The 9.75% Float Lock: Auditing the 1.96 Million BTC Sitting Inside US Spot ETFs

Hook

A single number crossed my terminal on September 14, 2025, and it was not a price. It was a stock figure.

US spot Bitcoin ETFs held 1,959,000 BTC on-chain. That is 9.75% of the mined supply. At the print, that stack was worth $221.4 billion.

Run the division before you read anything else: 221,400,000,000 ÷ 1,959,000 = $113,017 per BTC. That implied unit price is the timestamp. Bitcoin did not trade near $113,000 in September 2024, when 1.96 million coins would have been worth roughly $118 billion. It traded there for a large stretch of late 2025. The headline is not dated, but the arithmetic dates it. [Confidence: Medium]

Retail reads this as a green candle that has not printed yet. I read it as a custody concentration statement. Nine and three-quarter percent of a hard-capped, trust-minimized monetary asset is now sitting inside a compliance wrapper, behind a custodian's key ceremony, redeemable on a settlement cycle that has nothing to do with block time.

Alpha isn't in the headline. It's in the float.

Here is the part the headline buries: 1.959 million coins is larger than the estimated 1.1 million coins attributed to Satoshi Nakamoto's untouched wallets. The US ETF complex is now the single largest identifiable holder class in the history of the asset. That is a structural fact with a mechanical consequence, and almost nobody is pricing the consequence.

Context

What the instrument actually is

A US spot Bitcoin ETF is not a piece of the Bitcoin network. It is a registered fund — most are commodity trusts registered under the '33 Act, with a parallel set structured under the '40 Act — whose shares trade on a national securities exchange and whose underlying asset is held by a qualified custodian.

That sentence contains the entire risk architecture. The ETF is a TradFi-to-crypto bridge built out of legal paper and a custodian's cold storage. Chain-level finality is not what guarantees your claim. Delaware trust law, an SEC-registered prospectus, and a custody agreement are.

This is the fundamental trade every ETF buyer makes, consciously or not: they swap trust minimization for regulatory legibility. The network's security model is proof-of-work and key possession. The ETF's security model is a Big Four auditor, a custodian's SOC 2 report, and a compliance department.

The launch anchor

On January 10, 2024, the SEC approved eleven spot Bitcoin ETFs simultaneously, ending a decade of rejections and settling the underlying question after Grayscale won its mandamus petition against the Commission in August 2023. The launch set in motion a mechanical bid that had no precedent.

By the time of this print, the complex had absorbed 1.959 million BTC — approximately $221.4 billion in assets under management, spread across a dozen or so surviving products.

The settlement plumbing you must understand

This is where most readers lose the thread, so I will be precise.

US spot Bitcoin ETFs operate on a cash create / cash redeem model. In an in-kind model, an Authorized Participant hands the issuer shares of the underlying asset and receives ETF shares in return. In the cash model — which is what the SEC mandated at launch — the flow is different:

1. The AP wires US dollars to the issuer.
2. The issuer instructs the custodian to buy BTC in the spot market.
3. New ETF shares are created and delivered to the AP.
4. Redemptions run the reverse sequence: shares in, cash out, and the custodian sells BTC in the spot market to fund the wire.

Hold that sequence. It is the source of the most expensive analytical error in this entire sector, and I will return to it in the Core section.

The structural players

| Issuer | Flagship Product | Approx. Structural Position | Custody Model | Fee (typical) | |---|---|---|---|---| | BlackRock | IBIT | Largest share of complex, ~40%+ | Coinbase Custody | 0.25% | | Fidelity | FBTC | Second tier | Fidelity Digital Assets (self) | 0.25% | | Grayscale | GBTC | Legacy, structural net outflow | Coinbase Custody | 1.50% | | Ark / 21Shares | ARKB | Mid-tier | Coinbase Custody | ~0.21% | | Bitwise | BITB | Mid-tier | Coinbase Custody | ~0.20% | | VanEck, Invesco, Franklin, WisdomTree, Valkyrie, Hashdex | Various | Tail | Mostly Coinbase Custody | 0.19–0.25% |

[Confidence: Medium — the table reflects the industry's commonly cited structure, not the underlying data point, which disclosed only aggregate holdings. Treat the fee column as directional, not contractual.]

Every row of that custodian column matters more than the fee column. Notice how many times Coinbase Custody appears.

What the data point does and does not say

The print gives exactly four facts: 1.959M BTC held, 9.75% of supply, $221.4 billion in value, and a date that is not written down but is recoverable from the arithmetic.

It does not give you:

• Net flow direction for the week or the month
• Issuer-level breakdown
• The price context in which the milestone was crossed
• Any custody concentration figure
• Any ratio to exchange float or to realized liquidity

That absence is not accidental. A stock milestone at a price high is a marketing document. A flow table at the same moment is a risk document. The two are not interchangeable, and only one of them moves markets.

Core: The Order Flow Audit

1. The Float Reconstruction

The 9.75% figure needs a denominator before it means anything. Work backward: 1,959,000 ÷ 0.0975 = 20,092,308 BTC. That is slightly higher than the roughly 19.9 million coins mined by late 2025. The gap is a denominator artifact — Dune-style supply queries frequently exclude provably burned or permanently lost coins from the circulating base. [Confidence: Medium]

So the honest read is: 9.75% of circulating, addressable supply is held inside ETF custody vehicles.

Now build the real float table. This is the table that should appear in every article written about this milestone and almost never does:

| Supply Tranche | Estimated Size | % of Mined | Tradeable? | Notes | |---|---|---|---|---| | Total mined | ~19.9M BTC | 100% | — | Post-April-2024 halving | | Permanently lost (est.) | ~3.0–4.0M BTC | 15–20% | No | Unrecoverable keys | | Satoshi-era dormant (est.) | ~1.1M BTC | ~5.5% | Effectively no | Untouched since 2010 | | US Spot ETF custody | 1.959M BTC | 9.75% | Only via redemption | This data point | | Long-term holder cohort (proxy) | ~14–15M BTC | ~70–75% | Slowly | Includes the two rows above | | Exchange float (est.) | ~2.0–3.0M BTC | 10–15% | Yes | True marginal supply | | Annualized new issuance | ~164,250 BTC/yr | ~0.83%/yr | Yes | 450 BTC/day |

Now read the third and sixth rows against each other. The ETF complex holds 65% to 98% of the size of the entire exchange float. The thing that is supposed to be the marginal price-setting venue — exchange order books — is now roughly the same size as a single holder class that does not trade on block time.

That is the structural shift. Not that ETFs bought coins. That the coins they bought now rival the venue where price discovery happens.

2. The Absorption Engine

Let me make the squeeze math explicit, because it is the most quotable number in this entire analysis and it survives scrutiny.

Post-halving issuance is 3.125 BTC per block, 144 blocks per day:

Daily issuance = 3.125 × 144 = 450 BTC/day

Weekly issuance = 450 × 7 = 3,150 BTC/week

Annual issuance = 450 × 365 = 164,250 BTC/year

Annual inflation rate = 164,250 ÷ 19,900,000 ≈ 0.83%

Now compare against a modest ETF tape. At $113,017:

| Weekly Net Inflow (USD) | BTC Absorbed | Multiple of Weekly Miner Issuance | |---|---|---| | $500M | 4,424 BTC | 1.40x | | $1.0B | 8,848 BTC | 2.81x | | $1.5B | 13,272 BTC | 4.21x | | $2.0B | 17,698 BTC | 5.62x | | $3.0B | 26,547 BTC | 8.43x |

Read the middle rows again. During the 2025 inflow regime, the ETF complex was routinely absorbing three to six times the entire new supply schedule. Miners are the only protocol-level sellers of new coins. When a demand channel consumes three to six multiples of their output, the clearing price must be set by existing holders who choose to sell, not by new supply.

That is the mechanical definition of a float squeeze. It is not sentiment. It is arithmetic.

And now the second-order number, the one that reframes the milestone: 1,959,000 ÷ 164,250 = 11.9 years.

The ETF complex has absorbed the equivalent of roughly twelve years of post-halving miner issuance. It did this in under two years of trading.

We do not chase pumps; we engineer the squeeze. The squeeze here was engineered by an index-inclusion mechanic, not by a narrative.

3. The Cash Create/Redeem Lag — The Sector's Most Expensive Error

Here is where I depart from consensus analysis, and it is a lesson I paid tuition to learn.

When I ran the 2017 pre-sale arbitrage on TokenMarket and Nexus Mutual, executing over 400 transactions across mainnet and OTC desks, the single hardest problem was not finding the spread. It was reconciling two ledgers that settled on different clocks. Mainnet finality moved on block time. OTC settlement moved on banking hours. For roughly eighteen hours each cycle, I was flying partially blind, and my P&L was determined by how honestly I modeled that blind spot.

The ETF complex has the identical problem at a hundred times the scale, and the market is modeling it wrong.

In a cash create/redeem structure, when the AP redeems:

• The issuer does not need to move BTC on-chain immediately.
• The custodian can sell BTC on off-exchange desks, OTC venues, or through CME-adjacent liquidity and wire fiat to the AP.
• The on-chain address that Dune tracks may show no change at all.

Therefore: "ETF on-chain holdings" is a smoothed, lagging, partially cosmetic metric. It is the right metric for measuring long-run accumulation. It is the wrong metric for measuring real-time sell pressure. [Confidence: Medium-High]

This is the trap. In a redemptions cascade, on-chain ETF balances can hold flat or even rise while the custodian is quietly liquidating into the spot market. A holder watching the on-chain chart sees "stable accumulation." A desk watching exchange netflows and CME basis sees the distribution.

The two are looking at different slices of the same event, and only one slice clears.

4. The Custodian Concentration Problem

The 1.959 million coins are not spread across two hundred custodians. They are spread across a handful, and the dominant share runs through one.

Build the exposure ladder:

| Assumed Coinbase Custody Share of ETF BTC | BTC Under Single Custodian | USD Value at $113,017 | |---|---|---| | 50% | 979,500 BTC | $110.7B | | 60% | 1,175,400 BTC | $132.8B | | 70% | 1,371,300 BTC | $155.0B | | 80% | 1,567,200 BTC | $177.1B |

[Confidence: Medium — Coinbase Custody is the disclosed custodian for the majority of the products in the complex; Fidelity self-custodies for FBTC. Exact attribution should be verified against each fund's most recent custody disclosure before sizing risk.]

Even the conservative 50% case puts $110 billion of a bearer asset behind one company's key management, one company's operational security, one company's legal entity, and one company's public-market equity risk.

Note the reflexivity: Coinbase is a listed company whose stock correlates with crypto prices. A severe crypto drawdown pressures its equity, pressures its cost of capital, pressures its hiring and security budget, all while it holds the keys to the asset everyone is watching. That is not a hypothetical chain. It is a named dependency graph.

Concentration is leverage. Leverage cuts both ways, and the liquidator here is a custody incident.

5. The GBTC Drag and the Fee War

The aggregate 1.959 million figure hides a violent internal rotation that only appears when you look at issuer-level BTC holdings over time.

| Period | GBTC Approx. BTC | IBIT/FBTC Direction | Net Complex Effect | |---|---|---|---| | Jan 2024 (conversion) | ~620K BTC | Launching | Neutral | | Mid 2024 | Declining sharply | Rapid inflow | Net positive | | End 2024 | Materially lower | Strong inflow | Net positive | | 2025 | Continued bleed | Dominant inflow | Net positive, composition shifting |

[Confidence: Medium — GBTC's structural outflow has been a persistent, well-documented feature since conversion; exact balances should be pulled from the issuer's daily holdings disclosure rather than from memory.]

GBTC charges 1.50%. IBIT and FBTC charge 0.25%. On a $10 billion position, that is a $125 million annual fee differential. No allocator of a 60/40 sleeve survives explaining that line item to an investment committee twice.

The rotation is not sentiment. It is arithmetic, and it is mechanical. It also means the aggregate stock number is a blend of sticky new money and fleeing legacy money, and the blend ratio determines how fragile the 1.959 million actually is.

The fee war compounds it. Several issuers ran promotional fees down to 0.15–0.25% for early assets, and the equilibrium fee on a commodity trust with zero active management is probably heading toward single-digit basis points over the medium term. [Confidence: Medium]

Ask the question nobody asks: at 5 basis points on $221.4 billion, the entire complex throws off roughly $1.1 billion in annual fees, and a meaningful chunk of that has to fund custody, audit, legal, market-making incentives, and exchange listing. The wrapper's economic engine is thinner than the headline AUM suggests.

6. The CME Basis — Where the Real Buyer Sits

Most ETF flow is not directional conviction. A large slice is basis trade structural demand.

The classic carry:

1. Buy the spot ETF (or spot BTC in a margin account).
2. Short the CME front-month futures contract.
3. Collect the basis as it converges.

This trade is market-neutral. It is also a guaranteed spot buyer and a guaranteed futures seller, which produces two visible fingerprints:

• CME open interest and volume rise structurally, as institutions need a regulated hedge venue.
• The futures basis compresses toward the cost of financing, because the arbitrage is mechanical and crowded.

The consequence for price discovery is what I want you to hold onto: CME's pricing weight over Bitcoin has risen to rival and in some sessions exceed that of the offshore perpetual venues. If you are still reading funding rates on an offshore perp as your primary sentiment gauge, you are reading the smaller book.

But note the fragility. Basis trade demand is not price-insensitive. It is yield-sensitive. When the annualized basis drops below the cost of carry — risk-free rate plus financing friction — the trade unwinds, and the unwind is a coordinated spot seller across every account running it.

That is the hidden torque in the 1.959 million number. A meaningful slice of that stock is held by entities that will exit on a spread, not on a thesis.

7. The Correlation Regime Change

Here is the trade-off almost nobody puts in a client memo.

A 9.75% institutional lock changes who sets the marginal price and what they care about. A retail- and native-dominated Bitcoin market prices on halving cycles, on-chain supply dynamics, and crypto-native liquidity shocks. An institution-dominated Bitcoin market prices on the same factors the institution's other risk assets price on: real rates, the dollar index, liquidity conditions, and equity beta.

Directional implication, plainly stated: BTC's long-run realized volatility should keep compressing, and its correlation to the Nasdaq complex should keep rising.

That is not a bearish claim. It is a claim that the asset's portfolio role changed. Bitcoin was sold to allocators as an uncorrelated inflation hedge. A 9.75% custody lock inside SEC-registered vehicles is the mechanism by which it becomes a high-beta, high-duration risk asset with an inflation narrative attached.

The day a liquidity crisis hits, it will trade like a liquidity asset, not like a hedge. Watch the correlation print in the first 72 hours of the next macro shock. That is the real test of the ETF era, and it has not been honestly stress-tested yet.

8. The DeFi Siphon

Now the part that matters to me professionally, because I allocate into DeFi yield, not into trust shares.

Every dollar that enters BTC through an ETF buys a custody receipt backed by a coin the ETF will never lend, stake, bridge, or collateralize. That coin exits the composable economy permanently.

The chain-level consequence:

| Venue | Effect of ETF Absorption | Mechanism | |---|---|---| | Wrapped BTC supply (WBTC-adjacent) | Downward pressure | Demand rerouted to ETF instead of on-chain wrapper | | BTC-collateralized lending | Thinner collateral base | Fewer coins available to post | | Bitcoin L2 / Ordinals / inscriptions | Reduced native activity share | Institutions do not touch them | | CEX spot liquidity | Structurally tighter | Coins leave exchange balances for custody | | CME derivatives | Structurally stronger | Hedge demand concentrates here |

Recall the 2024 wrapped-BTC custody transition drama, when a change in a single custodian arrangement for the largest wrapped BTC product triggered a visible migration of on-chain liquidity toward newer, exchange-native wrapped assets. That episode taught the market something the ETF flow data confirms: Bitcoin's financialization is consolidating into fewer, more centralized venues, whether those venues are custodians, exchanges, or CME clearinghouses.

There is a real arbitrage consequence here, and it is the one I trade. As ETF absorption tightens the on-chain wrapped-BTC float, the borrow rate for BTC in DeFi lending markets decouples from the ETF complex's flow entirely. Two markets, one asset, two independent supply curves. That decoupling is a spread, and spreads are where I live.

Contrarian

The consensus interpretation of this data point is wrong in a specific, exploitable way.

The consensus says: 9.75% of supply is locked away, supply is tightening, therefore structurally bullish.

The mechanism says: 9.75% of supply is parked in a redeemable wrapper with a fee-sensitive holder base and a cash-settled exit path. That is not a lock. That is a collateralized floating overhang — the largest in the asset's history — and its release valve does not require a single on-chain transaction.

Three things the bullish framing gets backwards:

First, the overhang is mispriced as a lock. Compare this to genuinely removed supply. Satoshi's 1.1 million coins have not moved in over a decade. Lost-key coins are gone forever. Exchange-traded coins are liquid. ETF coins are the fourth category nobody names: structurally idle, legally mobile. They sit still in a bull market because the fee-and-mandate structure rewards sitting still. In a drawdown, the same structure rewards leaving, because the holder is an allocator with a risk committee, a drawdown threshold, and a mandate to reduce volatility exposure. An allocator does not HODL. An allocator rebalances — and rebalancing after a 30% drawdown means selling into the exact moment liquidity is thinnest.

Second, the stock-vs-flow confusion is being institutionalized. Since the milestone crossed 1 million, then 1.5 million, then 1.959 million, the marginal news value of each threshold has decayed. The market has fully absorbed the accumulation regime. The information content of "holdings at a new high" is now roughly zero — it is a confirming print in a tape that already priced 70–80% of the institutionalization thesis. The only variables with pricing power left are weekly net flow direction, basis, and custody concentration. The first two are published daily. The third is not published at all.

Third — and this is the part that stings — the retail trader reading this headline is the exit liquidity for the allocator reading the same print. Here is the asymmetry, laid out:

| Dimension | Retail Reading This Headline | Institutional Allocator | |---|---|---| | What they see | Accumulation, bullish | Allocation confirmation, rebalance trigger | | Time horizon | Weeks to months | Quarters to years, with hard mandates | | Reaction to -25% | Buy more, average down | Trim to target weight | | Cost sensitivity | Low | High (fee rotation) | | Exit mechanism | Sell on exchange, hit the bid | Redeem shares, custodian liquidates OTC | | On-chain visibility of exit | Full | Partial to none |

The last row is the kill shot. The retail exit is visible and adverse. The institutional exit is invisible and precedes the move. When the redemption wave arrives, the on-chain chart will show holdings holding flat while the spot market absorbs concentrate selling and price breaks. The retail reader will be told the lock is intact while the lock is unwinding through a channel they cannot see.

I have watched this exact structure before. In 2020 I shorted CKP exposure through ETH collateral because the oracle manipulation surface was obvious to anyone who read the liquidation cascade logic rather than the yield table. The market was chasing 200% APY and ignoring the fact that the collateral was pricing itself. The ETF version of that error is subtler but the same shape: a structure that looks like entrenchment but is actually a margin call waiting for a trigger.

And the trigger is not a hack. It is an allocation committee meeting in a week where the S&P is down 8% and the 60/40 sleeve is bleeding.

Now the equally contrarian counterpoint, because intellectual honesty requires both: the bear case here is weak on fundamentals. Unlike 2022's algorithmic stablecoin edifice, there is no reflexive death spiral in a spot BTC trust. There is no yield promise that must be paid from new deposits. There is no liquidation engine to cascade, no oracle to manipulate, no admin key to abuse, no treasury to unlock. Bitcoin has no issuer, pays no yield, and cannot be rugged by governance vote. The 1.959 million coins are not a Ponzi liability — they are a claim on the most liquid hard asset in existence.

That distinction is why this overhang is a timing risk, not a solvency risk. Different trade, different sizing, different stop.

Takeaway

Strip the narrative and you are left with a single mechanical fact and a small set of forward-looking tells.

The fact: a demand channel absorbing 3–6x daily new issuance has accumulated 11.9 years' worth of that issuance in under two years, and now holds a stock comparable to the entire exchange float. Price is being set by holders who choose to sell, not by miners who must sell. That is a structurally higher clearing price regime, and it does not revert because a chart looks extended.

The tells — the only five things worth tracking:

| Signal | Where to Look | Trigger Level | What It Means | |---|---|---|---| | Weekly net flow | Issuer daily holdings disclosures | Sustained two-week net outflow | Lock narrative inverts | | CME front-month basis | CME futures curve | Annualized basis < cost of carry | Basis-trade unwind, spot selling | | GBTC bleed vs IBIT inflow | Issuer-level BTC balances | Net complex negative despite IBIT green | Composition rot | | Custody concentration | Fund custody disclosures | Rising single-custodian share | Systemic risk accumulation | | BTC–Nasdaq correlation | Rolling 30d/90d | Sustained >0.6 | Hedging property gone |

Actionable framing, not a prediction: the milestone itself is not tradeable. The cascade of the first sustained weekly net outflow is. Historically, when the flow regime flips, the on-chain holdings print lags the price move by days, not hours — which means the chart retail watches will still look bullish for the first 3–5 sessions of distribution. That lag is the spread. That is where I will be positioned, and I will be positioned before the print confirms it, because by the time the print confirms it, the spread is gone.

The float is tightening. The release valve is a settlement cycle nobody can see.

The question is not whether institutions own 9.75% of Bitcoin. It is whether you are modeling that ownership as a vault or as a door — and which of those two readings you are using to size your position at $113,000.

Confidence-tagged where derived. Verify the denominated supply basis, issuer-level balances, and custody attributions against primary disclosures before acting. The arithmetic in this piece uses a $113,017 implied unit price and 1,959,000 BTC, both taken directly from the source data point.