Agentic commerce will run on crypto rails, PayPal and Google reps tell Consensus Miami

At Consensus Miami on May 10, 2026, senior representatives from PayPal and Google Cloud made a claim that would have sounded absurd five years ago: autonomous AI agents will conduct commerce on cryptocurrency rails, not through traditional banking infrastructure. The panel outlined three prerequisites for this future, namely open payment protocols, machine-readable merchant catalogs, and multi-party crypto custody. If they are right, the plumbing of global commerce is about to be rebuilt. If they are wrong, two of the largest corporations on Earth just spent valuable stage time talking about a fantasy.

The Agentic Commerce Thesis

The basic idea is straightforward. AI agents, software programs that act on behalf of humans, will increasingly handle purchasing decisions, price comparisons, subscription management, and supply-chain logistics. McKinsey estimated in late 2025 that agentic AI could automate up to 30% of business tasks by 2030, with commerce-related activities among the earliest targets. Gartner projected that by 2028, at least 15% of day-to-day work decisions would be made autonomously by AI agents, up from less than 1% in 2024.

The problem is that these agents cannot use existing payment systems efficiently. Credit card networks require human-in-the-loop authentication. Bank transfers rely on business-hours settlement windows. ACH payments take days. Wire transfers cost $25 to $50 per transaction. None of these systems were designed for software that needs to execute thousands of micro-transactions per hour across jurisdictions.

PayPal's representative argued that crypto-native payment protocols solve three problems at once. First, programmable money allows agents to set spending limits, enforce conditions, and execute escrow logic without human intervention. Second, blockchain settlement is 24/7 and global, matching the always-on nature of autonomous software. Third, transaction costs on layer-2 networks and newer chains have dropped below one cent, making micro-payments viable for the first time. PayPal's own stablecoin, PYUSD, launched in August 2023 on Ethereum and later expanded to Solana, positions the company to serve exactly this use case. PYUSD's market capitalization hovered around $500 million through early 2026, modest by stablecoin standards but meaningful as a signal of corporate intent.

Machine-Readable Commerce

Google Cloud's contribution to the panel focused on a less obvious but equally critical piece of infrastructure: machine-readable merchant catalogs. Today, when a human shops online, they read product descriptions, compare prices visually, and interpret shipping policies written in natural language. An AI agent can parse this information, but doing so is fragile, slow, and error-prone. Web scraping breaks when layouts change. Natural language descriptions are ambiguous. Pricing structures buried in PDFs or behind login walls are effectively invisible to automated systems.

The Google Cloud representative proposed standardized, open data formats that merchants could publish, allowing AI agents to query product availability, pricing, return policies, and fulfillment timelines through structured APIs rather than web scraping. This is not a new concept. Google's own Product Data Specification already defines hundreds of attributes for merchant feeds. Schema.org has offered structured data vocabularies since 2011. What is new is the urgency. As agent-to-agent commerce scales, the merchants who publish machine-readable catalogs first will capture disproportionate transaction volume.

The implication is a shift in competitive advantage. Today, e-commerce success depends heavily on SEO, visual design, and brand trust with human consumers. In an agent-dominated environment, success depends on data interoperability, API uptime, and structured pricing feeds. Small merchants with clean data could outcompete larger rivals with cluttered, human-oriented storefronts.

Multi-Party Custody and Trust

The third pillar discussed at the panel, multi-party crypto custody, addresses the most difficult question in agentic commerce: who holds the money? When an AI agent spends funds on behalf of a user, that agent needs access to a wallet. But giving a software agent unilateral control over funds is a security nightmare. A compromised agent could drain an account in seconds. A malfunctioning agent could execute thousands of erroneous transactions before anyone notices.

Multi-party computation (MPC) custody offers a partial solution. In an MPC setup, the private key controlling a wallet is split into multiple shares held by different parties, perhaps the user, the agent provider, and an independent custodian. No single party can move funds alone. The agent can initiate transactions, but the custodian or user must co-sign. Spending limits, velocity checks, and category restrictions can be enforced at the cryptographic level rather than relying on application-layer controls that can be bypassed.

Several firms already operate in this space. Fireblocks processes over $6 trillion in digital asset transactions and offers institutional-grade MPC custody. Coinbase's Wallet-as-a-Service product, launched in 2023, provides similar functionality aimed at developers building agent-driven applications. BitGo, Qredo, and Fordefi round out a growing field. The market for digital asset custody was valued at roughly $500 million in 2024 and is projected to exceed $2 billion by 2028, according to multiple industry estimates.

The counterargument, voiced by crypto-skeptics and some traditional finance executives, is that these problems are already solved by existing banking APIs. Plaid connects to over 12,000 financial institutions. Stripe can process payments in 135 currencies. The SWIFT network handles $5 trillion in daily transfers. Why rebuild this infrastructure on blockchains when incumbents already work?

The Skeptic's Case

The skeptic's case deserves honest engagement. Crypto payment infrastructure remains fragile in several respects. Stablecoin depegging events, most notably the TerraUSD collapse in May 2022 that erased $40 billion in value, demonstrated that not all programmable money is equally sound. Regulatory uncertainty in the United States, where the SEC and CFTC continue to dispute jurisdictional boundaries, makes corporate treasury departments hesitant to hold or transact in digital assets. The European Union's Markets in Crypto-Assets (MiCA) regulation, fully effective since June 2024, provides clearer rules but imposes compliance costs that could price out smaller players.

Performance is another concern. Visa processes roughly 65,000 transactions per second at peak capacity. Ethereum's base layer handles around 15 transactions per second, though layer-2 networks like Arbitrum and Optimism push effective throughput into the thousands. Solana claims theoretical capacity exceeding 65,000 TPS but has suffered multiple network outages, including notable incidents in 2022 and 2023, that undermine confidence in its reliability for mission-critical commerce.

There is also the question of whether PayPal and Google are the right messengers. PayPal has historically been aggressive about account freezes and censoring transactions that violate its terms of service. Google has shut down products and services with minimal notice. Both companies operate within regulatory frameworks that require compliance with sanctions lists, anti-money laundering rules, and content moderation policies. The crypto ecosystem's original value proposition, censorship resistance and permissionless access, sits in tension with the compliance apparatus these companies bring.

Bitcoin's Role in an Agent Economy

This tension points to a deeper question that the Consensus panel apparently left unaddressed: what kind of money should agents use?

PayPal's answer is its own stablecoin, a dollar-denominated token backed by U.S. Treasuries and bank deposits. Google's implied answer is whatever stablecoins gain the widest merchant adoption. Both answers assume the dollar's continued dominance as the unit of account for global commerce.

Bitcoin offers a different answer. A fixed supply of 21 million coins, a predictable issuance schedule halving approximately every four years, and a network secured by proof-of-work rather than corporate governance, these properties make Bitcoin the only digital asset that functions as genuinely sound money. Stablecoins inherit the inflationary characteristics of whatever fiat currency they track. The U.S. dollar has lost roughly 25% of its purchasing power since 2020, measured by the Consumer Price Index. An AI agent optimizing for its principal's long-term wealth should, in theory, prefer a non-inflationary settlement asset.

The practical objection is volatility. Bitcoin's price swings make it unsuitable as a medium of exchange for routine transactions today. But the Lightning Network, which now processes over 6,600 BTC in public channel capacity and supports instant, near-zero-fee payments, offers a path toward Bitcoin-denominated micro-commerce. If agents are already going to operate on crypto rails, the question is whether those rails will carry dollar proxies controlled by corporations or a neutral, non-sovereign monetary asset controlled by no one.

The Austrian economics framework clarifies the stakes. Ludwig von Mises argued that sound money is the bulwark of individual liberty against state encroachment. Friedrich Hayek proposed the denationalization of money, arguing that private currencies competing in an open market would produce better monetary outcomes than state monopolies. An agentic economy running entirely on corporate stablecoins merely replaces one set of gatekeepers with another. An agentic economy with a Bitcoin settlement layer preserves the possibility of genuine monetary sovereignty.

What to Watch

Three developments will determine whether the agentic commerce thesis materializes or stalls.

First, watch PayPal's PYUSD adoption metrics. If PYUSD's market cap breaks $1 billion in 2026 and merchant integration expands beyond early adopters, the stablecoin-as-agent-payment thesis gains credibility. If PYUSD remains a niche product, the market is signaling that demand for corporate stablecoins is weaker than the hype suggests. Tether's USDT, already exceeding $140 billion in market capitalization, and Circle's USDC, at roughly $60 billion, remain the standards to beat.

Second, watch for open-standard proposals around machine-readable merchant catalogs. If Google, Shopify, Amazon, or a consortium of smaller platforms publish open specifications that agents can consume, expect a rapid increase in agent-driven transaction volume within 12 to 18 months. If each platform builds proprietary APIs, fragmentation will slow adoption and favor large incumbents with the engineering resources to support multiple integrations.

Third, watch the regulatory trajectory in Washington. The U.S. stablecoin legislation that has been debated since 2023 would, if passed, provide legal clarity that corporate treasury departments need before committing to crypto payment rails. The EU's MiCA framework already provides this clarity for European companies. If the U.S. fails to pass comprehensive stablecoin rules by the end of 2026, expect agentic commerce innovation to migrate toward jurisdictions, including the EU, Singapore, the UAE, and Japan, that offer clearer regulatory environments.

The vision presented at Consensus Miami is plausible but incomplete. AI agents will almost certainly need programmable, global, always-on payment infrastructure. Crypto networks provide the best available candidate for that infrastructure. But the choice between corporate stablecoins and neutral, non-sovereign money like Bitcoin is not merely a technical decision. It is a decision about who controls the financial nervous system of an increasingly automated economy. PayPal and Google have told us where they think the future is heading. They have not told us, and likely would prefer not to be asked, whether that future should have a single point of corporate control or none at all.

---

Source: CoinDesk