AI agents need crypto rails because the traditional banking system cannot settle machine-speed, sub-cent, always-on payments between pieces of software — and macro investor Raoul Pal argues the layer-1 blockchains that can (Ethereum, Solana and Sui) become the ownable infrastructure of an emerging “invisible economy.” In a May 2026 presentation on his show The Journey Man, Pal predicted that within two years the majority of economic transactions on Earth will happen between machines, entirely beyond human perception.

That is a bold claim, and this analysis unpacks the mechanics behind it: what an “agentic economy” actually is, why bank plumbing breaks under it, and why Pal thinks the investable opportunity is not any single application but the base-layer rails themselves.

Key takeaways

  • Raoul Pal’s core forecast: within roughly two years, most economic transactions will be machine-to-machine and invisible to humans, running at “a million times faster than human neurons.”
  • Why crypto rails, not banks: banks require KYC, settle in hours or days, and stop at one cent — machines need instant settlement, 18-decimal precision and wallet-based identity that only blockchains provide.
  • The investment thesis: Pal argues you should “own the rails” — the layer-1 blockchains he names as capable of the load are Ethereum, Solana and Sui (he discloses he sits on the Sui Foundation).
  • The market-cap math: extending crypto’s historical growth trend takes total market value from ~$2.5 trillion to ~$100 trillion within a decade, which Pal calls conservative if blockchains become the settlement layer for the whole agent economy.
  • The human hedge: Pal’s advice is to own the infrastructure and use agents yourself, rather than be the labour an agent replaces.

Why AI agents need crypto rails instead of banks

AI agents need crypto rails because the existing financial system was built for humans moving human-scale money, and it physically cannot keep pace with software transacting at silicon speed. Pal’s argument is blunt: “a machine can’t use a bank.” KYC onboarding, settlement measured in hours or days, and a smallest unit of one US cent all assume a human is in the loop.

Autonomous agents break every one of those assumptions. They transact continuously — “24 hours a day, 365,” in Pal’s words — in amounts far below a cent, at a volume no card network or correspondent-banking chain was designed to clear. Pal’s contrast is personal: moving money from the Cayman Islands to the United States takes him three days, while a modern layer-1 can settle in around 300 milliseconds.

Crypto rails supply the four things machines require and banks cannot: near-instant final settlement, micro-denomination (down to 18 decimal places), a wallet that doubles as portable identity with no human required, and programmability so payments can be conditional and automated. That is why Pal frames stablecoins, DeFi, tokenization and on-chain identity not as separate crypto use cases but as features of a single machine-native payment system.

What is the agentic economy?

The agentic economy is Pal’s term for a financial system dominated by autonomous software agents that don’t just answer questions but perceive, decide, negotiate and transact on their own — signing transactions, spawning sub-agents, and earning and spending real budgets. Crucially, these agents replicate: a successful strategy is copied across millions of instances almost instantly, so the number of economic participants explodes far faster than any human-driven platform ever did.

Pal cites forecasts of an 80-to-1 ratio of agents to humans, and claims of up to 60% of crypto transactions being agent-driven by the end of 2026 (a figure he treats sceptically for this year but likely by next). Whether or not those exact numbers hold, the direction is what matters: he argues that every total-addressable-market estimate built on human activity is “wrong by orders of magnitude,” because it assumes human-scale adoption curves for a population of actors that self-replicates.

He also invokes network math. Human platforms tend to follow Metcalfe’s Law, where value scales with the square of users. Agent networks, Pal argues, can follow Reed’s Law — value scaling with group-forming combinations — producing a far steeper curve. We explore that distinction in our breakdown of Reed’s Law and the exponential age of crypto.

Why layer-1 blockchains, not layer 2s

Pal’s investment conclusion is that the value accrues to layer-1 blockchains specifically, because they are the settlement substrate the entire agent economy has to run on. He draws a sharp line between layers: Bitcoin (and Zcash) he classes as store-of-value, while layer 2s “rent” their security from a layer 1 rather than provide the base coordination themselves. Only the layer 1, in his framing, is the technology stack that actually runs the machine economy.

To value that, Pal proposes a thought experiment: if you could switch off Ethereum, you would destroy every layer 2, all of DeFi, roughly 70% of stablecoins, the NFT market and any real-world assets settled on it. Ethereum, he argues, is not a business to be judged on cash flow but a utility whose price should reflect the future economic activity it coordinates — which, on his agent thesis, points to a valuation “in trillions.”

The specific names he lands on as capable of the load are Ethereum plus its layer 2s, Solana, and Sui — with the disclosure that he sits on the Sui Foundation but says he arrived via his own research. He expects concentration like every prior technology wave: cloud, operating systems and smartphones all settled into a handful of dominant platforms, and he thinks base-layer blockchains will too.

Tokenization: everything becomes machine-readable data

A second pillar of Pal’s thesis is that “tokenization” means far more than putting stocks on-chain. He points out that the tokens in an AI model and the tokens on a blockchain share a name for a reason — both are packets of digital information — and predicts that essentially all data will become machine-readable and tradable in marketplaces that don’t exist yet.

Science data, university archives, climate and soil records, farm data — Pal argues each has value that agents will discover, buy and monetise using tokenized money, because on-chain settlement is the only mechanism fast enough for that volume of micro-transactions. This reframes tokenization as infrastructure for agents rather than a convenience for human investors, and it dovetails with the institutional side of the trend we cover in Ondo Finance and tokenized treasuries.

The $100 trillion number and the economic singularity

Pal’s headline figure comes from extending the long-run logarithmic growth trend of total crypto market capitalisation, which he says takes it from around $2.5 trillion today to roughly $100 trillion within ten years — and he calls even that conservative if blockchains genuinely become the coordination layer for all digital economic activity.

He ties this to what he calls the economic singularity: the point, which he pencils in around 2030–2032, where GDP as we measure it stops working. His reformulated growth equation replaces the old “population + productivity + debt” with “humans + robots + AI + debt + energy density + compute efficiency,” each component going exponential as falling energy costs and rising compute efficiency compound. We examine that framework in depth in our piece on Raoul Pal’s economic singularity thesis.

Importantly, Pal rejects the doom narrative that agents destroy demand and jobs. His counter-argument is that agents are billions of new economic participants that consume energy, compute, storage and data — so demand explodes rather than collapses, even if most of that activity is invisible to humans.

What this means for investors

Pal’s practical takeaway is deliberately simple and, as always, not investment advice: “own the rails.” Rather than trying to pick the winning application, he argues the base-layer blockchains are — for the first time, unlike the internet — infrastructure that anyone can own a fraction of, from any country and at any income level. His personal shortlist is Ethereum, Solana and Sui, with Sui flagged as the earliest and highest-risk of the three.

The second half of his advice is human: don’t be the labour an agent can replace. Pal leans on the philosophical idea of qualia — subjective human experience — as the one thing machines don’t have, and suggests people position themselves in culture, community and experience while their capital compounds on the rails underneath. Readers should treat these as one investor’s framework, do their own research, and size any exposure to volatile assets accordingly.

Frequently asked questions

Why do AI agents need crypto rails?

AI agents need crypto rails because banks require human KYC, settle in hours or days, and stop at one-cent minimums, while agents transact continuously in sub-cent amounts at machine speed. Raoul Pal argues blockchains uniquely provide near-instant settlement, 18-decimal precision, wallet-based identity and programmability — the features machine-to-machine payments require.

What is the agentic economy?

The agentic economy is a system in which autonomous AI agents perceive, decide and transact on their own — earning, spending, and spawning further agents — rather than only answering human prompts. Raoul Pal predicts that within about two years most economic transactions will be these machine-to-machine exchanges, invisible to humans and running at silicon speed.

Which layer-1 blockchains does Raoul Pal favour for the agent economy?

In his May 2026 presentation, Raoul Pal named Ethereum (with its layer 2s), Solana and Sui as the layer-1 blockchains he believes can handle the agent economy at scale. He discloses that he sits on the Sui Foundation but says he reached the conclusion through his own research, and describes Sui as the earliest-stage and highest-risk of the three.

Is Raoul Pal’s $100 trillion crypto forecast realistic?

Raoul Pal derives the figure by extending crypto’s historical logarithmic growth trend, which he says moves total market capitalisation from roughly $2.5 trillion to about $100 trillion over ten years. He argues this is conservative if blockchains become the settlement layer for the agent economy, but it remains a trend-extrapolation forecast, not a guarantee, and should be treated as one investor’s view.