Stablecoins are the practical, legally codified realization of the 1930s full-reserve-money proposal that banks successfully killed at the time.
Allaire says he was drawn to sound-money theory studying the aftermath of the 2008 financial crisis and the 1930s Chicago Plan (Irving Fisher's 100% Money), which proposed money fully backed by safe assets rather than fractionally re-lent. Banks lobbied it down and got FDIC insurance instead, leaving leverage risk intact (he cites 12-30x leverage in 2008). He argues the Genius Act now codifies the Chicago Plan's full-reserve model into law via dollar stablecoins like USDC, which must hold only short-duration Treasuries, repos, and overnight bank cash.
stablecoins-as-full-reserve-money
USDC's actual reserve composition is short-duration Treasuries, repos, and overnight bank cash, not a broad basket of assets.
Allaire specifies the average maturity of the T-bill portfolio is around 13 days, kept deliberately ultra-liquid so USDC functions like a cash instrument. Reserves are held at custodial institutions such as Bank of New York Mellon, with daily transparency reporting run through a system set up with BlackRock.
stablecoins-as-full-reserve-money
USDC already spans transaction sizes from sub-dollar AI-agent payments to multi-hundred-million-dollar institutional settlements on one uniform rail.
Allaire frames USDC as payload-agnostic the way email is: a 25-cent in-game purchase and a large electronic-trading settlement move through the identical infrastructure. He cites AI agents already paying fractions of a dollar for another agent's output as a live use case, alongside large firms doing capital-markets-scale transfers.
agentic-payments-infrastructure
The agentic economy requires new financial infrastructure because no existing rail supports instant, global, programmable, machine-initiated transactions at the necessary scale and cost.
Allaire argues AI agents will increasingly conduct work, transact with each other, and purchase specialized intelligence directly, requiring an infrastructure where agents can dynamically spin up financial endpoints, handle micro-transactions (cents), and scale into billions or trillions of transactions in real time - none of which legacy banking rails or even prior-generation blockchains could support economically.
agentic-payments-infrastructure
Circle designed ARC as a settlement layer purpose-built for regulated, real-economy financial institutions rather than the censorship-resistant, decentralize-everything ethos of earlier blockchains.
ARC uses a known, vetted validator set of major financial infrastructure companies held to high infosec/compliance standards, deterministic settlement finality in hundreds of milliseconds with no reorgs or hard forks, and USDC itself as the native gas token so costs are denominated in real dollars, not a volatile crypto asset. Allaire contrasts this with the 'alternative universe outside government reach' framing of the early blockchain era.
blockchains-as-operating-systems
Blockchains function as operating systems whose distinguishing properties - tamper-resistance, full auditability, and provable compute/state integrity - become newly critical once autonomous AI agents, not just humans, are transacting.
Allaire's original 2013 thesis was that blockchains would become operating systems analogous to mobile OSes, the web, or cloud platforms, but with unique guarantees: published code is tamper-resistant, every input/output is publicly auditable in real time, and transaction/compute integrity is provable. He argues these guarantees, originally valuable for financial fiduciary requirements, become essential once autonomous agents (not humans) are the ones initiating and executing transactions.
blockchains-as-operating-systems
Zero-knowledge proofs and off-chain compute verification are becoming production-critical, not just researchy, because billions of swarming AI agents will demand transaction scale current chains cannot handle natively.
Allaire describes ZK roll-ups and trusted execution environments as ways to move compute off-chain while still proving its correctness on-chain, calling this essential once agent populations reach the scale he anticipates. He also notes ARC ships with built-in privacy primitives from day one, since open/permissionless infrastructure still needs to let corporations avoid full public exposure of their activity.
agentic-payments-infrastructure
Real-world asset tokenization is moving through every layer of the securities stack, from share registries up through clearing (DTCC) to exchanges (NASDAQ, NYSE), with recent SEC guidance clarifying obligations at each layer.
Allaire says tokenization efforts now span record-keepers (transfer agents), depository/clearing infrastructure like DTCC, and the exchanges and brokers that will support trading tokenized securities. He notes the SEC issued guidance roughly a month before the taping laying out obligations across these layers, and that Circle's own stock is currently the most actively traded tokenized equity, ahead of Tesla or an S&P index product.
tokenization-of-real-world-assets
Tokenized securities today are adopted mainly to give non-U.S. investors access, inverting the old pattern where non-U.S. investors accessed foreign (e.g., Chinese) stocks via third-party wrapper instruments.
Allaire says most current tokenized-stock growth comes from people outside the U.S. who otherwise lack direct access to U.S. equities, and draws the parallel to how U.S. investors previously accessed Chinese listings through ETF-style wrapper products; he expects the more interesting long-run value to come from new utility (fractionalization, borrowing/lending against tokenized assets, novel packaging) rather than simply porting existing products onto new rails.
tokenization-of-real-world-assets
Emerging research on tying proof-of-work security to actual AI inference compute could turn Bitcoin-style mining's wasted energy expenditure into productive output.
Allaire highlights recent papers proposing that GPU work used to secure a proof-of-work cryptocurrency could simultaneously be the same GPU work used for AI inference, unlike Bitcoin mining where the proof-of-work computation itself has no other use. He frames this as potentially aligning with Bitcoin's original monetary principles while eliminating the 'waste' criticism, though he stresses no one knows what monetary asset will dominate a decade out despite Bitcoin's durability so far.
blockchains-as-operating-systems
AI diffusion is accelerating faster than existing social, political, and economic institutions can absorb, and Allaire expects on-chain organizational forms - blending human and agentic actors - to fill the resulting institutional gap.
Drawing an analogy to the Enlightenment and Industrial Revolution as periods that forced a renegotiation of the social contract, Allaire argues AI's accelerating diffusion (limited mainly by bureaucratic, legal, and human-risk factors rather than technical ones) will force a similar renegotiation. He expects a lag between disruption and new institutional forms, but predicts on-chain organizations with novel governance and contracting structures, mixing human and AI-agent participants, will proliferate and could become the most productive corporate forms in economic history.
ai-diffusion-and-the-social-contract
Allaire thinks double-digit global GDP growth in the 2030s is plausible from AI, but warns GDP could become a misleading metric if the gains mainly represent capital capturing value at labor's expense.
Asked to forecast GDP impact over roughly a five-to-ten-year horizon, Allaire says a discontinuous jump in productive output across industrial and commercial services is plausible, and that double-digit GDP growth in the 2030s (unevenly distributed globally) seems achievable based on current diffusion trends. But he flags the real risk: if GDP growth increasingly reflects capital capturing more of the economic pie at humans' expense, the metric loses its historical meaning as an indicator of broad economic well-being, absent a new 'social contract' to address that distribution question.
ai-diffusion-and-the-social-contract
Books referenced
100% Money - Irving Fisher - Allaire cites Fisher's book as the origin of the full-reserve-money idea behind the 1930s Chicago Plan, which he says stablecoins effectively implement today.
Lady Amazes - Gil references this sci-fi novel (title as heard in the transcript, possibly a mishearing of the actual title) about a post-AGI world where an overlay AI agent spawns representative agents for emerging demographic blocs to negotiate policy in a virtual senate; raised as an analogy for on-chain agentic governance. Allaire had not read it.
Companies
Circle - Allaire's company, issuer of USDC; founded 2013 on the thesis of a full-reserve, programmable dollar protocol for the internet.
BlackRock - Runs the system Circle uses for daily transparency reporting on USDC reserves.
Bank of New York Mellon - One of the large custodial institutions holding USDC's cash reserves.
Visa - Uses USDC on its own internal network to move money instead of the legacy banking system.
Stripe - Merchant platform integrating USDC for payments.
Shopify - Merchant platform integrating USDC for payments.
Ramp - B2B fintech that launched USDC as core to its treasury system, cited as a recent example of enterprise adoption.
Polymarket - Prediction market powered by USDC; cited as an example of crypto-native capital moving fluidly between derivatives and prediction markets.
DTCC - The Depository Trust & Clearing Corporation, the back-plane clearing system for U.S. securities, cited as moving toward tokenization.
NASDAQ - Cited among major exchanges building support for trading and distributing tokenized securities.
New York Stock Exchange - Cited among major exchanges moving toward on-chain settlement and tokenized securities.
Techniques and frameworks
Full-reserve money / Chicago Plan - 1930s proposal (Irving Fisher) that money should be fully backed and not fractionally re-lent; Allaire frames dollar stablecoins as the modern, legally codified (Genius Act) realization of this idea.
ARC (economic operating system) - Circle's own blockchain, positioned as a purpose-built settlement layer for the agentic economy: known/vetted validator set of financial infrastructure firms, USDC as the native gas token instead of a volatile asset, deterministic (non-reorgable) settlement finality in hundreds of milliseconds, and built-in privacy primitives.
Zero-knowledge proofs / ZK roll-ups - Discussed as the mechanism enabling off-chain compute with on-chain verifiability, which Allaire says becomes critical at the transaction scale billions of AI agents would require.
Proof-of-useful-work (inference as proof of work) - Emerging research direction Allaire flags: tying proof-of-work security to actual GPU inference computation instead of wasted energy expenditure (as in Bitcoin mining), potentially generating usable output while securing a network.
Summary
Circle co-founder and CEO Jeremy Allaire joins Elad Gil to trace the throughline from Circle's 2013 founding thesis to the current moment: that blockchains are, at bottom, operating systems for programmable, tamper-resistant, fully auditable money, and that this matters most once autonomous AI agents rather than humans become the primary economic actors transacting on them. Allaire roots USDC's design in a specific intellectual lineage - Austrian economics and Irving Fisher's full-reserve-money proposal from the 1930s Chicago Plan, which banks successfully killed at the time in favor of fractional reserve banking backed by FDIC insurance. He argues the Genius Act has now effectively codified that old full-reserve idea into U.S. law via regulated dollar stablecoins, which can only hold short-duration Treasuries, repos, and overnight bank cash (Circle's reserves run roughly a 13-day average maturity, custodied at institutions like Bank of New York Mellon with daily transparency reporting via a BlackRock-built system).
A large section of the conversation is about why existing financial rails cannot support an "agentic economy." Allaire's case: AI agents will increasingly do work, transact with each other, and buy specialized intelligence directly from other agents, at price points from a few cents to large institutional sums, in real time, globally, without needing to ask permission to plug into a payment rail. He argues this only became technically and economically feasible in the last couple of years with third-generation blockchains, citing USDC's transaction volume growth as evidence money velocity has picked up now that transaction costs have collapsed. This leads into Circle's own blockchain, ARC, which Allaire describes as an "economic operating system" purpose-built for regulated financial institutions rather than the censorship-resistant, government-evading ethos of earlier crypto: a known and vetted validator set of major financial infrastructure firms, deterministic (non-reorgable) settlement finality in hundreds of milliseconds, USDC itself as the native gas token so costs are denominated in real dollars, and built-in privacy primitives shipping from day one. Zero-knowledge proofs and off-chain compute verification come up as increasingly production-critical (not merely research) infrastructure, since Allaire expects transaction volumes to scale into the billions or trillions as agent populations grow.
The discussion moves to tokenization of real-world assets - securities, money markets, and currencies - which Allaire says is happening at every layer of the stack, from transfer agents up through clearing infrastructure like the DTCC to exchanges such as NASDAQ and NYSE, with the SEC having issued clarifying guidance about a month before the taping. He notes Circle's own stock is currently the most actively traded tokenized equity, ahead of Tesla or an S&P-500 product, and that most current tokenized-stock demand comes from non-U.S. investors seeking access they otherwise lack, inverting the older pattern of U.S. investors accessing Chinese equities via wrapper instruments. Gil and Allaire also touch on prediction markets (USDC powers Polymarket) as a parallel venue to traditional financial markets, and on emerging research tying proof-of-work cryptocurrency security to genuinely useful GPU inference computation, potentially eliminating the "wasted energy" critique leveled at Bitcoin mining.
The conversation closes on a more speculative, macro register. Gil raises the analogy of a sci-fi novel he calls "Lady Amazes" (as transcribed; the exact title is uncertain), in which emergent demographic blocs spawn representative AI agents that negotiate policy in a virtual senate - a governance analogy for the on-chain organizational forms Allaire expects to proliferate, blending human and agentic actors under novel contracting and governance structures. Allaire frames the current moment as one of several historical junctures (alongside the Enlightenment and Industrial Revolution) that force a renegotiation of the social contract, driven by an AI diffusion rate he sees as accelerating and limited mainly by bureaucratic, legal, and human-risk factors rather than technology. Pressed for a GDP forecast, he says double-digit global growth in the 2030s is plausible but unevenly distributed, while cautioning that GDP could become a misleading metric if the gains mostly represent capital capturing more value at labor's expense, absent new institutions to address that distributional question.
Notable Quotes
"Blockchains are operating systems, and they have compute engines. They have virtual machines, and you can write Turing Complete code." - Jeremy Allaire
"The kind of agentic economy is being born as we speak. And in that world, we need a different infrastructure for the financial intermediation." - Jeremy Allaire
"The most active tokenized stock today is not Tesla. It's not the S&P index. It's actually Circle." - Jeremy Allaire
"The risk here is that GDP effectively... is a sort of capital, capturing more capital at the expense of humans, like that's the real risk." - Jeremy Allaire