US spot bitcoin ETFs pulled in $1.71 billion across September 21–22, 2026, per Farside Investors, with Sept. 21 alone at $999.0 million. On a flat $10,000 position, IBIT’s 0.25% fee costs $25 a year forever; buying spot on Kraken at the 0.80% taker tier costs $80 once. Self-custody breaks even in 3.2 years.
That is the whole bitcoin ETF vs self-custody question in one line of arithmetic. An ETF charges rent on your position every year. Holding coins yourself charges a toll once, at the door.
The flow rebound this week makes the comparison urgent, because a lot of money just chose the expensive side. Below is the math, the tickers, and the part of the cost that never shows up in an expense ratio.
What happened with bitcoin ETF flows on September 21–22, 2026?
Flows turned sharply positive. Farside Investors logged $999.0 million of net creations on Sept. 21 and $714.7 million on Sept. 22 — $1.71 billion in two sessions. Bloomberg reported on Sept. 23 that the complex had turned positive after a $4.6 billion rebound. Bitcoin traded at $84,366 on Sept. 23, per CoinGecko.
The Coin Republic called Sept. 21 the largest single inflow day since October 2025. Cumulatively, according to Farside, the US spot complex has taken in $56.94 billion since launch, against an average daily figure of just $84.2 million.
The concentration is the story. Across those two days, IBIT took $731.7 million and FBTC took $496.2 million — $1.23 billion, or 71.7% of the total. Bitwise’s BITB, which charges less than both, took $21.6 million: 1.3%.
Grayscale’s legacy GBTC sits at negative $27.84 billion cumulative, per the same Farside table. IBIT’s cumulative take is $64.86 billion.
Why the flow split matters to a cost comparison
It proves buyers are not optimizing for fees. If they were, the 0.15% and 0.20% funds would be absorbing the creations. Instead the two 0.25% funds took roughly seven dollars in every ten.
Investors are paying up for liquidity and brand. That is a defensible trade — but it is a trade, and it should be priced deliberately rather than by default.
Is a bitcoin ETF or self-custody cheaper to hold?
Self-custody is cheaper on a long holding period; the ETF is cheaper on a short one. The crossover sits between 1.6 and 5.3 years depending on which fund and which order type you use. The ETF’s cost recurs annually on market value. The self-custody cost is one-time and fixed in dollars at purchase.
The recurring cost: sponsor fees
The iShares product page lists IBIT’s sponsor fee at 0.25%, with net assets of $67.08 billion as of Sept. 23, 2026. Fidelity’s documentation puts FBTC at 25 basis points.
Bitwise discloses 0.20% for BITB, with $3.32 billion in assets as of Sept. 21, 2026. Grayscale’s Bitcoin Mini Trust page lists 0.15% and calls itself the lowest-cost bitcoin exchange-traded product in the world.
The one-time cost: exchange fees and spread
Kraken’s published fee schedule puts entry-tier spot trading at 0.40% maker and 0.80% taker for accounts with $0-plus in 30-day volume. A market order pays the taker rate. A resting limit order pays the maker rate.
Add a bitcoin network withdrawal fee to move coins off the exchange. Kraken states these vary by network conditions rather than publishing a fixed figure, so treat it as a small variable line item, not zero.
What does a $10,000 bitcoin position actually cost each way?
On a flat $10,000 held five years: Grayscale’s BTC costs $75, BITB costs $100, IBIT and FBTC cost $125 each. A market-order self-custody purchase costs $80 once, then nothing. A limit-order purchase costs $40. Here is the arithmetic, using the fees above.
| Route | Rate | Year 1 | 5 years | 10 years |
|---|---|---|---|---|
| Grayscale BTC (Mini) | 0.15%/yr | $15 | $75 | $150 |
| Bitwise BITB | 0.20%/yr | $20 | $100 | $200 |
| iShares IBIT | 0.25%/yr | $25 | $125 | $250 |
| Fidelity FBTC | 0.25%/yr | $25 | $125 | $250 |
| Self-custody (taker) | 0.80% once | $80 | $80 | $80 |
| Self-custody (maker) | 0.40% once | $40 | $40 | $40 |
The breakevens fall out directly:
- Market order vs IBIT: $80 ÷ $25 = 3.2 years
- Market order vs Grayscale BTC: $80 ÷ $15 = 5.3 years
- Limit order vs IBIT: $40 ÷ $25 = 1.6 years
- Limit order vs Grayscale BTC: $40 ÷ $15 = 2.7 years
- Limit order vs BITB: $40 ÷ $20 = 2.0 years
One adjustment matters more than any of this. The sponsor fee is charged on market value, not on your cost basis. If bitcoin triples, the ETF’s annual dollar cost triples with it while the self-custody toll stays fixed at $80.
That asymmetry cuts both ways. In a drawdown the ETF’s dollar fee falls, and the self-custody buyer is stuck having paid 0.80% on a higher entry. The point is that the crossover is a moving target, not a fixed 3.2 years.
Which bitcoin ETF has the lowest fee?
Grayscale’s Bitcoin Mini Trust (BTC) at 0.15% is the cheapest sponsor fee among the funds we verified from issuer pages, ahead of Bitwise’s BITB at 0.20% and the 0.25% charged by both IBIT and FBTC. But the headline fee is not the whole holding cost.
| Fund | Sponsor fee | Assets | Sept 21–22 flows |
|---|---|---|---|
| Grayscale BTC (Mini) | 0.15% | Not disclosed on page | Not separately listed |
| Bitwise BITB | 0.20% | $3.32B (Sept 21) | $21.6M |
| iShares IBIT | 0.25% | $67.08B (Sept 23) | $731.7M |
| Fidelity FBTC | 0.25% | Not disclosed on page | $496.2M |
| Grayscale GBTC | Legacy vehicle | -$27.84B cumulative flow | $3.3M |
IBIT’s scale buys tightness. Its page lists a 30-day median bid/ask spread of 0.02% — on a $10,000 round trip, roughly $2 of friction. A smaller fund with a wider spread can hand back the 10 basis points it saved you on the fee in a single entry and exit.
The same page showed a premium/discount of -0.21% as of Sept. 22, 2026. That is a $21 mark below NAV on $10,000 at that snapshot. It is not a recurring fee, and it can invert — but it is a real cost of entering and exiting that no expense ratio captures.
We ran similar arithmetic on the commodity side in our gold ETF fee comparison, where the fee gap between the cheapest and most popular fund was wider than it is here.
What does self-custody cost that the fee table doesn’t show?
Operational risk, and it is not small. Two events this month priced it. On Sept. 7, 2026, CoinDesk reported a $320 million exploit on the Liquid Network, used by exchanges. On Sept. 18, CoinDesk reported a targeted attack on Haruko, a London institutional crypto infrastructure provider.
Haruko said 15 non-whitelisted clients were affected, with read-only exchange API credentials, trading data and “a small amount of client funds” taken. The firm serves more than 80 clients globally.
Those were institutions with security staff. A retail holder running their own keys has a smaller attack surface but no recovery desk, no insurance and no one to call.
The costs you should add to the self-custody column
- Hardware. A dedicated signing device is a one-time purchase, plus a backup.
- Network fees. Every withdrawal and every future consolidation pays the chain.
- Time. Seed backup, geographic redundancy, periodic verification, an inheritance plan.
- Tail risk. Loss of keys is absolute. There is no 0.25% that buys it back.
Against that, self-custody removes counterparty exposure to a sponsor, a custodian and an authorized participant chain. The honest framing is that you are swapping one risk for another, and paying $25 a year for the swap.
Is the bitcoin ETF vs self-custody choice different by account type?
Yes, and account type often decides it before cost does. Tax-advantaged accounts cannot hold coins directly at most brokers, which makes the fee comparison irrelevant there. Taxable accounts have a real choice, and short horizons favor the wrapper.
| Investor situation | Cheaper on the numbers | Why |
|---|---|---|
| IRA or 401(k) | ETF | Direct coin custody generally unavailable |
| Holding under 2 years, taxable | ETF | 0.80% entry toll not yet amortized |
| Holding 5-plus years, taxable | Self-custody | One-time 0.40–0.80% beats 0.15–0.25% a year |
| Trading around a position | ETF | 0.02% median spread on IBIT; no on-chain fees |
| Wants no counterparty | Self-custody | No sponsor, custodian or AP chain |
| Wants recourse if something breaks | ETF | Regulated wrapper, broker statements, estate handling |
One tax wrinkle is moving. Congress advanced a digital-asset tax package in mid-September 2026 that would apply wash-sale rules to crypto — we covered the vote in our piece on the crypto wash sale rule. If it becomes law, one of self-custody’s practical advantages over the ETF wrapper narrows.
Investors comparing wrappers on other assets may find our MSTR vs IBIT breakdown and our Solana and XRP ETF comparison useful, since both deal with the same wrapper-versus-asset cost question.
Frequently asked questions
Does a bitcoin ETF actually hold bitcoin?
US spot bitcoin ETFs hold bitcoin through a custodian. You own shares in the trust, not keys. That is the structural difference the fee is buying.
Which spot bitcoin ETF is largest?
IBIT, by a wide margin. Its iShares page listed $67.08 billion in net assets as of Sept. 23, 2026, against $3.32 billion at Bitwise’s BITB as of Sept. 21.
How long until self-custody is cheaper than IBIT?
On a flat $10,000, 3.2 years at Kraken’s 0.80% taker rate and 1.6 years at the 0.40% maker rate, before network fees and hardware. Rising prices shorten the payback, because the ETF fee scales with value.
Do bitcoin ETFs trade at a premium or discount?
They can do both. IBIT’s page showed -0.21% as of Sept. 22, 2026 — a discount to NAV. Creation and redemption normally keeps the gap narrow, but it is never exactly zero.
Is the lowest-fee bitcoin ETF always the cheapest to own?
No. Total cost is the sponsor fee plus the spread you pay entering and exiting, plus any premium or discount. A wider spread on a smaller fund can outweigh a 10 basis point fee saving for an active holder.
Can I hold bitcoin directly in an IRA?
Most mainstream brokerage IRAs do not support direct coin custody, which is why the ETF wrapper dominates retirement allocations regardless of the fee math. Specialized self-directed custodians exist and carry their own fee schedules.
What did bitcoin ETFs take in this week?
$1.71 billion across Sept. 21–22, 2026, per Farside Investors — $999.0 million then $714.7 million. Bloomberg reported the complex had turned positive after a $4.6 billion rebound.
The bottom line
On cost alone, self-custody wins any holding period longer than roughly three years, and wins it decisively by year ten: $80 against $250 in IBIT on a flat $10,000.
It depends on exactly two things, and they are both knowable. First, your account type — an IRA settles it for you. Second, whether you will actually execute key management correctly for a decade, because the failure mode there is total and the ETF’s is not.
The uncomfortable finding is in the flow table. Buyers put 71.7% of $1.71 billion into the two most expensive funds on the list while the 0.20% option took 1.3%. Whatever this week’s money was optimizing for, it was not cost.
This article is journalism, not investment advice. Do your own research before investing.
Sources
- iShares Bitcoin Trust ETF (IBIT) fund page — sponsor fee, net assets, spread, premium/discount
- Farside Investors — Bitcoin ETF Flow — daily and cumulative net flows
- Bitwise Bitcoin ETF (BITB) — 0.20% sponsor fee, AUM
- Grayscale Bitcoin Mini Trust ETF (BTC) — 0.15% sponsor fee
- Fidelity Wise Origin Bitcoin Fund (FBTC) — 25 basis point expense ratio
- Kraken fee schedule — entry-tier maker/taker spot fees
- CoinDesk — Haruko cyberattack, Sept. 18, 2026
- CoinGecko — Bitcoin price — $84,366 as of Sept. 23, 2026