How AI Agents Will Transform the Financial System with Circle Co-Founder and CEO Jeremy Allaire
In a Nutshell
Jeremy Allaire explains Circle's USDC stablecoin as full-reserve, programmable digital dollars enabling instant, global, low-cost transactions on blockchains, positioning them as the foundation for an AI-driven agentic economy where autonomous agents conduct micro-payments (e.g., cents for AI outputs) and complex financial activities. Blockchains act as tamper-resistant operating systems for machines intermediating economic value, with Circle's ARC network—backed by USDC and run by trusted validators—delivering scalable, privacy-enabled infrastructure for real-world adoption, including tokenized assets like treasuries (USYC) and RWAs. Over the next decade, AI agents will explode transaction volumes, tokenize global assets, and spawn new on-chain organizations, potentially driving double-digit GDP growth amid a renegotiated social contract.
These notes were generated by AI and may contain inaccuracies.
Jeremy Allaire, co-founder and CEO of Circle, discusses cryptocurrency, AI, agentic payments, AI evolving on the blockchain.
Circle founded in 2013. Goal: create a protocol for dollars on the internet, inspired by Bitcoin. Enable instant, global, frictionless, no-cost value storage and movement. Vision of programmable money where blockchains become operating systems for machines intermediating economic activity, including autonomous software.
Blockchains would become like operating systems and you could actually have machines that intermediate economic activity and financial activity on the internet.
Commoditize payment utility with safe digital dollar currencies and tamper-resistant programmability to improve financial system: safer, more accessible, efficient, new utility for money.
Early crypto aimed to divorce from traditional finance, e.g., Bitcoin response to 2008 crisis and bailouts. Allaire believes in Austrian economic thought, studied in 1990s. Global financial crisis drew him to build safer system via full reserve money.
Bitcoin as full reserve: no fractional lending. Full reserve money differs from fractional reserve banking; backed by government obligation (e.g., US dollar) but no fractional lending against it.
1930s Great Depression led to Chicago Plan by economists like Irving Fisher (book: 100% Money). Proposed full reserve government money. Banks lobbied against; resulted in FDIC insurance instead, preserving risk-taking.
Great Financial Crisis: 30x, 12x, 14x leverage. Current system depends on reserve currencies like dollar for 30-50 years. Stablecoins as full reserve money; codified in Genius Act last year: narrow money model, no fractional reserve lending.
Stablecoins backed 1:1 by safe, liquid assets per payment laws. Now federal law: short-duration US government treasuries or overnight treasury collateral with global banks, plus cash for liquidity in custodial institutions like BNY Mellon.
USDC transparent via daily reports with BlackRock. Purchasable for $1, backed by treasuries, repos, short-duration T-bills (average ~13 days duration), super liquid cash-like instrument.
General protocol for dollars on internet. Micro-transactions: 25 cents for digital game objects; AI agents paying 20-50 cents for AI output. Macro: electronic trading firms settling multi-million dollar capital markets transactions.
Payload-agnostic like email. Used by Stripe/Shopify merchants, Visa (internal money movement bypassing legacy banking), neo-banks, remittance, B2B fintech treasuries, invoice payments.
24/7 availability (e.g., fundraising on weekends). Internet-native: works like WhatsApp/WeChat globally. Low fees relative to volume. Store of value/exporting digital dollars worldwide (Genius Act motivation, geoeconomic strategy).
Programmable money: public API on internet. Developers add global dollar settlement without permission via smart contracts.
Smart contracts: code wrapped around money for conditional payouts, financial instruments.
Blockchains as OS like mobile/web/cloud/AI foundation models: Turing-complete code, virtual machines.
Key attributes: tamper-resistant code, perfectly auditable (public blockchain), transaction/compute integrity. Provable inputs/outputs/state crucial for AI-driven economy and fiduciary integrity.
Early speculation: AGI from blockchain agents in multi-turn economic games with incentives/game theory.
Current shift: 3 months into dramatic tech capability jump. AI agents will conduct white-collar/service work, collaborate, purchase specialized intelligence/output.
Agentic economy needs new financial infrastructure: global, interoperable, instant, programmable by arbitrary software, dynamic financial endpoints, billion/trillion-scale/micro-scale (5-10 cents) real-time transactions.
Crypto delivers: USDC transaction volumes exploding (most transacted digital currency), sub-cent costs reliable on third-gen blockchains like ARC (transactions for millionth of penny).
AI agents need trustworthy medium for coordination, value storage, contracts: real-time, provable via blockchain.
Enables organization of compute work, new corporate forms.
ARC: economic OS for storing/moving value, instantiating corporate forms, complex contracts. Moving from speculation to real economic activity with stablecoins, regulatory clarity.
Drivers: machines in economic output. Substance of contracts/corporations as software machines.
Bitcoin less adaptable. Solana/Ethereum for smart contracts.
Early blockchains from early adopter phase focused on completely
Decentralization is a good goal, building an alternative universe censorship-resistant and outside government reach. As adoption scales to major companies like Walmart or households storing wealth, intermediaries will persist; people won't all be their own bank. Intermediaries need robust infrastructure. ARC features a known validator set run by major financial infrastructure companies held to high standards for SOC compliance, reliability, and availability.
Examples of ARC validators haven't been announced yet but will be soon. The model involves financial infrastructure companies, possibly large technology companies, running the distributed infrastructure. This provides assurances that bad actors aren't processing transactions and guarantees deterministic settlement finality—no hard forks or reorgs—in hundreds of milliseconds for securities, cash, or other assets.
ARC uses real money as the foundation: USDC as the default native token, legally electronic money, avoiding volatile gas tokens. Companies budget like AWS credits, simplifying treasury, operations, and compliance. Primitives support payment systems, capital markets, privacy, and compliance. Circle has worked with leading institutions like Visa, BlackRock, Bank of New York Mellon, and central bankers worldwide to incorporate requirements for running the financial system on internet infrastructure.
These new distributed network operating systems must support the real economy, not a shadow economy.
Scaling models are key for network computers excelling at public recordkeeping and computing on those records. For billions of AI agents, scaling is critical. ZK rollups and zero-knowledge proofs prove off-chain compute on-chain. Off-chain compute, trusted execution environments, and cryptographic proofs enable privacy while maintaining open, interoperable, permissionless infrastructure. Corporations need privacy; individuals avoid doxxing. ARC ships day one with built-in privacy primitives.
As large-scale financial infrastructure like NYSE or derivatives clearing houses moves on-chain, scaling becomes critical. Compared to early internet (1990s to 2001, 10 years of grinding), blockchain is at the broadband moment with Wi-Fi, broadband, usable devices enabling software, media, real-time communications delivery. Now, societal demands, financial system needs, and agentic economy converge.
Tokenization of stocks like fractional Berkshire Hathaway shares using crypto is happening, globally accessible via USDC success. Check RWA.xyz for tokenized stocks; most active is Circle, not Tesla or S&P index. Tokenized money markets like on-chain treasury bills: Circle operates largest USYC (US Yield Coin), fastest-growing tokenized treasury product, and largest tokenized euro EURC.
Tokenization efforts span the stack: record-keepers (Computershare), depositories/clearing (DTCC), brokers/exchanges (NASDAQ, NYSE). SEC issued guidance a month ago on obligations. Growth enables non-US access to US assets, mirroring past Chinese stocks via third-party instruments, ETFs, funds.
Beyond porting existing products, tokenization unlocks fractionalization, borrowing/lending, new packaging. AI plays a significant role. Ties to prediction markets: stock/financial markets are world's biggest; prediction markets become parallel infrastructure. Market makers in prediction markets determine reality for companies/equities. USDC powers Polymarket; traders move USDC for events alongside oil/Bitcoin derivatives.
Papers explore tying proof of work to generic inference work: GPU-efficient inference generates revenue. Proof of work's exhaust is energy waste; inference compute as proof of work makes it productive, aligning with Bitcoin principles. Opens new paradigms with energy infrastructure shifts, intelligence conversion, compute performance—unavailable 15 years ago. Bitcoin has network effects but future money in 10 years uncertain.
Pace of AI diffusion accelerates, limited by bureaucracy, legal, human risk. Opportunity for new social, political, economic structures amid renegotiated social contract, like Enlightenment/Industrial Revolution. Lag between disruption and new institutions; emergent on-chain organizations with governance, contracting mixing humans/agents could be most productive corporate forms, overlaying political systems.
Book reference: Lady Amazes (sci-fi, post-AI world)—as demographics emerge, overlay AI spawns viewpoint agents for virtual senate negotiating policy.
GDP Impact of AI
Technology impacts on GDP hard to measure (internet productivity vs. growth); reasons include measurement issues, deflation. AI could drive discontinuous GDP jumps via productive output in industry/services. GDP may lose meaning; risk of capital capture at humans' expense without new social contract. Potential for double-digit GDP growth in 2030s (e.g., Cathie Wood's view), achievable in large world parts.
Keep No Priors: AI, Machine Learning, Tech, & Startups in your library
Save the videos and channels worth coming back to, and find them again in one place.
