
Coinbase publishes a custody fee of 50 basis points annualized, with a $500,000 minimum balance and an implementation charge between zero and $10,000. Traditional securities custody safekeeping runs in the low single digits of basis points and is frequently bundled toward zero. An 80 percent compression takes 50 basis points to 10, which is still several times the conventional benchmark, so the headline number in this thesis is less aggressive than it sounds. The mechanism matters more than the magnitude, because the mechanism determines what an allocator ends up holding.
The thesis is usually told as a scale story: volume arrives, providers compete, fees fall. That reads the structure backwards.
Commodity markets compress margins because supply is substitutable. Crypto custody has the opposite configuration. Supply is concentrated, with one provider holding roughly 84 percent of the US spot Bitcoin ETF complex, and it is not substitutable, because key material generated under one provider's scheme does not transfer to another. Concentration plus switching costs is the textbook description of pricing power.
So scale is the reason a provider can hold 50 basis points, not the reason the number falls. Any version of this thesis resting on competitive pressure alone has to explain why a dominant provider facing clients who cannot leave would cut price.
The traditional market already ran this experiment, and the answer is not competition. Safekeeping became a loss leader because a second revenue pool grew large enough that custodians preferred to win the mandate at near zero and earn on the adjacency: securities lending, collateral reinvestment, cash spreads and fund administration.
Three conditions had to hold, and they are the right test here. A second revenue pool at genuine scale. Enough substitutability that a provider fears losing the mandate. And an adjacency large enough to fund the cross-subsidy.
In crypto the first condition is met and growing, through staking, financing, conversion and administration. The third is forming. The second is absent, which is the structural constraint of this category. That combination produces a specific forecast: if the fee compresses, it compresses because the provider found somewhere better to make money rather than because a competitor forced it. And in this asset class that somewhere is your assets doing something.
When the explicit fee falls and the revenue migrates, the all-in cost of custody stops appearing as an expense line.
A pseudonymised case published by one investigative outlet illustrates the end state precisely, and the figures are that publisher's own reporting rather than regulator findings. A custody bank running securities lending for three pension plans earned in three layers: a disclosed lending agent fee of roughly 30 to 40 percent of gross lending income, which appeared in trustee reports; management fees of roughly 15 to 25 basis points on affiliate short-duration funds holding borrower collateral, deducted before the net split with the plan; and spreads on uninvested cash swept into a vehicle the bank administered.
The allocation is the part worth copying into a diligence file. The plans received roughly 38 to 47 percent of gross economic value and the bank and its affiliates captured roughly 53 to 62 percent, with affiliate vehicles holding between 68 and 84 percent of total collateral. Most importantly, the disclosed fee was approximately half of the bank's total take. The other half sat in affiliate fund fees and cash spreads that never presented as a fee.
That is what a commoditized custody fee looks like after commoditization. The number on the invoice was real and it was not the cost.
Which turns the thesis into an evidentiary problem. Four economics have to be reconstructed because no document reports them voluntarily.
The staking spread: gross protocol reward earned on the assets against what is credited, stated as a percentage rather than a net yield. The conversion spread on fiat movements and inter-asset trades, measured against a reference rate at execution time. Net interest retained on fiat balances held between instructions. And any affiliate vehicle fee deducted before a net figure is calculated, which is the layer most likely to be invisible because it is netted upstream of the statement.
The deliverable is a fee waterfall: gross economic value generated from the assets, every deduction in sequence with the entity taking it named, and the net credited. In the pseudonymised case that disclosure arrived as remediation, alongside a requirement for written client authorisation before affiliate collateral investments, which tells you the sequence. The waterfall is produced after enforcement unless it is written into the contract beforehand.
For an allocator the practical consequence is that a custody fee reduction should trigger a diligence cycle rather than conclude one.
The thesis fails outright if key-share portability never arrives and no adjacency reaches scale, because 50 basis points then holds on pricing power alone. It fails in a subtler and more likely way if the fee compresses 80 percent while all-in cost stays flat, which is not a failure of the forecast but a failure of the metric, and it is detectable only with a waterfall.
Three monthly markers, all public or obtainable. Published fee schedules against the low-single-digit-basis-point traditional benchmark, which is a real series now that one provider posts 50 basis points openly. Whether staking or financing participation becomes a condition of the headline custody rate, since bundling is the visible fingerprint of cross-subsidy. And disclosure quality on the three spreads, which is the only one of the three that requires asking.
The constructive note is that published pricing is itself progress. A public rate is a comparable, and comparables are the precondition for the competition that switching costs currently suppress. The honest statement of this thesis is narrower than its title: the explicit fee very likely compresses on something like that order, the compression is funded by revenue moving to activities performed with the assets, and an allocator who banks the fee saving without pricing the adjacency has not saved anything.
For informational purposes only. Not an offer to buy or sell any security. Available only to accredited investors who meet regulatory requirements.