How Spot Bitcoin ETFs Actually Work: The Mechanics, the Fees, and the Fine Print

The one-sentence take

A spot Bitcoin ETF is a grantor trust that holds real bitcoin and sells you shares in it — the cleanest Wall Street wrapper for BTC exposure, at the price of a yearly fee, a custodian in the middle, and a tax quirk most buyers never see coming. It is not "Bitcoin in your brokerage account." It is a fund that holds Bitcoin for you, and the fine print — fee drag, phantom income, custody concentration — is where the actual cost lives.

What it actually is

A spot Bitcoin ETF is a grantor trust registered under the Securities Act of 1933 — not a 1940-Act investment company, not a commodity pool (verified — IBIT prospectus). That structure matters: it is passive, holds no yield, runs no leverage or derivatives, and exists for one job — to hold bitcoin and track its price.

The load-bearing sentence is in the prospectus: "Individual Shares will not be redeemed by the Trust." You never hold bitcoin, you never hold a key, and you can never take the coins out. You own a share of a fund that owns the keys. That is the entire trade, stated plainly.

How shares get created (verified — IBIT S-1/424B3): shares are issued and redeemed only in blocks of 40,000 (a "Basket"), and only by Authorized Participants — registered broker-dealers and DTC participants. At launch the process was cash-only (APs deliver cash, the trust buys BTC via trading counterparties like Cumberland DRW, Flow Traders, JSCT, Virtu, or Coinbase Prime); in-kind (bitcoin) creation/redemption was added later. The AP arbitrage — buying the ETF when it trades below NAV and redeeming, or creating when it trades above — is what keeps the market price pinned to the bitcoin it holds.

Who holds the bitcoin

The custody stack is the part most buyers never read (verified — IBIT prospectus + issuer disclosures):

• Coinbase Custody Trust Company is the Bitcoin Custodian; Coinbase Inc. is the Prime Execution Agent; BNY Mellon is the Cash Custodian.
• No rehypothecation: the trust "will not loan, pledge or rehypothecate the Trust's assets" (with a narrow Trade Credits carve-out for cash-redemption settlement). The bitcoin is not lent out, not re-pledged, not used to generate yield.
• But the concentration is real: roughly 84% of US spot BTC ETF assets ($77B of the ~$91.7B market, Apr 2026, Reported — secondary market-data sources; Coinbase itself disclosed custody of 9 of 11 at launch) sits with Coinbase. At the fund level, most of the 12 US spot BTC ETFs custody via Coinbase or Coinbase-affiliated custody (Reported — issuer disclosures, 2026-08-20); only two are fully independent — FBTC (Fidelity Digital Assets, sole custodian) and HODL (Gemini). One custodian, most of the market.

The fee is the second price

Every spot ETF charges a sponsor fee, paid daily in bitcoin — the trust sells BTC to pay it. That is the mechanism behind the fee's real cost: it is not a cash deduction from your account, it is a slow leak from the bitcoin itself.

The fee ladder (verified — issuer disclosures, 2026-08-20):

• Low-cost tier: MSBT 0.14% (Morgan Stanley), BTC 0.15% (Grayscale Mini), EZBC 0.19% (Franklin Templeton)
• Mid tier: BITB 0.20%, HODL 0.20%, ARKB 0.21%, IBIT/FBTC/BTCO/BRRR/BTCW 0.25%
• The outlier: GBTC 1.50% (Grayscale's converted trust)
• All launch fee waivers have expired — HODL's, the last, ended Jul 31 2026. Fee comparison is now apples-to-apples.

The compounding cost (illustrative — Reported, btcetfcalc): on $10K held 10 years, the 0.14%→0.25% gap is roughly $350; the GBTC 1.50% gap is roughly $4,500. Fee drag compounds in bitcoin terms, because of the next point.

BTC-per-share decay (verified — prospectus): "The amount of bitcoin necessary for the creation of a Basket... will decrease over the life of the Trust, due to the payment or accrual of fees." Every share's bitcoin backing shrinks by the fee, forever. Two funds holding the same BTC price can drift apart on BTC-per-share purely on fee. That is the correct metric for comparing fee drag — more than the headline expense ratio.

The tax quirk nobody mentions: phantom income

This is the most differentiated fact in the whole category (verified — VanEck official blog, Cohen & Co, CoinTracking):

Because a spot ETF is a grantor trust, the IRS "looks through" it — shareholders are treated as owning a pro-rata share of the underlying bitcoin. And because the trust sells bitcoin to pay its sponsor fee, that sale is allocated to shareholders as taxable gain or loss even if you never sell a share. You can owe tax on a fund you haven't touched. That is phantom income — and GBTC's 1.50% fee means a larger expense-driven sale, and a larger phantom event, every year.

Two more tax wrinkles (US-specific, evolving — consult a tax professional):

• Wash-sale: direct BTC is property and the wash-sale rule currently doesn't apply; ETF shares are securities, and brokers apply wash-sale adjustments in practice, with IRS guidance still unclear.
• Form 1099-DA is new in 2026 — digital-asset reporting is being formalized as you read this.

The honest caveats

1. "ETF shares are not self-custody" — Pledge's own methodology page says it. No keys, no redemption, no "not your keys not your coins" escape hatch. The custodian is a single point of institutional trust, and most of the market shares the same one.
2. Closure risk is real. Hashdex DEFI was liquidated Aug 2026 — the first-ever spot BTC ETF closure (ceased trading Aug 17, cash distribution ~Aug 24). Small funds can close; scale is a feature.
3. Tracking is tight but not perfect. Benchmark is the CME CF Bitcoin Reference Rate (NY variant); IBIT tracks to ~3 bps (repo, 2026-06-20) and its premium/discount sits near zero (IBIT −0.07% Aug 2026). But tracking error and premium/discount are their own moving parts — the ETF can deviate from the bitcoin price even when the bitcoin price is flat.
4. Institutional money is real but shallower than headlines. 13F-filer (professional) holdings fell from 24.7% → 20.8% of spot ETF AUM in Q1 2026, and ~4 of every 5 dollars in the category are still non-professional (Reported — CoinShares 13F Q1 2026). The "institutions are all in" story is a directional truth, not a completed one.

How to read Pledge's ETF score

Pledge scores spot ETFs on eight weighted factors (verified — repo data-integrity.test.ts + methodology page):

The 10/10 anchor (verified — repo perfect-10 blueprint): a durable 15 bps fee (the market floor — nothing charges less durably), greater than $1B/day liquidity with sub-2 bps spreads, less than 2 bps tracking error, price pinned to NAV, greater than $50B AUM, a Tier-1 issuer, fully transparent fees, and mature multi-custody. The repo's one-liner: "A perfect ETF score would be cheap, liquid, tightly tracked, and structurally boring."

The closest real product — and the split that defines the category: IBIT scores 8.7 (verified — repo, 2026-06-20): 0.25% fee, ~$47B net assets (ishares.com, Aug 14 2026), ~735K BTC (Reported — third-party, Jul 15 2026), tightest spreads (0.03%), deepest options market, BlackRock behind it. MSBT (0.14%, Morgan Stanley) is the fee leader but tiny ($419M, Aug 14 2026). No fund has both the floor fee and the scale — MSBT has the fee, not the scale; IBIT has the scale, not the floor fee. That split is the tradeoff the score exists to show. Live range ~4.3–8.7 (DEFI 4.3, liquidated Aug 2026).

Read it like this: the score is category-relative (an ETF 8.7 is not a loan 8.7), point-in-time (repo DATA_AS_OF 2026-06-20), and a map, not a verdict. A high score means the fund is boring in the ways that matter — cheap, liquid, tightly tracked, well-custodied. It does not mean "safe," and it is not a buy rating. Read the factor breakdown, not the headline.

The take

A spot Bitcoin ETF is the cleanest way to hold BTC inside a brokerage or IRA — and the fee, the custodian, and the phantom-income quirk are the price of that convenience. Compare on fee (in BTC terms), custody concentration, and tracking, not on the ticker. The wrapper is the product; the bitcoin is the asset. Risk is the product.