Bitcoin ETF Premium and Discount to NAV: What It Means and When to Worry

May 27, 2026• 18 min

Spread, premium and closing drift measured across all twelve US spot Bitcoin ETFs, the GBTC discount that reached 48.9%, and what to check before you place an order.

TL;DR. A US spot Bitcoin ETF trades within a few basis points of net asset value, and the gap is closed by authorised participants rather than by anything you can trade. On 22 September 2026 the eleven funds that publish the figure showed a 30-day median bid/ask spread between 0.02% and 0.07%, against 0.37% to 0.46% at the small altcoin trusts. Where they differ is at the closing bell: measured against IBIT over the twelve months to that date, Fidelity's FBTC sat 0.16% from its usual relationship on a typical day and WisdomTree's BTCW 0.64%, four times as far. GBTC is the extreme case, a closed-end trust that closed at a 48.9% discount on 13 December 2022 and has given up $27.84bn since it became redeemable in January 2024. Practical version: use a limit order, skip the opening minutes (five-minute ranges in IBIT run about 2.5 times the midday figure), and read the 30-day spread before anything large.

What NAV, premium and discount actually mean

A spot Bitcoin ETF has two prices, built differently. Net asset value is an accounting number struck once per business day: take the bitcoin the trust holds, value it against a reference rate, subtract the accrued sponsor fee, divide by shares outstanding. That reference rate is an afternoon New York average rather than a single exchange print, which is why NAV stays calm on a violent day; IBIT prices against the CME CF Bitcoin Reference Rate, New York Variant, over the hour to 4pm Eastern. Market price is whatever somebody last paid on the exchange. It updates from 9:30 to 16:00 Eastern and stops dead outside those hours, while bitcoin keeps trading.

The premium or discount is the difference expressed against NAV: (market price − NAV) ÷ NAV. Above NAV is a premium, below is a discount. On a well-arbitraged fund it flips sign constantly and means nothing on a single day; persistence and size are what matter. If you're new to the wrapper, start with what a Bitcoin ETF actually is.

Where the numbers live, and which one to trust

Every US issuer publishes the same small set of figures on the fund page, with a one-day lag and an as-of date. They're not equally useful.

Closing premium or discount. One day's snapshot. Noisy, sign-flipping, and the number most often quoted out of context.
30-day median bid/ask spread. The best single figure for a buyer, because it's what a round trip costs right now, and a median doesn't move on one bad afternoon.
Quarterly premium/discount table. Days closed at a premium versus a discount over the last four quarters. This is the tracking-quality number, and it separates a fund that wanders from one that doesn't.

One warning before comparing issuers on a single day's premium. On 22 September 2026 five funds published a figure within a hair of each other at −0.20% to −0.21%, while Grayscale published exactly 0.00% and Invesco +0.01%. That isn't real dispersion, it's a convention difference: some measure the closing price against NAV, others the 4pm bid/ask midpoint, and NAV strike times vary. Compare spreads and quarterly tables instead.

How far the twelve US Bitcoin funds actually drift

Issuer pages give you one fund at a time, on the issuer's chosen date, so we put them side by side and added a measurement of our own. All twelve hold the same asset, so any two should keep an almost constant price ratio, drifting only by the difference in their fees. When a fund's close leaves that ratio, the move is its own pricing noise: a real premium or discount, a wide closing quote, or a stale last print. We took IBIT as the yardstick and measured how far each fund's daily price ratio to it sat from that pair's trailing 20-session median, over the twelve months to 22 September 2026, a window that keeps the Grayscale Mini Trust spin-off of July 2024 out of the sample.

Fund	Fee	Net assets, issuer page	30-day median bid/ask spread	Our measured daily dispersion vs IBIT	Widest single close
IBIT (BlackRock)	0.25%	$68.34bn	0.02%	benchmark	—
FBTC (Fidelity)	0.25%	$13.87bn (31 Aug)	0.04%	0.16%	0.81% (30 Jul 2026)
GBTC (Grayscale)	1.50%	$10.97bn	0.02%	0.18%	0.74% (14 Nov 2025)
BTC (Grayscale Mini)	0.15%	$5.42bn	0.03%	0.18%	0.60% (13 Jul 2026)
BITB (Bitwise)	0.20%	$3.32bn (21 Sep)	0.02% (20 Sep)	0.23%	1.18% (1 Jul 2026)
ARKB (ARK 21Shares)	0.21%	$2.66bn (31 Aug)	0.04%	0.23%	1.63% (7 Mar 2026)
HODL (VanEck)	0.20%	$1.27bn	0.04%	0.30%	1.14% (24 Oct 2025)
MSBT (Morgan Stanley)	0.14%	$759m (see note)	0.04%	0.60%	4.50% (28 May 2026)
EZBC (Franklin)	0.19%	$479m	not published	0.55%	2.16% (5 Feb 2026)
BTCO (Invesco Galaxy)	0.25%	$440m	0.07%	0.58%	2.57% (5 Feb 2026)
BRRR (CoinShares)	0.25%	$437m (17 Sep)	0.05% (17 Sep)	0.61%	3.15% (5 Feb 2026)
BTCW (WisdomTree)	0.25%	$182m	0.06%	0.64%	3.03% (7 Apr 2026)

Fees, net assets and spreads come from each issuer's own fund page on 22 September 2026 unless the cell says otherwise. Dispersion and widest-close figures are our own calculation from daily closing prices, 1 September 2025 to 22 September 2026, roughly 270 sessions per fund and 96 for MSBT, which listed on 8 April 2026. Morgan Stanley's page contradicts itself on that date, reporting $758.94m of net assets beside a holdings line worth about $858m, so treat its AUM as unconfirmed.

Read the last two columns against each other, because they disagree by an order of magnitude and both are right. Eleven of the twelve quote a 30-day median spread between 0.02% and 0.07% (Franklin doesn't publish one), so by that measure spot Bitcoin ETFs are a solved problem: the widest, BTCO at 0.07%, is three and a half times the tightest, and the whole range sits five to twenty times inside the small altcoin trusts, where Canary's LTCC publishes 0.46%, its Hedera fund HBR 0.39% and its Sui fund SUIS 0.37%. Yet our closing-price dispersion runs 0.16% at FBTC and 0.64% at BTCW, a four-to-one gap that tracks assets closely and ignores fees completely.

The difference between the two is the useful part. A quoted spread is what a market maker shows when it's showing; dispersion is where the day's last print actually landed, which also absorbs staleness and the mismatch between a 4pm close and a NAV strike. In a fund that trades all day the two nearly coincide, and in one that doesn't they can't: across 206 sessions to 22 September 2026 BTCW printed a trade in only 28.1% of the 78 five-minute windows in a session, against 100% for IBIT and 99.1% for GBTC. A tight quote you can't reach isn't a tight market.

The issuers' quarterly tables agree. IBIT closed 2025 at a premium on 121 days and a discount on 124, which is what good tracking looks like: a coin flip. ARKB's third quarter of 2026 ran 36 premium days to 20 discount days, widest premium 0.42% on 3 September, widest discount 0.71% on 29 July, with 98.28% of days inside half a percent of NAV. And GBTC's 1.50% fee buys no tradability penalty at all: at 0.02% its spread ties IBIT's and beats the Mini Trust's, because the fee shows up in NAV rather than in the quote.

Which leaves one practical reading: paying 0.6% to enter a $182m fund in order to save six basis points a year is a bad trade unless you'll hold for a decade. Fees and assets fund by fund are in the full directory of US spot Bitcoin ETFs, with live pages on our funds directory.

GBTC: the most expensive discount in crypto

Everything above describes funds where the arbitrage works. Grayscale's trust is the case study in what happens when it doesn't, and the reason this topic gets attention at all. The Bitcoin Investment Trust launched in 2013 and began trading on OTCQX in 2015, with shares created through private placements to accredited investors who could resell publicly after a lockup. What the structure never had was redemption: you couldn't hand shares back and receive bitcoin. Creation worked in one direction only.

That asymmetry produced both extremes. When brokerage demand outran the slow, gated supply of new shares, GBTC traded above the value of its coins, and buying at NAV in a placement to sell at a premium six months later became a crowded trade at several large crypto lenders. When demand reversed in 2022 nothing could pull the price back up, and the discount widened to 48.9% on 13 December 2022, weeks after FTX failed and Grayscale's sister company Genesis halted withdrawals. A share representing $20 of bitcoin fetched about $10, and the discount stayed below 30% through most of 2022 and the first half of 2023.

Nothing about bitcoin caused that. The trust held the coins throughout and its NAV tracked bitcoin normally. The 48.9% was purely the price of a broken redemption mechanism, which is the most important thing this article has to say: a premium or discount is a statement about the wrapper, not about the asset.

The discount closed on the calendar of the legal case, narrowing through the second half of 2023 as the DC Circuit's August 2023 ruling made conversion look likely, and reaching single digits by the end of December. On 11 January 2024 GBTC began trading on NYSE Arca as a redeemable ETF and the gap vanished, because that morning an authorised participant could finally buy a discounted share, redeem it for bitcoin and sell the bitcoin.

Then the bill arrived. Holders trapped for years could finally leave, and a 1.50% fee gave them a reason to. By our own daily flow data GBTC lost $7.44bn in its first 30 sessions as an ETF, $21.50bn across 2024, and $27.84bn through 22 September 2026, with 181 of 2024's 245 sessions negative and a worst day of $642.5m out on 18 March 2024. Most of it rotated to cheaper funds rather than leaving the category, a pattern unpacked in our Grayscale GBTC analysis. The same film ran on the ether side, where Grayscale's ETHE traded at a reported premium in the hundreds of percent in mid-2020 and near a 59% discount at the end of December 2022.

How the arbitrage closes the gap, and why a small fund closes it slower

Retail buyers never close a premium. Authorised participants do, and only in whole baskets: above NAV an AP delivers bitcoin or cash to the trust, receives new shares and sells them into the premium, and below NAV the trade runs in reverse. Each round trip is riskless in principle and costs real money in practice, which is where the size effect comes from. The full walkthrough is in how authorised participants work. Three frictions set the width of the band.

The basket is indivisible. IBIT's creation unit is 40,000 shares, EZBC's 50,000, and several small funds use 5,000. A 0.3% premium on 40,000 IBIT shares is worth chasing; the same 0.3% on 5,000 shares of a fund that turns over a few hundred thousand dollars a day is worth a few hundred dollars, and the AP still has to unwind into a book that can't absorb it.
The hedge costs something. Bitcoin moves between the moment an AP quotes and the moment the trust prices the order. Someone pays for that, and the wider bitcoin's realised volatility, the wider the quote.
The book has to exist. This is the one people underestimate, and it's the 28.1% figure above. In a fund that goes untraded for most of the day, the price on your screen can be an hour old, and the gap between it and NAV isn't a premium so much as an absence.

MSBT shows it from the launch side. Morgan Stanley's fund, at 0.14% the cheapest in the category, printed a close 4.5% away from the rest of the market on 28 May 2026, seven weeks after listing. Whether that was a genuine premium or one thin trade at a bad moment hardly matters: a market order could have hit it.

What in-kind creation changed

The 2024 cohort launched cash-only: an AP handed over dollars and the fund itself bought the bitcoin, which put the fund's execution slippage inside NAV and forced the AP to carry a hedge across the gap. Both widen the band the price can wander in. On 29 July 2025 the SEC permitted in-kind creations and redemptions for spot crypto ETPs, letting APs deliver and receive the coins directly.

Our dispersion measure across nine funds, pooled by day, moved in the expected direction: the median session ran 0.326% from September 2024 to 28 July 2025 and 0.244% afterwards, a tightening of about a quarter.

Period	Sessions	Median daily dispersion (9 funds vs IBIT)	Median absolute daily bitcoin move
Sep 2024 to 28 Jul 2025 (cash only)	225	0.326%	1.64%
29 Jul 2025 to 22 Sep 2026 (in-kind permitted)	289	0.244%	1.48%

Our own calculation from daily closing prices and our bitcoin price series. Be careful with the conclusion: bitcoin also got calmer, with the median absolute daily move falling from 1.64% to 1.48%, so some of the tightening has nothing to do with settlement mechanics. Dispersion fell by roughly 25% while volatility fell by roughly 10%, which is consistent with in-kind helping and is not proof of it.

When the gap actually widens

Volatility days

The worst session in our twelve-month window was 5 February 2026, when average dispersion across the ten non-IBIT funds hit 0.96%, four times a normal day's 0.24%, and EZBC, BTCO and BRRR all recorded their widest close of the year on that one date. What followed is the point: bitcoin fell 14.7% the next day, from $73,066 to $62,293. The gap widened before the move, not after it, because market makers priced the risk first.

Weekend gaps

Bitcoin trades continuously and the funds don't, so every Monday open absorbs whatever happened since Friday's close. The largest such gap in our series was 6 to 9 February 2026, plus 12.9%, from $62,293 to $70,307 with no session in between; 18 to 21 September gapped 6.5%. A premium printed in the first minutes of a session like that is a stale reference point, and it resolves as soon as real size trades.

The first minutes of the session

This one is measurable, and it's the advice that actually saves money. In five-minute candles across recent sessions, the high-to-low range inside each candle decays steadily through the day.

Minutes after the 9:30 ET open	IBIT median 5-min range	GBTC median 5-min range
0 to 5	0.42%	0.37%
5 to 15	0.47%	0.31%
15 to 30	0.36%	0.30%
30 to 60	0.31%	0.24%
60 to 330 (mid-session)	0.20%	0.14%
330 to 385 (final hour)	0.16%	0.11%

Our own calculation from five-minute candles: IBIT over 26 sessions to 22 September 2026, GBTC over 42. In both, the opening quarter hour runs roughly two and a half times as wide as mid-session, and the quietest stretch is the final hour, not the first. The familiar advice to avoid the close as well doesn't hold up for the large funds, and there's a reason to prefer late in the day: NAV is struck in the New York afternoon, so a trade near the close is priced against a reference rate already being set. In a thin fund the calculus differs, because the closing auction is where a stale book prints.

The first weeks of a new listing

A new fund has few APs, no float and no natural two-sided flow. MSBT's first fifteen sessions from 8 April 2026 ran about 0.25% above where the rest of the market implied, peaking at 0.65%, and Morgan Stanley's own table shows the same skew over a longer run: since inception the fund has closed at a premium on 82 days against a discount on 26. That's a one-directional bias rather than the sign-flipping noise a mature fund shows, and a reason not to rush a launch.

What it costs you, and what to do about it

Buy at a 0.3% premium and sell at a 0.3% discount and you've paid 0.6% in slippage, which on a 0.25% fund is more than two years of management fee surrendered on two trades. Over five years the fee dominates and slippage rounds away; on a position you rebalance quarterly it's the larger cost, and nobody quotes it to you. That's why we work in total cost of ownership rather than the headline number in Bitcoin ETF expense ratios compared. Four rules cover almost all of it.

Use a limit order, always. A market order is an instruction to accept whatever the widest quote in the book happens to be. In the tight funds that costs a few basis points; in a $182m fund it can cost more than a year of fees.
Don't trade the first fifteen minutes. The table above is the argument. Let the overnight bitcoin move get priced first.
Read the 30-day median spread, not yesterday's premium. On a Bitcoin fund anything above 0.07% is an outlier; on a small altcoin trust the same figure runs 0.4%.
Below about $1bn, size matters more than fee. Six basis points saved is roughly $6 a year per $10,000; half a percent of entry slippage is $50, once, up front.

Order mechanics broker by broker are in how to buy a Bitcoin ETF, and the wider screen we run before choosing a fund is the selection checklist. The one signal worth acting on is a persistent, one-directional premium across weeks, because it means creation isn't keeping up. The fastest way to check is our Bitcoin ETF flows page, which is nothing more than the daily change in shares outstanding at each fund.

Strip the noise out and a spot Bitcoin ETF returns bitcoin's spot return minus the sponsor fee, give or take a few basis points of friction, which is more than futures-based products manage: their roll costs have nothing to do with the fee and have missed bitcoin by percentage points, as spot versus futures Bitcoin ETFs sets out. So a small premium or discount is normal and self-correcting, a persistent one is a liquidity statement about a particular fund, and a huge one means the redemption mechanism is broken. Only the third case is a reason to worry, and no US spot Bitcoin ETF has been in it since January 2024.

FAQ

What is a Bitcoin ETF premium or discount to NAV?

It's the difference between the ETF's market price and its net asset value, the accounting value of the bitcoin behind one share, expressed as a percentage of NAV. Above NAV is a premium, below is a discount. US spot Bitcoin ETFs trade within a few basis points either way, and the number flips sign from day to day.

How big is the premium or discount on a typical spot Bitcoin ETF?

Small. Eleven of the twelve US spot Bitcoin ETFs published a 30-day median bid/ask spread between 0.02% and 0.07% on 22 September 2026. Closing prices wander further: measured against IBIT over the prior twelve months, FBTC deviated about 0.16% on a typical day and the sub-$500m funds 0.55% to 0.64%, with widest single closes of 2% to 4.5%.

Why did GBTC trade at a 50% discount?

Because until January 2024 it was a closed-end trust with no redemption mechanism. Shares could be created in private placements but never handed back for bitcoin, so when demand fell there was no arbitrage to pull the price up. The discount reached 48.9% on 13 December 2022, after FTX collapsed and Genesis halted withdrawals, and it closed only when GBTC converted to a redeemable ETF on 11 January 2024.

What happened to GBTC after the discount closed?

Holders locked in for years could finally exit, and the 1.50% fee gave them a reason to. By our own daily flow data GBTC lost $7.44bn in its first 30 sessions as an ETF and $27.84bn through 22 September 2026, with 181 of 2024's 245 sessions negative. Most of it rotated into cheaper funds rather than leaving the category.

How do I check the premium or discount before I buy?

Go to the issuer's fund page. Every US issuer publishes the closing price, end-of-day NAV, the premium or discount, a 30-day median bid/ask spread and a quarterly table of days closed at a premium versus a discount. Use the spread and the quarterly table. A single day's closing premium is noise, and issuers don't even calculate it the same way.

Does the premium or discount eat into my returns?

On a multi-year hold the spread averages out and the sponsor fee dominates. On anything you rebalance often it is the larger cost: buying at a 0.3% premium and selling at a 0.3% discount costs 0.6%, more than two years of a 0.25% fee. Entry and exit slippage belongs in total cost of ownership alongside the expense ratio.

When is the worst time of day to buy a Bitcoin ETF?

The first fifteen minutes. In five-minute candles across recent sessions, IBIT's median intra-candle range was 0.42% in the opening five minutes and 0.47% over the next ten, against 0.20% mid-session and 0.16% in the final hour, and GBTC shows the same shape. The overnight bitcoin move has to be priced in first, and a market order into that is an expensive way to find out where the price settled.