When gold got its exchange traded fund in November 2004, the rules were simple. You want to create ETF shares? Bring gold bars to the vault. Real metal, in the building, or no deal. There was no shortcut. Every dollar of investor demand that flowed into the GLD trust required someone to physically source and deliver gold bullion to HSBC's vault in London. Supply and demand worked exactly as intended. Gold was trading around $450 when the ETF launched. Seven years later, it hit $1,921.
When Bitcoin got its ETF in January 2024, the SEC initially imposed an even stricter model: cash only. Authorized Participants, Wall Street market-makers like Jane Street and Cantor that serve as middlemen between the ETF and the underlying asset, could not deliver Bitcoin directly. Instead, the AP handed cash to the ETF issuer, and the custodian (Coinbase, in the case of IBIT) went and bought Bitcoin on the open market. The AP had no control over when or how that purchase happened. This was more forceful than gold's model at driving spot demand, because every creation forced a visible, real-time purchase of Bitcoin by the custodian, not by the AP. And it worked. Bitcoin went from roughly $46,000 to a peak above $120,000 in eighteen months.
Then in July 2025, the SEC decided to "improve" the system. It approved in-kind creation and redemption for Bitcoin ETFs, the same model that gold had always used. The stated goal was efficiency: lower costs, tighter spreads, better alignment with how other commodity ETFs operate. On paper, it made perfect sense. Same rules as gold. Should work the same way.
But the switch did something that looked technical and was in fact transformative: it moved the control of Bitcoin sourcing from the custodian to the Authorized Participant. Under cash, the custodian bought Bitcoin immediately on the open market. Under in-kind, the AP decides when to source, where to source, and through which channel. That transfer of control is the actual mechanism that changed everything.
The Escape Hatch That Gold Never Had
The difference comes down to one thing: Bitcoin has a deep, liquid futures market on the CME. Gold futures exist too, but they were never a practical substitute for delivering physical bars. When a gold Authorized Participant needs to create ETF shares, the path of least resistance is straightforward. Get gold bars, put them in the vault, receive your shares.
When a Bitcoin Authorized Participant needs to create shares, the path of least resistance is completely different. To be clear, the AP must still deliver real Bitcoin to the ETF trust eventually. The accounting has to balance. But the in-kind model gives the AP total control over when, where, and how it sources that Bitcoin. Under the old cash model, the custodian bought Bitcoin on the open market immediately. Under in-kind, the AP can short the ETF first, hedge its exposure using CME Bitcoin futures in the interim, then source actual Bitcoin quietly from an over the counter desk at a time and price of its choosing, and settle later. The real Bitcoin still enters the trust, but the visible spot buying pressure that would have moved the price is replaced by a quiet, delayed, private transaction. The futures market gives APs an escape hatch that lets them satisfy ETF demand on their own schedule without ever placing a buy order on the public spot market.
This is not illegal. It is a structural feature of how ETF market making works under Regulation SHO, which exempts Authorized Participants from the usual short selling constraints that apply to everyone else. The exemption was designed to keep ETFs functioning smoothly. It works perfectly well for stock ETFs. It works fine for gold. But when applied to an asset with a hard supply cap of 21 million units and a parallel derivatives market, it produces a perverse outcome: investor demand flows into the ETF, but the buying pressure never reaches the actual Bitcoin market.
Jeff Park, CIO of ProCap and a Bitwise advisor, draws the distinction: no Authorized Participant suppresses Bitcoin's price. The structure suppresses price discovery itself. The first would be manipulation. The second is worse — it is invisible.
The result is visible on any chart. During the cash only era, from January 2024 to July 2025, Bitcoin's price roughly tripled. Since the switch to in-kind, it has fallen from above $120,000 to roughly $65,000. The SEC applied gold's rules to an asset with a fundamentally different market structure, and the price action since tells the story.
Why Futures Don't Fix the Problem
In theory, futures prices must converge to spot at expiry. That is a hard rule. On the settlement date of a CME Bitcoin futures contract, the futures price equals the spot price. No exceptions.
But this convergence does not create spot buying pressure, and that distinction is the whole story.
CME Bitcoin futures are cash settled, not physically settled. When a contract expires, nobody delivers actual Bitcoin. The two parties simply settle the difference in cash. This means an AP can hedge its short ETF position with futures, let the futures expire, settle in cash, roll into a new contract, and repeat. The AP still owes real Bitcoin to the trust at some point, but the futures hedge lets it delay that delivery indefinitely while remaining economically neutral. The spot purchase that investors expect to happen when they buy the ETF can be deferred for weeks or months.
Another difference with gold. Many gold futures contracts allow for physical delivery. Hold a gold future to expiry and you can receive gold bars. That creates a direct mechanical link between the futures market and the physical market. Bitcoin futures on the CME have no such link. They are entirely paper settled.
So when the textbooks say "futures converge to spot," what actually happens is that the futures price adjusts to match the spot price at expiry. But the spot price itself is being set in a market where the biggest potential buyers, the Authorized Participants, are not participating because they are hedging in futures instead. It becomes circular. Spot is quiet because the big players are in futures. Futures converge to the quiet spot price. The mechanism that is supposed to align price with real demand instead reinforces the gap.
Where APs Actually Source Their Bitcoin
When an Authorized Participant does eventually need to deliver real Bitcoin for an in-kind ETF creation, they draw from several sources, roughly in order of preference.
First, their own inventory. Large trading firms like Jane Street hold Bitcoin on their books as part of their market making operations. They can deliver coins they already own without making any new market purchase.
Second, OTC desks. These are private brokerages that match large buyers and sellers outside of public exchanges. An AP can call an OTC desk and request thousands of Bitcoin, and the desk fills the order from its own reserves or sources it from miners, funds, or other institutional sellers. None of this appears on exchange order books. None of it moves the visible spot price. The OTC pool is the invisible buffer. When it empties, the buying becomes visible.
Third, other Authorized Participants. APs trade with each other. One might have excess inventory while another needs coins. These bilateral trades happen entirely in private.
Fourth, prime brokerage lending. APs can borrow Bitcoin from custodians or prime brokers who hold coins on behalf of clients, including, in some arrangements, the ETF trusts themselves. Borrowed coins still satisfy the accounting today, but they leave the lender exposed if the borrower cannot return them.
Fifth, and only as a last resort, public exchanges. Going to Coinbase or Binance and placing a large buy order is the option APs prefer to avoid because it moves the market, signals intent, and raises their costs.
The ordering reveals everything about why the current system suppresses price discovery. Authorized Participants will exhaust every private, invisible channel before they bid on a public exchange. They use inventory first, then OTC, then bilateral trades, then borrowing. Only when all of those sources run dry do they show up in the visible market where price is actually discovered.
Watch the OTC balances. They are the last quiet channel before AP demand reaches the open market. When the desks run dry, two options remain: borrow coins and inherit counterparty risk, or bid on a public exchange and move the tape.
Why None of This Changes the Long Term Outcome
The ETF plumbing does not delete demand. It delays it.
Every ETF share in existence must ultimately be backed by real Bitcoin sitting in a custodian's vault. Auditors verify this. The SEC verifies this. The accounting must balance. What the Authorized Participant grey window does is control when and how visibly that Bitcoin gets purchased. It does not change whether it gets purchased.
Think of it as a dam on a river. The water keeps flowing in. ETF inflows, corporate treasury buying from companies like Strategy (which now holds over 717,000 Bitcoin), self custody withdrawals from exchanges, mining supply that keeps shrinking after each halving. The dam can hold the water back and control the release. But it cannot make the water disappear. And the reservoir behind it keeps rising every single day.
Three forces are converging that will eventually overwhelm the dam.
The OTC pool is almost empty. Over the counter desks are the quiet supply channel that Authorized Participants depend on to source Bitcoin without moving the spot price. These desks held roughly 480,000 Bitcoin in September 2021. By mid 2025, that number had fallen to around 156,000 according to CryptoQuant's on-chain tracking. Strategy alone purchased over 200,000 Bitcoin in 2025, more than what the visible OTC pool held at the time. When an Authorized Participant tries to source Bitcoin for ETF settlement and the OTC desk says there is nothing available, the firm has exactly one option left: bid on the open market like everyone else.
Mining cannot keep pace. Following the April 2024 halving, only about 450 Bitcoin are mined per day. That is roughly 164,000 per year of new supply entering the system. Meanwhile, as of early 2026, ETFs collectively hold 1.27 million Bitcoin and Strategy holds another 717,000. Annual issuance covers less than a tenth of what these two pools already absorb. And they keep buying.
Futures must converge to spot. Every futures contract has an expiry date. On that date, the price of the future must equal the price of spot Bitcoin. If Authorized Participants have been hedging with futures while the spot market stayed artificially quiet, that divergence snaps back at settlement. The longer the suppression persists, the more violent the convergence.
Gold went from $450 to $1,921 over seven years with an ETF model that worked properly from day one, backed by a commodity with elastic supply where higher prices incentivize more mining. Bitcoin has a fixed supply cap, a halving schedule that keeps tightening issuance, and an ETF model that is currently broken but mathematically self correcting.
The plumbing is delaying the inevitable. The 21 million cap is real. The supply is draining. The demand keeps building. The only question has ever been timing.