, not tomorrow’s, and by the time a flow number is a headline, the transactions behind it are settled. And it inherits every distortion of its own plumbing: settlement mechanics, AP inventory decisions, and reporting conventions all sit between the raw number and the truth.

One further habit rounds out a durable practice: keep a written baseline. Flow analysis rewards memory, and the difference between a scary number and a normal one is usually the recent average. A reader who knows that June’s outflows ran near $200 million per session recognizes instantly that a $50 million red day is quiet and a $600 million one is an event, while a reader without the baseline treats every print as breaking news. A simple running log of daily net flow, the streak count, and the leader among funds takes a minute to maintain and converts the evening headline from a mood into a measurement.

The practical posture is the one every durable market indicator earns: a permanent tile on the dashboard, weighted heavily during streaks, read fund by fund, and never traded alone. Retail sentiment gauges make a useful contrast: when searches for Bitcoin going to zero hit records, that told you how the crowd felt, while the flow table told you what the institutions actually did, and the difference between feeling and doing is where most of the information lives. The investors who handled June best were not the ones who ignored the record outflows or the ones who panicked on them. They were the ones who knew exactly what the number measured, noticed what it could not measure, and went looking for the rest of the picture.

Five mistakes beginners make with flow data

The dataset’s popularity guarantees a steady supply of misreadings, and the common ones cluster into five patterns worth naming.

Trading the headline. By the time a flow figure is news, the transactions behind it settled hours earlier and the APs are already hedged. The evening print is a positioning record, not an actionable event, and strategies built on reacting to it are structurally late by a full session.

Confusing volume with flows. A billion dollars of ETF shares trading hands is churn between investors; the fund neither grew nor shrank. Heavy-volume, zero-flow days mean positioning battles, and reading them as institutional entry or exit misreads the mechanics entirely.

Reading the net and missing the gross. A quiet net figure can hide enormous two-way traffic, with one allocator entering as another exits. The published number nets it all; when basket-level or fund-level detail is available, disagreement inside the total is often the day’s real story.

Treating all funds as one investor. The complex spans low-cost vehicles favored by advisors, trading-oriented funds favored by hedgers, and converted legacy products with their own exit dynamics. IBIT bleeding while FBTC absorbs, the exact shape of the July 2 turn, describes rotation between investor types, not a single mind changing. Fund-level reading is the difference between narrative and analysis.

Extrapolating flows to the whole market. The wrapper holds a large minority of institutional exposure, not the whole of it, and June proved the point at record scale: the dataset showed a historic exit while the asset’s largest holders were absorbing supply through channels no flow table reports. The flow data was accurate, complete, and insufficient, which is the permanent condition of every indicator and the reason dashboards have more than one tile.

Internalize the five and the dataset becomes what it should be: the clearest single window into one specific cohort, consulted daily and trusted exactly as far as its plumbing extends.

Frequently asked questions

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Digital asset markets are volatile and you can lose your entire investment. Always do your own research. Information current as of July 4, 2026.