The press release contained five facts. One of them was that there were no facts.
That is not a joke about PR copy. It is a structural observation about a specific artefact: the announcement by Raffles Family Office, an Asian multi-family office, that it is opening a wealth management service aimed at the region's "stablecoin billionaires." No assets under management figure. No named custodian. No licence number. No launch date. No executive quoted. No fee schedule. No client count. A sentence of positioning and a category label.
I have spent most of my working life reading records that are defined as much by their missing fields as by their present ones. In 2017 I traced ETH out of fourteen wallet clusters attached to an ICO team that had published a governance structure on a blog and implemented something materially different in the contract. The published material looked complete. The omissions were where the control actually lived. You learn to read the blank columns.
So this is not a report on a product. It is a report on a trace with very few bytes in it — and on what those bytes imply about where institutional stablecoin money is actually moving.
Family offices are, by construction, the least legible layer of finance. A single-family office manages one family's balance sheet and in most jurisdictions escapes licensing entirely. A multi-family office serves several unrelated families and, in the same jurisdictions, generally does not. Hong Kong's Securities and Futures Commission regulates Type 1 (dealing), Type 4 (advising on securities) and Type 9 (asset management) activities; an MFO running discretionary mandates for third-party clients needs Type 9. Singapore's MAS draws a similar line — the 13O and 13U tax exemption schemes are built for single-family vehicles, while a fund management company serving outside clients needs an RFMC, LFMC or CMS licence.
The gap between those two structures is where most of the industry's ambiguity lives. Say "family office" and the listener hears unlicensed, private, low-profile. Say "multi-family office" and the same listener often still hears it. Licence status is a factual field that gets drowned out by a label. On-chain equivalents exist: a contract's owner() function tells you more than a marketing page ever will. The same is true of a corporate registry number.
Then there is the client category. "Stablecoin billionaire" is a phrase built for headlines rather than for a spreadsheet. Real crypto fortunes are denominated in volatile assets — tokens with float, unlock schedules, market-depth constraints. Someone who has converted nine figures into USDT and left it sitting there is a specific kind of person: usually post-exit, usually with a tax event behind them, usually with a jurisdictional problem they have not solved. That is a small population. Estimates of genuinely nine-figure stablecoin holders in Asia run in the hundreds, perhaps low thousands, and many of them are entities rather than individuals — market makers, OTC desks, payment processors serving cross-border trade.
Asia is the right geography. Hong Kong and Singapore spent 2023 through 2025 building the licensing perimeter — the SFC's VASP regime, the stablecoin ordinance taking effect in 2025, MAS's stablecoin framework from 2023. Dubai sits alongside. Regulatory clarity is real, which means the grey space is closing, which means offshore holders now have a reason to migrate into a supervised wrapper.
The trouble is that this story arrived as a press release rather than a filing. Announcements and filings have different information properties, and I treat them differently. A filing has consequences for being wrong. A press release does not.
Strip the marketing and a family office offering stablecoin wealth management is a conversion machine. Assets arrive as USDT on Tron or USDC on Ethereum; they must leave as something a private bank, a tax authority and a next-generation beneficiary can all recognise. Every step in that chain is a third-party dependency, and none were disclosed.
Start with custody. A firm like this almost certainly holds no keys itself. The realistic configurations are MPC wallets (Fireblocks, Copper), a regulated trust company (Hex Trust, Anchorage, Sygnum), or exchange custody (HashKey, OSL). Each carries a different risk profile and the difference is not cosmetic. MPC means the institution controls shards and the client has, at best, contractual recourse. Trust custody means a segregated legal structure — bankruptcy-remote in principle. Exchange custody means you are a creditor. The announcement does not say which. For a client deciding whether to hand over forty million USDC, that is the only question that matters.
Second, aggregation. Stablecoin balances are not sitting on one chain. Tron carries the majority of USDT transfer volume; Ethereum carries most USDC and the deepest DeFi liquidity; Solana carries a growing share of payments flow; BSC carries retail and Asia-facing activity. A desk handling this must maintain treasury operations across four or more address spaces, rebalance between them, and reconcile gas costs in native tokens it may not want to hold. My 2020 work on Compound versus Aave taught me how fast the obvious answer changes when you follow actual address behaviour — I tracked more than 500 unique wallets over three months and found roughly 70% of the booked yield was arbitrage flow rather than durable deposits. The same measurement problem applies here. A treasury team moving $200 million between chains weekly generates a great deal of transaction activity and very little alpha. Yields don't lie, but volume does.
Third, the yield question, which is where the model either works or collapses. Stablecoins do not appreciate. A client holding $100 million in USDC and paying a 0.75% management fee is losing 0.75% a year in real terms unless something generates return. So the desk must convert idle balances into yielding instruments: tokenised Treasury bills (BENJI, USDY), money market funds, or lending venues. Each introduces counterparty exposure the client did not previously have. Tokenised T-bills are the cleanest option and still involve a fund administrator, a custodian bank, a blockchain issuer and a smart contract. Lending venues are not clean at all.
Fourth, the arithmetic. At 50 basis points on $1 billion, revenue is $5 million a year. That does not cover a licensed platform with multi-chain treasury operations, KYT screening, legal architecture across three jurisdictions and a relationship team. At 1% it covers roughly the operating cost and nothing else. To be a meaningful business the desk needs either $3–5 billion in stablecoin AUM or a much higher fee loaded into architecture work — one-off structuring fees booked in the first eighteen months and never repeated. The first is implausible on the client population described. The second is realistic but does not compound.
Which surfaces the concentration problem. This is not a retail product with a long tail. A handful of families represent most of the revenue. One client leaving is a visible event on the P&L. Traditional private banks absorb that because they hold thousands of relationships. A focused multi-family office does not.
And then the piece nobody puts in a press release: provenance. Stablecoin balances at this scale have a history. Some touched mixers. Some came through offshore exchanges with thin KYC. Some are the proceeds of OTC settlement networks that exist precisely because the banking system would not process them. A licensed desk taking these clients must prove source of funds — not to a wallet, to a person, to a bank account, to a tax filing. That is the hardest operational requirement in the entire business, and it was the largest blank in the announcement. KYT tooling (Chainalysis, Elliptic, TRM) produces risk scores, not exculpation. A score of 40 does not tell a regulator who wired the original fiat.
There is also the segregation and insurance question, which nobody asks until it matters. In March 2023, USDC traded to $0.87 on a Saturday because a fraction of its reserves sat at a failed bank. It recovered within days. But for those hours, every desk holding client USDC was making decisions about disclosure in real time. A wealth management business holding client stablecoins needs a written answer to that scenario: who eats the 13% drawdown, how fast, and under what contractual clause. I would want to see that document before I believed the service exists beyond a slide.
My 2024 work comparing BlackRock's IBIT inflows against Coinbase institutional vault deposits produced a 0.85 correlation between ETF creations and Layer 2 fee activity. Plenty of people read that as causation. It was not. It was two series responding to the same underlying allocator behaviour. The same discipline applies here. A family office launching a stablecoin desk and stablecoin supply expanding are correlated through a third variable — institutions deciding that on-chain settlement is now survivable — and neither causes the other.
Now the part that will annoy the people who forwarded the article.
An announcement is not an adoption signal. It is an intent signal, and the two have wildly different statistical properties. I spent two weeks in 2022 mapping UST into Curve pools and calculated that 12 million LUSD were burned in the final 48 hours of the de-peg. The mechanism did not fail because people announced it would; it failed because the redemption curve was structurally concave. Announcements told you nothing. Pool balances told you everything. Chaos is just data waiting for the right query.
Applied here: what we have is one firm, in one region, saying it will serve a client category whose size is asserted rather than measured. The correct inference is not that institutional stablecoin adoption is accelerating. The correct inference is that one licensed intermediary believes there is enough demand to justify a business line. Different claims, different confidence intervals.
Second correction: the competitor that matters is not the private bank. It is the VIP desk at the exchange where the client already keeps their assets. Those desks hold the deposit history, the trading relationship, the KYC file and the wallet whitelist. Extending from that position into wealth management is trivially easy. An independent family office must win the relationship from scratch and differentiate on exactly two things — cross-venue neutrality and multi-jurisdictional tax structuring. The first is a real advantage, because an exchange desk has an obvious conflict when advising a client to move assets elsewhere. The second is a real moat, because it requires lawyers in three cities, not software.
Third: the structural tension nobody prices. The client's assets are permissionless and portable. The service is hyper-permissioned and sticky. Every regulatory tightening on one side makes the other less attractive, because the entire reason a stablecoin holder wants a licensed wrapper is to escape the friction of being unlicensed — and the wrapper reintroduces exactly that friction, with reporting obligations attached. This is why these businesses scale slowly. Not because of technology. Because their value proposition is a compromise the client is permanently reconsidering.
Four fields will resolve this story, and all four are public. A licence number with a scope attached — Type 9 versus an advisory-only arrangement, or the MAS equivalent. A named custodian, because the custody model is the risk model. An AUM figure within two reporting cycles, since family offices can disclose aggregate numbers when they choose to. And a senior compliance hire, which is the most honest leading indicator available: firms build the AML function before they scale the book, or they never scale it.
If those appear, the announcement was a genuine capability statement. If a year passes and the desk is still described only by its target client, treat the filing as marketing and move on.
The question worth sitting with is not whether Asia's stablecoin wealthy want a compliant home. They do. The question is whether that home can be built at a price they will pay, by an institution small enough to need them. Trust the hash, not the headline. And when there is no hash — read the blank columns.
