Arc’s mainnet launch tests settlement, not scale

Circle has set Sept. 16 for Arc’s public mainnet launch, and half a billion test transactions have already settled the capacity question. The unresolved part is whether eleven founding institutions controlling block production can deliver settlement finality that does not depend on central-bank money or legacy payment rails.

Arc is designed as a bank-run chain for tokenized assets and stablecoins, with those eleven founding institutions still controlling block production when the mainnet opens. The governance is the product, a permissioned chain run by regulated banks meant to give institutions a shared settlement layer without the permissionless risk of public blockchains. Sept. 16 tests whether that structure can produce something no permissioned rail has yet produced.

Circle’s half-billion test transactions are evidence the network can handle volume, but volume was never the binding constraint. A chain that processes that many test runs has proven it can count; it has not proven that a transfer of tokenized deposits between two banks is final, irreversible, and acceptable to regulators in the way central-bank money is final. That test cannot be simulated in a sandbox.

A group of eleven founding institutions is small enough to coordinate and small enough to be hard to trust as a neutral settlement layer. A bank will accept finality from that group only if the group’s own regulatory standing substitutes for a central bank’s; Sept. 16 tests that hypothesis. A regulator may be comfortable with eleven banks running a chain without a bank being comfortable marking an interbank transfer as done because those eleven banks said so.

The closest live example came when HSBC and Standard Chartered completed a cross-bank tokenized deposit transfer over Swift’s blockchain network: the tokens moved, but the final settlement still ran on existing systems. When the ledger entries were done, the money moved the old way. That describes the architecture rather than criticizing the test; a tokenized deposit transfer that settles on legacy rails is a mirage of finality, with the innovation confined to the message.

JPMorgan’s addition of euro payments to JPM Coin, with Siemens as the first client, approaches the same gap from the other direction. The euro leg sounds like a tokenized payment rail, but JPM Coin moves bank liabilities alongside central-bank money, and the new euro payments run in parallel with it. The token is a representation of the bank’s promise to pay.

The Bank for International Settlements’ Agorá project selected 41 institutions from seven central banks for live tokenized payments testing, which sharpens the distinction. Agorá is testing whether tokenized payments can be settled in central-bank money, with seven central banks among the participants; Arc runs bank-controlled block production without a central bank at the center. The two projects are asking different questions: Agorá is asking whether central-bank money can be tokenized, while Arc is asking whether tokenized money needs central-bank finality at all.

The regulatory clock arrives one day before the technical one, with the CLARITY Act’s Sept. 15 cloture vote determining the statutory framework for stablecoins and tokenized rails. Circle’s launch follows a day later. If the Senate advances the bill, Arc opens with a clearer path to legal recognition for its stablecoin and settlement layer; if the bill stalls, Arc opens anyway, into a less defined regulatory field. The law can define what a tokenized deposit is, but it cannot make a transaction final.

Finality remains the weak point on every permissioned rail. The HSBC-Standard Chartered transfer netted to legacy systems, and JPM Coin’s euro leg runs alongside central-bank money. Arc’s test transactions measure how many times the system can do something without measuring whether a bank will treat that something as done. Finality is not a technical luxury: if a bank cannot mark a tokenized transfer complete and irrevocable at the moment of consensus, it must keep the legacy entry open, hold capital against settlement risk, and reconcile two ledgers. That dual-ledger cost is why every permissioned rail so far has been a demonstration without becoming a replacement.

Sept. 16 begins to answer the finality question, though throughput will not answer it. The first real transfer between two of Arc’s eleven validators will matter more than any test count. If that transfer does not touch central-bank money and both banks treat it as final, Arc will have done what no permissioned rail has done. If it still reconciles to a legacy rail at end of day, Circle has built a faster messaging layer without building a settlement layer. The distinction between moving information and moving money is the entire ballgame for tokenized institutional settlement.