Update (23 September 2026): Tempo has been live on mainnet since 18 March 2026, and Stripe’s acquisition of OpenRouter — announced 19 August 2026 — has moved from rumour to signed deal. This piece stress-tests Emad Mostaque’s claim that AI agents will never use “crypto rails”, and will instead settle on Tempo.
Every few years someone announces that the real winner of crypto is not crypto. This time it’s Emad Mostaque, interviewed on the Risk Takers podcast in September 2026, arguing that AI agents — the autonomous software that will soon buy tokens, APIs and compute on our behalf — will not transact on crypto rails at all. They will use Tempo, which he describes as Stripe’s payment rail: fee-free, with OpenAI and Anthropic in the coalition behind it.
His reasoning is behavioural, not ideological. Agents have no loyalty, no attachment to “their bags”, and therefore no endowment effect — the human bias that makes us overvalue what we already own. They will simply go wherever money is easiest, safest, and least likely to be lost to a mistake.
Here is the line that has travelled furthest: “why would an agent use anything except for Tempo?”
It is a provocative claim, and worth taking seriously — RWA investors have a lot riding on whether the next wave of machine money settles on open chains or inside a Stripe walled garden. So let’s take the thesis apart.
What Mostaque Actually Said
From the interview transcript, his argument runs like this:
“I think crypto rails make coordination a lot easier, but really, why would an agent use anything except for Tempo? Like Tempo is directly integrated into Stripe. OpenAI and Anthropic are part of the coalition there and it is basically fee-free.”
He frames the deciding factor as institutional gravity rather than technology:
“something with the institutional heft of a Stripe paradigm where OpenAI and others are with basically zero fees would be the most sensible place to transact.”
And on agent motivation:
“How do you attract an agent when they do not have a utility function? They will just go for wherever the bulk of money is that is easy to do, where they are sure they will not mess up.”
On loyalty:
“the agents have no loyalty… they are loyal to their users and then soon to themselves.”
On stablecoins specifically, he expects the money itself to become wrapped for machine use, with monetary velocity going “off the charts”.
The quotes are punchy. But the brief for any investor is the same: verify the claims, don’t repeat them. So let’s.
What Tempo Actually Is
Mostaque’s description is broadly accurate — with one detail he glosses over.
Tempo is a Layer 1 blockchain built specifically for stablecoin payments, incubated by Stripe and the crypto investment firm Paradigm (whose co-founder Matt Huang leads the project). It launched a public testnet in December 2025 and went live on mainnet on 18 March 2026.
Its design choices are all aimed at payments rather than speculation:
- No native gas token. Transaction fees are paid in USD stablecoins, so an agent never needs to hold a volatile token to move money.
- Near-instant settlement with no re-orgs — finality you can actually rely on, which matters enormously for machine-speed payments.
- Sub-cent fees. Tempo targets roughly one-tenth of a cent per transaction.
- EVM-compatible, so existing Ethereum tooling and applications port across.
The “coalition” claim also checks out, at least as design input: Visa, Deutsche Bank, Shopify, Revolut and OpenAI are among the firms that fed into Tempo’s design, alongside Paradigm’s portfolio. So OpenAI and Anthropic being “part of the coalition” is fair — with the caveat that “design partner” is not the same as owner, shareholder, or exclusive integration. That distinction matters more than it sounds, and we’ll come back to it.
The “fee-free” claim needs the same precision. Tempo’s settlement layer is effectively free — fractions of a cent, and no gas token to buy. But the wider Stripe stack an agent would actually traverse and the fees around a payment — merchant processing, currency conversion, treasury management — do not vanish. “Fee-free” describes the rail, not the whole journey.
The Catch: Tempo Is Itself a Crypto Rail
Here is the sharpest problem with the thesis, and it isn’t a quibble.
Tempo is a blockchain. It settles in stablecoins. It is, in the plainest sense, a crypto rail.
What Mostaque is really arguing is narrower and more defensible than the headline: agents won’t use open, permissionless, volatile-gas crypto rails — the ones where a wallet can be drained, a transaction can be front-run, or a chain can halt. They will use a permissioned, corporate-governed crypto rail where the complexity is hidden and the counterparty is Stripe.
That is a real distinction, and arguably a correct one. But it is a distinction about who controls the rail, not about whether the rail is crypto. For anyone holding RWA exposure, that reframing is the whole ballgame: the question is no longer “crypto or not-crypto” but “which chains capture the agentic economy — and who owns them”.
Stripe’s Real Play: Own the Agent’s Money Path
Look at what Stripe has actually assembled, and the strategy is unmistakable.
- Agentic Commerce Protocol (ACP) — an open standard co-developed with OpenAI that lets agents negotiate checkout and share payment credentials, powering Instant Checkout inside ChatGPT.
- Link agent wallet — lets agents pay on a user’s behalf with controlled spending approvals and one-time-use cards, without exposing real payment details.
- Machine Payments Protocol (MPP) — co-authored by Stripe and Tempo, an open standard for agent microtransactions and recurring payments, accepting stablecoins as well as fiat.
- OpenRouter — agreed to be acquired in August 2026 for a reported $7bn+ (some reports higher). OpenRouter routes traffic across 400+ AI models from 80+ providers. As Stripe CEO Patrick Collison put it, “tokens are the central currency for companies building with AI”.
Read together, Stripe is trying to own three layers at once: the checkout (ACP), the settlement (Tempo), and the compute spend (OpenRouter). That is not a payment company hedging; it is a company building the pipes through which machine money will flow. Mostaque’s instincts about Stripe’s trajectory look sound.
Where Agent Money Is Actually Settling
Here the evidence is more ambiguous than either camp admits.
The crypto side is real but small — and possibly overstated. x402, Coinbase’s open protocol that revives the HTTP “402 Payment Required” code for stablecoin payments, processed roughly 165 million agent transactions and about $50m of volume by April 2026, with an average payment of around $0.31. By August 2026 it was running ~29 million transfers a month at about 6 cents each. Impressive growth — but one widely cited analysis suggests only 0.6% to 7.5% of x402 payment value is genuinely agentic rather than human-driven or scripted. That is the weakest link in the crypto bull case, and it deserves to be stated plainly.
The stablecoin side is enormous but mostly not agentic yet. Stablecoin settlement reached $7.2 trillion in February 2026, surpassing US ACH for the first time, and around $7.5tn in March — but the vast majority of that is exchange and treasury flow, not machine commerce. Visa’s own stablecoin settlement programme hit a $7bn annualised run rate across nine blockchains by April 2026, and card spending linked to stablecoins was around an $18bn annualised clip.
And the incumbent is not standing still. Mastercard launched Agent Pay for Machines in June 2026, handling machine micropayments with both cards and stablecoins; Visa is building on x402 while integrating stablecoin settlement. The likeliest outcome is not “crypto wins” or “Stripe wins” but a hybrid: agents defaulting to whichever rail is cheapest and safest for the job, with stablecoins doing the sub-cent settlement and cards holding the consumer-dispute layer.
What Tempo Cannot Replicate
The article would be a puff piece if it stopped here, because Tempo’s design advantages come bundled with properties crypto rails have and a corporate chain structurally cannot:
- Permissionless access. Anyone, anywhere, can build and transact on a public chain without approval. Tempo is governed by Stripe and Paradigm; access is ultimately a commercial decision.
- No single point of control — or de-platforming. A corporate rail can, in principle, freeze funds, block a merchant, or exit a jurisdiction. That is a feature for compliance and a risk for anyone whose business model Stripe dislikes. Public chains have no CEO who can switch you off.
- Composability and programmability. Open rails let anyone build on top — lending, collateral, derivatives, yield — without permission. A payments-first walled garden deliberately avoids that complexity; that is what makes it safe, and also what makes it limited.
- 24/7 settlement with no gatekeeper. Truly neutral settlement means no counterparty deciding whether your transaction clears. Tempo promises settlement reliability, but reliability delivered by an operator is a different promise from settlement that no operator can revoke.
None of that makes Tempo the wrong answer — for a mainstream agent buying a coffee or a token bundle, a governed rail is better. It just means the thesis is really an argument about convenience beating neutrality, and those two properties don’t always trade in the same direction for investors.
The Regulatory Wildcard: Stablecoin Law and Tokenised Assets
Mostaque’s “wrapped money” prediction lands right in the middle of the biggest policy shift in digital finance.
The GENIUS Act, signed in July 2025, created the first US federal framework for “payment stablecoins”. Its implementation has run through 2026: federal agencies faced a July 2026 statutory deadline to finalise frameworks, the OCC expects to finish its implementing rules by November 2026, and the prohibition on issuing payment stablecoins without a licence takes effect January 2027. Reserves must be 100% backed by cash or short-term Treasuries, and issuers cannot pay holders yield.
Two things follow. First, legitimised stablecoins are the raw material for exactly the agentic economy Mostaque describes — a fee-free corporate rail is only credible because the dollars moving on it are now legally ring-fenced. Second, the same legislation is turbocharging tokenisation: tokenised real-world assets reached roughly $38.9bn in distributed on-chain value by late September 2026, with tokenised US Treasuries and money market funds alone at $15–18bn and BlackRock’s BUIDL in the $2.3–2.8bn range. If agents transact in wrapped dollars, they are transacting in a tokenised asset — every time.
What to Watch
- Whether agent volume migrates from x402 to Tempo/MPP. The first real scoreboard: does Tempo’s Mainnet carry meaningful agent traffic by mid-2027, or does x402 keep the developer mindshare?
- How “design partner” converts into exclusivity. If OpenAI or Anthropic ever route payments only through Tempo, Mostaque’s thesis hardens; if they stay multi-rail, it softens.
- The OCC’s November 2026 rules and the January 2027 licensing cliff. These set which stablecoins can legally be used in US agentic commerce — and therefore which rails are viable.
- The counter-argument in practice. Watch whether any high-profile de-platforming or freeze on a corporate rail sends developers back to permissionless chains. That single event would do more for the crypto side than any white paper.
- Tokenised treasuries as agent collateral. If agents start holding tokenised T-bills as their working capital, the RWA thesis and the agentic-payments thesis become the same trade.
The Bottom Line
Mostaque is probably right about which rail wins the default, and wrong about the framing. Agents will gravitate to the cheapest, safest, most reliable option — and right now that is a Stripe-governed stablecoin chain, not an open one. But that chain is still a blockchain settling tokenised dollars, which means the agentic economy doesn’t bypass crypto: it quietly absorbs one particular, permissioned version of it, owned by a payments incumbent and its partners.
For RWA investors the takeaway isn’t “agents don’t use crypto rails”. It’s that the fight over the agentic economy is a fight over who owns the rail, who issues the dollar on it, and who is allowed to build on top — and the tokenised assets underneath are the same ones already on your watchlist.
This article draws on the September 2026 Risk Takers podcast interview with Emad Mostaque, Tempo and Paradigm public materials, Stripe newsroom and Sessions 2026 announcements, rwa.xyz data as of 23 September 2026, x402/Chainalysis and PYMNTS analysis, and US Treasury/OCC GENIUS Act implementation guidance. Quotes are attributed to the interview and its claims independently verified where possible. Not financial advice.