Bitcoin ETF Authorized Participants: How Creation & Redemption Actually Works

May 25, 2026• 13 min

Who the authorized participants are, how big a creation basket is at each US spot Bitcoin ETF, what changed when the SEC allowed in-kind orders on 29 July 2025, and what happens when creations stop.

TL;DR. Authorized participants are the only firms that deal directly with a spot Bitcoin ETF, and they deal only in whole baskets. Those baskets run from 5,000 shares at ARKB, BRRR, BTCO and BTCW to 50,000 at Franklin's EZBC, which at 22 September 2026 closing prices is $143,000 to $2.49m per order. Issuers name their APs unevenly: BlackRock's Q3 2025 10-Q names Jane Street, Virtu Americas, JP Morgan Securities and Marex Capital Markets as the four able to settle IBIT in kind, while most filings just say "the Authorized Participants". The big change came on 29 July 2025, when the SEC approved in-kind orders in Release 34-103571; IBIT bought $9,362,376,537 of bitcoin in kind for shares issued in the first half of 2026 alone. It isn't universal, since Grayscale's GBTC was left out of that order and still runs on cash. The arbitrage works: IBIT closed 21 basis points below NAV on 22 September 2026, and through the whole first half of 2026 it never ended a day more than 1% away from it.

What an authorized participant actually does

An ETF has two markets stacked on each other, and most investors see only one. On the exchange, shares change hands between investors and the fund's bitcoin holdings don't move by a satoshi. Underneath sits the primary market, where a short list of broker-dealers hand the trust bitcoin or dollars and receive newly issued shares, or hand back shares and take bitcoin or dollars out. Those firms are the authorized participants.

That second market is the entire reason an ETF tracks anything. If the share price drifts above the value of the coins behind it, an AP mints more shares and sells them into the rally; if shares sag below the coins, an AP buys them cheaply and hands them back for coins worth more. Nobody does this out of duty. They do it because the gap is money, and what's left when they're done is a share price sitting on net asset value. The mechanics of that gap are in premium, discount and NAV.

Three consequences matter. Flows and volume are different quantities, because only AP activity changes what the trust holds. Flows arrive in lumps, because APs can't deal in odd lots. And the arrangement is contractual, not statutory: nothing obliges an AP to show up on a given morning. For the wider picture, start with what a Bitcoin ETF actually is.

Who the authorized participants actually are

This is harder to answer than it should be. Some issuers name the firms in a 10-K or prospectus; others write "the Authorized Participants" throughout and never print a name. Anyone offering a full roster for all twelve US spot Bitcoin funds is guessing. Here's what's on the record.

Fund	Firms named in public filings	Where it's disclosed
IBIT (iShares Bitcoin Trust ETF)	Jane Street Capital, Virtu Americas, JP Morgan Securities, Marex Capital Markets, named as the APs able to support in-kind activity	Form 10-Q, quarter ended 30 Sep 2025
ETHA (iShares Ethereum Trust ETF, same sponsor)	Ten APs, among them Goldman Sachs, JP Morgan, Jane Street, Virtu, Jefferies, Macquarie and UBS	Form 10-K for 2025
BTC (Grayscale Bitcoin Mini Trust ETF)	Jane Street and Virtu, whose agreements were amended for in-kind orders	Form 8-K filed 6 Feb 2026
GLNK (Grayscale Chainlink Trust ETF, same sponsor)	Jane Street, Virtu Americas, Macquarie Capital (USA), ABN AMRO Clearing USA	Prospectus
GBTC, FBTC, BITB, ARKB, HODL, BRRR, EZBC, BTCO, BTCW, MSBT	No firms named in the filings reviewed; the AP role is described generically	10-K / 10-Q / prospectus

Read that as a floor rather than a census, but the pattern is the point. The same two names, Jane Street and Virtu, appear in every list that exists, at two sponsors, across bitcoin, ether and chainlink funds. Concentration in ETF market-making isn't a crypto quirk, but it does mean the category's arbitrage rests on a handful of balance sheets. A dozen APs on paper also isn't a dozen firms competing on a given Tuesday: an agreement costs nothing to keep open, while assembling a basket needs a desk that can move several million dollars of bitcoin inside a session.

The basket is the unit of everything

A creation unit, or basket, is the minimum block an AP can deal in. It's fixed in the trust's agreement and doesn't change with the price, so its dollar size floats with bitcoin. Here are the twelve US spot Bitcoin funds, each basket priced at the 22 September 2026 close, set against how often that fund actually moved over the last 90 trading sessions in our own flow series.

Ticker	Basket (shares)	One basket at 22 Sep 2026	Sessions with a flow (of 90)	Median size of a moving day
EZBC	50,000	$2,491,000	13	$6.6m
IBIT	40,000	$1,953,200	89	$143.6m
FBTC	25,000	$1,874,500	81	$30.7m
HODL	25,000	$609,500	36	$5.3m
GBTC	10,000	$666,700	36	$43.9m
BITB	10,000	$469,500	46	$10.1m
BTC (mini)	10,000	$381,400	38	$13.0m
MSBT	10,000	$246,153	49	$6.9m
BTCW	5,000	$455,500	5	$4.0m
BTCO	5,000	$427,650	11	$6.4m
ARKB	5,000	$142,646	54	$20.2m
BRRR	5,000	$121,500	2	$2.8m

Basket sizes come from the funds' prospectuses and annual reports; prices and flow counts are our own data for the 90 sessions from 14 May to 22 September 2026. A small basket doesn't buy you an active fund: ARKB and BRRR both deal in 5,000 shares, and one printed a flow on 54 sessions while the other managed two. Fund size dominates. What the basket controls is granularity, which is why EZBC sits still for weeks and then prints one round number, while IBIT's flow looks smooth at about 74 baskets on a median day.

The second reading is that a zero says something about basket economics, not investor interest. BRRR's two moving days in 90 don't mean nobody owns it. They mean net demand never reached $122,000 on a day an AP felt like doing the work. That's the most misread thing in ETF flow data, and how to analyze ETF flows unpacks it.

In-kind versus cash, and what changed on 29 July 2025

For their first eighteen months, US spot Bitcoin ETFs settled every order in dollars. An AP wired cash, the trust bought coins through an agent and sent them to the custodian; redemption ran the film backwards.

It's often said the SEC banned in-kind in January 2024. Not quite. The approval order of 10 January 2024, Release 34-99306, addresses in-kind exactly once, in a footnote, and only to say that the proposals before the Commission contemplated cash creation and redemption and that in-kind was therefore outside the order's scope. Cash-only was the shape of the filings, not an explicit prohibition, which is why a rule change could reverse it. On 29 July 2025 the SEC granted accelerated approval to nine filings from Nasdaq, Cboe BZX and NYSE Arca in Release 34-103571, covering sixteen bitcoin and ether funds. The proposals, in the order's words, "would allow the Trusts, in addition to their ability to transact with authorized participants in cash, to engage in creations or redemptions of their shares in an in-kind transaction of spot bitcoin or spot ether".

What in-kind actually changes

Nasdaq's filing argued it operationally: in-kind lets the AP, not the trust, decide how, where and whether to source or sell the bitcoin, which takes the fund out of the market on both sides of every order. BlackRock's risk factors make the point in reverse, warning that limited ability to do in-kind "could result in the exchange-traded product arbitrage mechanism failing to function as efficiently as it otherwise would", with wider bid/ask spreads and greater costs to investors. Adoption was staggered, gated by broker-dealer compliance rather than fund readiness: ARK 21Shares enabled it on 30 July 2025, BlackRock on 31 July, VanEck not until 20 November. Volumes aren't marginal: IBIT's 10-Q for the quarter ended 30 June 2026 records $9,362,376,537 of bitcoin purchased in kind for shares issued over six months, against $5,491,156,370 paid out in kind for redemptions.

The tax claim that doesn't survive the filings

Here's where the standard explanation goes wrong. In-kind redemption is genuinely tax-efficient for a conventional 1940 Act ETF, and the habit is to carry the argument over. These funds aren't 1940 Act funds. They're grantor trusts, and their financial statements show realized gains and losses booked on in-kind redemptions anyway: a realized loss of $771,263,116 at IBIT on $3.85bn of in-kind redemptions in the June 2026 quarter, and a net realized gain of $0.5bn at FBTC on bitcoin distributed for redemptions over the first half. The SEC's order says only that in-kind "may enhance tax efficiencies", hedged and unquantified. So the honest version is narrower: in-kind removes a trading step and a spread from the fund's side, which is a real saving, while what lands on your return runs through Bitcoin ETF tax in the USA.

The funds still running on cash

The July 2025 order didn't cover everything. Grayscale wasn't in it, and its prospectus of 12 September 2025 said the trust was "not at this time able to create and redeem shares via in-kind transactions". The Bitcoin Mini Trust got there in early 2026, disclosing in an 8-K filed 6 February that NYSE Arca had approval and that AP agreements with Jane Street and Virtu were amended accordingly. GBTC didn't follow: its 10-Q for the quarter ended 30 June 2026 still states that its AP agreements don't provide for in-kind, so every order runs as cash and the trust does the selling. Franklin's EZBC is a softer case, with the ability to transact in kind but cash orders in practice. Permission and practice are different columns, and the fund-by-fund picture is in the major Bitcoin ETF funds overview.

One creation, start to finish

Take an in-kind creation of ten IBIT baskets, 400,000 shares, worth about $19.5m at the 22 September 2026 close.

IBIT trades a few basis points above the bitcoin behind it, and an AP decides the gap covers its costs.
The AP submits a creation order before the fund's daily cut-off, which falls earlier than the 4pm close, and it prices at that day's benchmark rate rather than at whatever the shares last printed.
The AP sources the bitcoin itself and delivers coins to the trust's custody address. Before 31 July 2025 this step was a dollar wire and the trust did the buying.
The trust issues 400,000 new shares, and the AP sells them into the market over the following sessions, closing the gap.

Scale it up and you get the filings: IBIT created 2,523 baskets and redeemed 4,684 in the June 2026 quarter, 100.9m shares in and 187.4m out. One quirk trips up first-time readers of flow data. Most issuers book a creation into the next day's share count, so a Tuesday order usually appears in Wednesday's figures, as how to read Bitcoin ETF flows explains.

Does the arbitrage actually work? The numbers say yes

Readings below come from issuer pages on 22 September 2026, except BITB, whose page was dated 20 September.

Fund	NAV	Market price	Premium / discount	30-day median bid-ask spread
IBIT	$48.93	$48.83	−0.21%	0.02%
FBTC	$75.16	$75.03	−0.20%	0.04%
BITB	$46.86	$46.96	+0.21%	0.02%
ARKB	$28.65	$28.59	−0.21%	0.04%
GBTC	$66.63	$66.63	0.00%	0.02%
BTC (mini)	$38.11	$38.11	0.00%	0.03%

A single day proves little, so read the quarterly distributions. IBIT closed at a premium on 29 days and a discount on 32 in the first quarter of 2026, then at a premium on 43, at NAV on 3 and at a discount on 16 in the second. Every bucket beyond plus or minus 1% is empty in both, and the half's widest readings were +0.99% on 3 February and −0.71% on 17 June. That's 93.4% of days inside half a percent of NAV in the first quarter, 90.3% in the second, and 98.3% at ARKB in the third to 22 September. Which way the balance leans follows flows rather than any tracking defect: BITB spent 92 days at a discount and 87 at a premium across 2025, then flipped to 40 premium against 18 discount in the second quarter of 2026.

For contrast, here's the absence of this machinery. Before conversion GBTC was a closed-end trust with no redemption mechanism, so shares could be created but never handed back. Grayscale's 2023 10-K records a maximum discount of 49% between May 2015 and December 2023, an average discount of 25%, and shares quoted below NAV on 718 separate days; at the end of 2022 the discount stood at 45%. The fund converted on 11 January 2024, redemptions became possible, and by 19 February the same filing reports the shares trading 0.02% from NAV. Five weeks of working arbitrage closed a gap nine years of investor patience couldn't, and our GBTC analysis takes it apart.

What can break

Nobody bothers. An AP decides a dislocation is too small to justify assembling a basket, so thin funds print zeros for weeks and quote wider spreads than IBIT.
An AP steps away. Agreements can be terminated, and capital or compliance limits can make a desk drop a product. With a handful of firms doing most of the work, that's more concentrated than long AP lists imply.
Settlement and venue trouble. A cash order leaves the trust exposed to the bitcoin price between order and execution; in-kind moves that risk to the AP. Either way an outage at a major venue during the pricing window is a real hazard, which is part of why custody arrangements matter.
Creations stop altogether. The end state, and it happened in this market in 2026.

When creations stop: the BWOW case

Bitwise's Dogecoin ETF is the cleanest recent example of the mechanism being switched off on purpose. BWOW listed on NYSE Arca in November 2025 and never gathered assets: $1,152,815 of net assets at the end of 2025, $604,292 at 31 March 2026, $473,547 at 30 June. On 10 September 2026 Bitwise filed an 8-K notifying the exchange of its decision to close, de-list and liquidate. Creation of new shares ceases before the open on 15 October 2026, trading halts at the same moment so the last session is 14 October, and on 22 October remaining shareholders receive the net asset value of their shares as of 21 October.

Once creations stop, the share count is capped and the arbitrage works in one direction only. In a liquidation the gap stays small, because the terminal value is a known cash figure weeks out. In a vehicle that halts creations and keeps trading indefinitely, the price can wander a long way from the assets. That's the structural difference between an ETF and a closed-end trust, and it's what produced GBTC's 49% discount.

How to watch AP activity yourself

You don't need filings for the day-to-day. Every US spot Bitcoin ETF publishes shares outstanding and NAV daily, and the change in shares outstanding priced at NAV is that day's creation or redemption activity. That's the calculation behind our Bitcoin ETF flows page, broken out by issuer so you can see which funds an AP actually worked in, while the funds page carries fees, exchanges and listing dates.

Across the category's history our series counts $56.94bn of net creations over 676 sessions from 11 January 2024 to 22 September 2026, split 402 positive days to 274 negative. IBIT accounts for +$64.86bn of that and GBTC for −$27.84bn: one fund's APs building a position for two and a half years while another's unwound one. Read flows in baskets rather than dollars when comparing small funds, treat a zero as "no basket" rather than "no interest", and check the issuer's premium page before a large order.

FAQ

What is an authorized participant in a Bitcoin ETF?

A broker-dealer that has signed an agreement letting it create and redeem ETF shares directly with the trust, in fixed blocks called baskets. For US spot Bitcoin ETFs those run from 5,000 shares (ARKB, BRRR, BTCO, BTCW) to 50,000 (EZBC). APs are the only route by which a fund's bitcoin holdings change.

Who are the authorized participants for spot Bitcoin ETFs?

Disclosure is uneven, so the honest answer is partial. BlackRock's Form 10-Q for the quarter ended 30 September 2025 names Jane Street Capital, Virtu Americas, JP Morgan Securities and Marex Capital Markets as the IBIT APs able to support in-kind orders; Grayscale's 8-K of 6 February 2026 names Jane Street and Virtu for its Bitcoin Mini Trust. Most other funds describe the role without naming firms.

Do Bitcoin ETFs use in-kind or cash creation?

Both, depending on the fund. Until 29 July 2025 every US spot Bitcoin ETF settled in cash. On that date the SEC approved Release 34-103571, permitting in-kind orders for sixteen bitcoin and ether funds. In-kind is now the norm: IBIT bought $9,362,376,537 of bitcoin in kind for shares issued in the first half of 2026. GBTC is the exception, since its June 2026 10-Q says its AP agreements still don't provide for in-kind.

Did the SEC ban in-kind creation when it approved spot Bitcoin ETFs in 2024?

Not in those words. The approval order of 10 January 2024, Release 34-99306, mentions in-kind only in a footnote, stating that the proposals before the Commission contemplated cash creation and redemption and that in-kind was outside the order's scope. Cash-only was the shape of the filings rather than an explicit ban, which is why a rule change could reverse it in July 2025.

Can a retail investor create or redeem ETF shares?

No. Only authorized participants deal with the trust, and only in whole baskets worth hundreds of thousands to millions of dollars. Everyone else trades on the exchange through a broker, and a share bought from another investor changes nothing about how much bitcoin the fund holds. That is why volume and net flow are different numbers.

Why does a Bitcoin ETF show zero flow on a day when it clearly traded?

Because no basket was created or redeemed. Flow measures the change in shares outstanding priced at NAV, and shares move only in blocks. Over the 90 sessions to 22 September 2026, BRRR recorded flow on 2 days and BTCW on 5, against 89 for IBIT. Those funds had buyers and sellers every session; demand never reached one basket.

What happens if authorized participants stop creating shares?

The share count is capped and the price can drift from the value of the underlying bitcoin, because the arbitrage then works in one direction only. The clearest recent case is Bitwise's Dogecoin ETF BWOW: its 8-K of 10 September 2026 set creations to cease before the open on 15 October 2026, with a last trading day of 14 October.