Anatomy of a tokenized fund: the stack and what it costs

Eight jobs, eight firms

The idea is simple: a fund is a bundle of jobs. Someone must own the assets, someone must choose them, someone must count their value, someone must keep them safe, someone must record who owns the fund, and someone must find buyers. Tokenization changes how two of these jobs are done and leaves the rest alone.

BUIDL shows it clearly, because BlackRock rented almost everything. When the fund launched on 20 March 2024, BlackRock contributed the portfolio and the brand. BNY Mellon took custody and administration, Securitize took the shareholder register and the token, Securitize Markets took distribution, and PwC took the audit. That rent-a-stack model is what most of the institutional wave copied afterwards.

The table below is the map for the rest of the chapter.

The off-chain half: owning, managing and counting

The first five layers existed long before blockchains, and they carry every legal promise a holder has.

A legal wrapper is the entity that legally owns the assets, so that the investor owns a piece of the entity rather than the bills directly. BUIDL's wrapper is a fund company. Ondo's USDY uses a special-purpose vehicle, a company created only to hold the collateral and issue notes against it, and Paxos holds the gold behind PAXG in a trust. Chapter 4 shows why the choice decides what happens in a bankruptcy.

The manager picks the assets and takes the largest share of the fee. In a Treasury fund the job is narrow, since short government bills leave little to decide.

The administrator keeps the fund's accounts and calculates its NAV, the net asset value: everything the fund owns minus what it owes, divided by the number of shares. For BUIDL, BNY Mellon does both the administration and the custody. The NAV is a number produced off the chain, once a day, by a bank. Getting it onto the chain without delay or error is the subject of chapter 5.

The custodian holds the actual Treasury bills and may not use them as its own. The price of that job has been driven down over decades. A traditional fund custody agreement charges around 1 basis point a year on the first billion dollars, and less above that.

The on-chain half: registering, tokenizing and selling

The last three layers are where tokenization acts, and they bring one new job and one new cost.

The transfer agent keeps the official list of who owns the fund. In BUIDL, Securitize, a transfer agent registered with the SEC, mints a token for every share an investor buys and burns it on redemption. Each token can move only between wallets on an allow-list, a list of addresses whose owners have passed identity and eligibility checks. Securitize keeps that list, so every transfer checks it first.

That check has a price you can measure. On 4 September 2026 I read the gas used by ordinary transfer calls on Ethereum. A BUIDL transfer used about four times the gas of a plain USDC transfer, because it has to consult the compliance contracts before moving.

The token also brings a second kind of custody. Token custody means holding the private keys to the wallet in which the tokens sit, a separate contract with a separate firm such as Anchorage, BitGo or Fireblocks. This job is young and priced accordingly: digital custody is quoted at 10 to 60 bps of the assets held, many times the traditional rate for the same bills.

Distribution closes the stack. Securitize Markets acted as BUIDL's placement agent. The largest single buyers have been DeFi protocols. In 2025 Sky, the successor of MakerDAO, put 500 mn USD into BUIDL as part of a competition to allocate its reserves to tokenized Treasuries. Chapter 8 follows that money.

What it costs to run

No RWA infrastructure provider publishes a price list, so the cost of a stack has to be modelled from public benchmarks. One such model, for a tokenized Treasury fund with 100 mn USD of assets on one chain, comes to about 435,000 USD a year once the fund is running, an estimate built on 2026 market rates.

The surprise is where that money goes. Digital custody, priced in the model at a negotiated 25 bps, accounts for about 250,000 USD, or 57% of the total. Lawyers, auditors and the blockchain itself are small lines next to it.

One layer does get cheaper. A classic transfer agent charges per account, historically around 12 USD per open account a year, so a fund with a hundred thousand small investors pays over a million dollars a year just for its shareholder list. A digital transfer agent charges roughly the same flat fee however many holders there are. For a fund with many small investors that saving is decisive. For BUIDL, which had 89 investors on Ethereum when I read its compliance contract on 4 September 2026, it barely matters.

Who takes the fee

A basis point (bps) is one hundredth of a percentage point, so a 50 bps fee is 0.50% of the assets a year. Fund fees come out of the fund's value daily, so the investor never sees a bill, only a slightly lower yield.

A fee can be split by layer in a model. For a 500 mn USD fund charging an all-in 100 bps, built on public market rates, the manager keeps about 35 bps, distribution takes about 15, and the platform and transfer agent take about 12. Administration, compliance, custody, audit and on-chain infrastructure share most of the rest. The issuer is left with about 5 bps.

The platform and transfer agent line is the one layer a traditional fund does not pay for in this form. It is the direct price of being on a chain. How BlackRock actually divides BUIDL's fee among Securitize, BNY Mellon, PwC and itself is not public.

The prices investors see today are also subsidized, from three pockets.

1. Blockchain foundations. Since 13 November 2024, BUIDL has charged 50 bps on Ethereum, Arbitrum and Optimism but 20 bps on Aptos, Avalanche and Polygon. The foundations behind those three chains pay BlackRock the difference to attract the fund.
2. Fee waivers. Issuers switch off their own fees to gather assets. Ondo's OUSG earns Ondo no management fee while its waiver runs, because all the yield flows into the fund's value.
3. Investors' capital. Securitize, now a public company, reported net losses of about 8 mn USD in the first quarter of 2026 and about 22 mn USD in the second, in its SEC filings of 12 August 2026.

None of the three is permanent, so today's fees are not a stable price.

Break-even: why there are no small tokenized funds

Break-even is the fund size at which fee income covers running costs. The idea is simple, but one detail matters. Some costs are fixed: legal work, audits, the transfer agent and gas cost about the same whether the fund holds 10 mn or 1 bn USD. Others are variable: custody is charged as a percentage of assets, so it grows with the fund.

Splitting the modelled budget this way gives roughly 175,000 USD of fixed cost plus about 26 bps of variable cost on every dollar, almost all of it custody. A fund charging 50 bps keeps 24 bps per dollar after those variable costs, and those 24 bps cover the fixed cost at about 73 mn USD of assets. A fund charging 15 bps never breaks even under these assumptions, because its fee is lower than its variable costs.

That arithmetic sorts the market. Tokenized Treasury funds charge between 15 and 50 bps, and a newcomer paying list prices for custody has a hard time surviving at the low end. The funds that thrive belong to firms whose custody and administration cost them far less, because they already run them at scale. BUIDL held about 2.8 bn USD on 4 September 2026, according to DefiLlama, which at up to 50 bps is up to about 14 mn USD a year of fees for a fund that runs on infrastructure its partners already own.

The scarce resource is not software either. RWA.xyz counted more than 200 tokenization platforms in September 2026. What is scarce is licences: Securitize operates through separate US entities registered as broker-dealer, trading venue operator, transfer agent and fund administrator, and those registrations take years to assemble.

What to watch next

• Custody pricing: digital custody is the largest cost in the stack. Published or sharply negotiated rates would lower break-even for everyone and open room for smaller issuers.
• Foundation subsidy deals: the payments that cut BUIDL's fee on some chains are commercial agreements. When they end or change, the fee on those chains should move.
• Securitize's quarterly results: as the one pure tokenization firm that reports publicly, its split between issuance revenue and servicing revenue shows whether the platform layer can make money.
• Incumbents taking layers in-house: a BlackRock fund filing from May 2026 names BNY Mellon, not a crypto-native firm, as transfer agent for an on-chain share class whose official register is the chain plus BNY's off-chain identity file. If large banks absorb the transfer agent layer, the independent platforms lose their core job.