Circle's Arc vs Tempo: two stablecoin chains, two theories of payments

Tempo has been settling transactions since 18 March 2026. Circle's Arc opens its mainnet on 16 September. Any comparison written this week is therefore a comparison between a chain with six months of production traffic and a chain with a launch date, and there is no way to write around that asymmetry honestly. What can be compared now is the design each has committed to, and the two are genuinely different bets about what a stablecoin chain is for.

As of 10 September 2026, the independent directory Built on Arc (builtonarc.app) tracks [N] projects associated with Circle's Arc blockchain and records [LIVE COUNT] as live on mainnet.

Two theories

Tempo, backed by Stripe and Paradigm, is a distribution bet. It launched on the strength of Stripe's existing merchant and platform relationships, raised a $500 million Series A at a $5 billion valuation, and went to mainnet in March with a machine-payments protocol as its distinguishing feature — payments initiated by software agents rather than people, with the fee and identity plumbing designed for that case from the start. The theory is that the chain that gets real transaction volume first learns fastest and becomes the default before anyone else ships.

Arc is an issuer bet. Circle is the issuer of USDC, and Arc's design puts that stablecoin at the base layer: fees are paid in USDC rather than a volatile native asset, the execution layer is EVM so the existing Ethereum toolchain works unmodified, and consensus runs a permissioned validator set of named institutions rather than an open one. The theory is that enterprise settlement flows follow the issuer and the counterparties, and that a chain operated by the company that mints the dollar token is where regulated institutions will be willing to settle.

Both are coherent. They are not the same argument, and they are aimed at different first customers — Tempo at merchants, platforms and agentic commerce, Arc at banks, card networks, custodians and market infrastructure.

Fees, and what each chain assumes about stablecoins

The fee model is where the philosophical difference becomes a technical one.

Tempo defines its own token standard, TIP-20, which builds fee payment, transfer memos and compliance policy into the token itself, and allows a transaction to be paid for in whatever TIP-20 stablecoin the sender is holding, converted automatically at the protocol level. It is stablecoin-agnostic by construction.

Arc's gas is USDC by default, with a paymaster layer available to extend fee payment to other assets. It is USDC-first, with flexibility as an addition rather than a premise.

That difference is not cosmetic. A payments business that holds several stablecoins across corridors will read Tempo's model as the one that matches its treasury. A business whose entire exposure is USDC, and whose compliance posture depends on a single fully-reserved issuer, will read Arc's as the simpler one. Neither answer generalises.

Performance claims, marked as claims

Tempo publishes its own benchmarks: block finality averaging 508 milliseconds, throughput described as 21,200 settled transfers per second, average fees below a tenth of a cent, and a 99.999 per cent availability target. These are vendor figures under conditions the vendor chose. They have not been independently audited, and this piece treats them as claims worth tracking rather than settled facts.

Circle's published Arc figures are a target rather than a measurement: roughly 780 milliseconds to finality via Malachite, a Tendermint-derived BFT engine, across an initial network Circle describes as around 100 validator nodes. No mainnet throughput or per-transaction fee benchmark had been published as of 10 September 2026, and none can exist until the chain is carrying load.

The fair statement is that Tempo has numbers from production and Arc has numbers from a specification. That is a real advantage for Tempo today. Whether it survives contact with Arc's first month of traffic is exactly what the next four weeks will show.

The story is the hedging

The most useful thing in either chain's partner list is the overlap.

Mastercard is one of the eleven institutions Circle named as founding validators of Arc on 5 August 2026, and was a design partner on Tempo during its testnet phase. In March 2026 Mastercard also acquired the stablecoin infrastructure firm BVNK, reported at $1.8 billion, which deepens its position in the category rather than in either chain specifically. MoneyGram is an Arc founding validator and has been reported as Tempo's anchor remittance validator. Visa appears in Arc's validator cohort and among the organisations Tempo lists on its own site.

None of these companies has publicly explained the overlap, and neither Circle nor Stripe has commented on it. The reasonable reading is not that anyone is being duplicitous; it is that the payments incumbents do not believe the category has a winner yet, and are buying optionality on both sides at a cost that is trivial relative to their balance sheets. For anyone choosing a chain, that is a more honest signal than either launch narrative: the institutions with the most information are declining to choose.

The structural difference between the two validator sets is worth separating from the names in them. Arc's is permissioned by design and institutional in composition, weighted toward market infrastructure — a depository, two card networks, an exchange operator, an asset manager. Tempo's reported corporate validator group is smaller and weighted toward its own backers and early commercial partners. Concentration risk exists in both; they are concentrated around different kinds of institution.

What is actually live on each, dated

As of 10 September 2026, Tempo has been on mainnet for just under six months, following a public testnet that opened in December 2025. Arc has been on public testnet since 28 October 2025 and has no public mainnet. Circle has said more than 100 builders are active on a private mainnet ahead of the public date, which is a real signal about commitment but not a claim any third party can verify on an explorer.

That means the day-one question for Arc is not whether the chain works — it is how much of the announced ecosystem actually ships on the day. From 16 September the checkable version of that question is what is live on the Arc blockchain, counted against evidence rather than against press releases, and this section will be rewritten with dated figures for both chains once mainnet opens.

Which theory wins

Nobody can answer that from here, and a comparison that pretends otherwise is selling something. What can be said is what each chain has to prove. Tempo has to show that Stripe-adjacent volume becomes settlement depth that other builders can use, rather than Stripe's own flows running on their own rails. Arc has to show that an institutional validator cohort and issuer proximity convert into transactions, not just into logos on a press release — and its first test is whether the names Circle announced in August are visibly doing anything on 16 September.