The viral rise of OpenClaw^[1] (formerly Clawdbot) marks a generational leap in autonomy. When these AI agents began interacting with each other — in some cases negotiating and transacting — the agentic future shifted from science fiction to operational reality.

 

OpenClaw is just one step on an accelerating journey. Trillions of dollars are being invested to build an AI-enabled world. AI spending by US hyperscalers alone is projected to exceed $650bn in 2026, roughly ten times the inflation-adjusted cost of the Apollo program.^[2]

 

What began as simple chatbots is rapidly evolving toward agentic, fully autonomous AI systems. These AI agents will not just generate content but become economic actors – ones that can reason, act, transact, debate, coordinate and more, all without real-time human supervision. The effects of this buildout will be felt everywhere, but perhaps nowhere more than in commerce.

 

Some estimates suggest AI agents could mediate $3 to $5 trillion in global consumer commerce by 2030.^[3] If even 10% of that volume becomes agent-to-agent, programmatic commerce, that implies hundreds of billions in annual machine-native settlement flows.

 

This naturally leads to the question: what financial and coordination rails make sense for AI agent native commerce?

 

Commerce today is built for humans and involves personal identity verification, banking intermediaries, legal contracts, settlement windows, and manual oversight. Autonomous software cannot walk into a branch to open a bank account, physically sign documents, or wait days for ACH clearance. Agents require infrastructure that is programmable, always-on, globally accessible, permissionless, and machine-verifiable by default.

 

Blockchains can satisfy those constraints, and we are already seeing that dynamic emerge.

 

Coincident with OpenClaw going viral in January, Solana transactions and active addresses also began climbing. Evidence on Moltbook, the social network for its AI agents, suggests they may have contributed to this growth.

 

 

x402 is an internet-native payment protocol developed by Coinbase that allows AI agents to pay for digital resources in real time without accounts or complex, high-friction authentication. Since its launch in 2025, transactions have accelerated.

 

 

It is still early and the examples today are more directional than definitive. But If investors are excited about the possibility of AI innovation, it would be remiss to overlook why we believe blockchain rails will be foundational to unlocking a world of fully autonomous agents.

 

Levels of Autonomy

Many will rightly point out that AI agents today do not need blockchain. This is true near-term, but we believe it is a short-sighted view.

 

McKinsey recently published a framework^[4] describing six levels of automation for AI-driven commerce, from basic subscription assistance (Level 0) to fully autonomous agent-to-agent commerce (Level 5). The key insight is that Levels 0 through 4 do not require new financial rails. In each case, a human identity sits behind the transaction. The user has authenticated with ChatGPT, Amazon or Perplexity. They have a credit card on file. When the agent transacts, it’s acting as a proxy for that human and inheriting their identity, their payment credentials, and their legal standing.

 

 

 

 

The foundations for this type of commerce – shared payment tokens, chargeback systems, fraud detection infrastructure – already exist via Visa or Stripe and they work reasonably well.

 

Blockchain rails become critical at Level 5 and beyond: when agents transact directly with other agents without human direction; when there is no human identity to inherit, when payments must be programmatic, conditional, and settled in milliseconds; and when agents need portable reputation across platforms.

 

As long as humans remain economically liable, legacy rails suffice. Once agents become economically independent actors, the constraint set changes.

 

Agentic Finance

To understand where value accrues, and why blockchain matters, we must imagine the logical end state of agentic AI. We are moving toward a world where agents are not merely human assistants, but independent economic actors. Some will be created by companies or individuals. Others will be generated by agents themselves, forming increasingly independent systems that reason, allocate capital and transact without real-time human supervision.

 

If no human is specifying a transaction channel (e.g. go to a bank, use Stripe, spin up a blockchain wallet), then agents will rationally select rails that maximize speed, reliability and global reach while minimizing friction and dependency. When the alternative is opening a bank account and waiting for ACH settlement during limited banking hours, agents will naturally choose permissionless, 24/7 blockchain rails.

 

We see three key constraints that will push agents to blockchain rails:

 

• Identity and Access: How do we track unique identities of AI actors that are transacting with each other and registering for services? What does a new reputation system look like when legacy credit scoring and fraud detection systems were built for humans with physical footprints operating within jurisdictional borders?

• Currency and Payments: What forms of money are needed when agents are making countless micropayments, executing conditional payments, and greatly increasing the need for cross-jurisdictional commerce? What forms of accounts are needed when agents can’t walk into a bank branch to open an account?

• Trust-minimized transactions: How do AI agents avoid the friction of disputes that require human arbitration or other forms of centralized trust, systems they may not be able to or choose to access?
•  

Identity and Access

Before an agent can pay for something, the counterparty must know who, or what, it is dealing with.

 

Traditional identity systems were built for humans. They rely on government IDs, physical signatures and other credentials that assume a legal person is on the other end.

 

An autonomous AI agent has none of these. It can’t walk into a bank to open an account or legally sign a contract. And yet, if we want agents to transact autonomously, they need some way to prove they’re legitimate and authorized to act.

 

If you connect an agent to your bank account, the questions multiply. How do you run anti-money-laundering checks on software? Where does liability fall if the agent acts autonomously? What if it was manipulated?

 

In simple cases, an agent can inherit its owner’s credentials (e.g. ChatGPT Checkout). But this model breaks down at scale. Multiple agents require separable permissions and spending limits. Misbehavior must be isolated without freezing all agents. These scenarios require agents to have their own verifiable identities, not borrowed human ones.

 

This is where blockchain-based identity becomes useful. Using cryptographic techniques, an agent can prove it’s authorized to act on behalf of a specific person or company without revealing sensitive information about that person. Think of it like a digital power of attorney that can be verified instantly by anyone, anywhere, without calling a lawyer or checking a database.

 

Emerging standards such as Ethereum’s ERC-8004 propose on-chain registries where agents can establish verifiable credentials and build transaction history and reputation over time. An agent that has successfully completed thousands of transactions without disputes becomes meaningfully different from a brand-new agent with no history, and that reputation becomes portable across platforms.

 

 

This matters because trust is a prerequisite for commerce. Merchants have spent years building systems to block bots and scrapers. In an agent-driven economy, now they need to figure out how to let the right bots through. A cryptographically secure and verifiable identity gives merchants confidence without requiring human sponsorship.

 

Programmable Money and Micropayments

Traditional payment rails were designed for human-scale transactions. When you pay for a coffee or a pair of jeans, a credit card’s transaction fees (typically 2-3% plus around 30 cents per transaction) are immaterial.

 

But agent-to-agent commerce operates at a completely different scale. An agent writing code may make 10,000 API calls in a single task. An agent that is comparison shopping may check across hundreds of data providers. Payments need to occur in milliseconds, repeatedly, and in fractions of a cent.

 

Credit card networks are not optimized for this behavior. Minimum fees make micropayments uneconomic. Fraud systems are tuned to freeze accounts exhibiting high volume machine-like activity. Transaction speeds are fractions when compared to high-performance blockchain protocols.

 

Stablecoins and programmable money are genuinely useful here. On-chain transactions can be fractionalized down to small units with settlement costs near fractions of a penny. More importantly, because the payments are programmable, they can be conditional: pay X only if the API returns valid data, release funds only when the compute job completes, stream payments in real-time as a service is consumed rather than paying upfront for a block of capacity you might not use.

   
                                         

 

Programmability also improves capital efficiency. Today, you typically have to prefund an account for your agent to access a new service. You need to approximate usage and lock up capital in advance. With smart contracts and on-chain collateral, an agent can prove solvency without transferring payment until the service is delivered.

 

Blockchain enables financial infrastructure that matches how agents should work: autonomous, high-frequency, conditional, and capital-efficient.

 

Trust-Minimized Transactions

Traditional commerce embeds trust in intermediaries. Payment processors manage chargebacks. Banks provide settlement guarantees. Courts adjudicate disputes. Contracts ultimately rely on human legal systems for enforcement.

 

This framework becomes inefficient when billions of low-value transactions occur across jurisdictions. An AI agent transacting with another AI agent may not have access to, or choose to rely on, a specific jurisdiction’s legal system. Cross-border enforcement can be slow, expensive, and uncertain.

 

Blockchains reduce reliance on these fallible trust systems by using smart contracts to directly encode enforcement. For example, smart contracts allow funds to be escrowed programmatically and released only when predefined conditions are met. Settlement is deterministic rather than subject to chargeback risk. Rules are transparent and verifiable in advance by both parties. There is no need to rely on legal remedies.

 

For autonomous agents operating at scale, minimizing reliance on centralized intermediaries and human arbitration reduces friction, increases predictability, and allows commerce to scale programmatically. This lower friction infrastructure may increase the surface area of economic activity that would otherwise be uneconomical under legacy enforcement models. Agentic commerce, enabled by blockchain rails, could accelerate global GDP growth.

 

This is the Beginning

The question isn’t whether agentic commerce is coming. It’s what infrastructure it will run on.

 

As AI agents become autonomous economic actors, the number of economic actors in the global economy increases exponentially. Agents will require a digitally native financial rails, a technology stack that can handle programmatic settlement, high volume micropayments, permissionless coordination, and trust-minimized identity systems. These principles are fundamental to the design of blockchains.

 

We believe it’s fair to say that the rapid adoption of AI agents is a strong secular tailwind for blockchain activity. There’s already evidence this is happening, and it is a value creation opportunity we believe most investors are underestimating.

 

---

1. https://x.com/MindBranches/status/2021816290516582480 ↩

2. https://www.carsongroup.com/insights/blog/big-tech-capex-plans-are-ballooning-is-that-good-or-bad/ ↩

3. McKinsey & Company; The LeadGen Economy ↩

4. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-automation-curve-in-agentic-commerce ↩