Cathie Wood Says AI Agent Spending Could Reveal Where Demand Is Heading

Cathie Wood thinks investors need to start paying attention to a new source of demand signals soon: AI agent spending.

Speaking at Robinhood’s Summit in Houston, the ARK Invest CEO argued that investors should think about spending by autonomous agents. Wood’s revised her old maxim (“follow the developers”) about finding value by watching who makes useful technology. Investors may increasingly be able to “follow the agents,” she argued, to see which technologies are gaining traction.

Related: XRP Scarcity Index Turns Negative: What Does It Mean for XRP Price?

Background: What Cathie Wood Means by “Follow the Agents”

The basic idea behind “follow the agents” is straightforward.

Right now, investors looking to identify emerging sources of technology adoption typically look to user behavior.

A large and growing user base, active developers, and increasing downloads or revenue can all be signs of wider acceptance.

Autonomous agents could generate a similar signal.

An agent may need to spend money to fulfill its programming, whether that takes the form of an API call, buying a good or service, or even paying another software agent to perform a task.

If millions of agents begin using the same third-party API to access a data set, that signals the high data value.

That is why AI agent spending is potentially valuable to investors: it takes software adoption to a new level.

It transforms it into payments.

Why AI Agent Spending Could Become a Demand Signal

There is one economic metric that is hard to fake: cash.

If autonomous agents begin regularly spending cash on a service, that has value to the service provider.

For example, let us say that a research agent needs to buy market data in order to perform its analysis.

It has a choice of three providers, one of which it has previously used.

When the agent renews its API key, it is making a statement about value: the most reliable data for the price.

This sort of transaction could be valuable to investors, as it indicates which firms are capturing demand from autonomous agents.

It may be much harder to predict, but it could provide more insight into long-term value trends.

That, according to Wood, is why investors should care about “following the agents.”

Autonomous Agents Are Already Being Used To Make Purchases

The “payment” part may sound far-fetched, but companies are already working on ways for autonomous agents to make live transactions.

In July, Visa announced that its European agents were already testing autonomous transactions in controlled real-world environments, using live cards and real merchants.

Amazon has also been working on the concept, announcing in August 2026 that Amazon Bedrock AgentCore Payments had become generally available.

The tool enables agents to discover and pay for APIs, Model Context Protocol services, and digital content without human input.

Related: Trump Could Put AI Under a 10-Member Watchdog as US Renames It “Super Intelligence”

Once that process becomes widespread enough, AI agent spending will become an economic force.

How Big Could This Get?

According to ARK estimate, agents could consume more than $8 trillion worth of goods and services online by 2030.

ARK also estimates that their share of total digital expenditures could rise from around 2% in 2025 to 25% by 2030.

ARK argues that agents could drastically shorten the purchase cycle, accelerating the path from discovery to spending.

Rather than having a human buyer perform each step of the process, an agent could handle it all, from scanning for available products to comparing prices and completing the purchase.

It is important to note that none of this is guaranteed to happen.

Even so, if even 1% of ARK’s predicted value is realized, it would represent an extraordinary opportunity for investors.

That is why they should be watching AI agent spending closely.

What Will Autonomous Agents Spend Their Money On?

But there is another category that they will consume much more frequently: software.

A sophisticated agent may need to rent or purchase access to multiple models, or call an external API to perform an action it was not programmed to do.

The same agent may also buy data, rent computing power, or pay another agent to perform an analysis.

For example, let us imagine that an autonomous agent is performing financial research. It may need to:

• Subscribe to a market data API in order to analyze prices
• Pay another firm to analyze blockchain wallet data
• Buy access to a specific model
• Scan a news database for relevant articles
• Pay a payment processor to settle each transaction

This has obvious implications for the traditional software sales model.

If agents can pay by the query, by the token, or by the analyzed job, it could capture a much larger share of the market.

By contrast, software with seats or licenses that only renew annually is far less appealing to agents.

Why Crypto Is a Good Fit for Agent Payments

Autonomous agents work best with payment systems that can operate programmatically without requiring human intervention for every transaction

That is why stablecoins and other blockchain-native financial instruments could be particularly attractive for agent payments

Coinbase’s x402 protocol, for example, effectively revives the HTTP “402 Payment Required” status code by allowing computers to request direct payment when encountering a paid API.

Amazon Bedrock AgentCore Payments already supports x402 payments, alongside the Machine Payments Protocol (MPP). Meanwhile, Coinbase’s x402 implementation allows agents to make payments with USDC, including on the Base and Solana blockchain networks.

If the number of autonomous agents explodes, so will the number of machine-to-machine transactions, many of which will be settled in stablecoins.

Of course, that may not happen.

It is entirely possible that traditional card networks will retain their dominance in the payments space, even when it comes to autonomous agents.

Visa and Stripe, in particular, have both been developing their own tools for agents that can authorize payments using the same tokens that humans use.

Both companies are also partnering with OpenAI to facilitate payments using models such as GPT.

Stripe has already announced the launch of a number of tools that allow agents to spend money on behalf of humans.

It suggests that companies are already preparing for the dawn of agentic commerce, even as crypto startups rush to grab the largest share of the market.

In the end, it is not about card payments versus stablecoins: it is about which payment infrastructure can attract the largest number of autonomous agents.

What Metrics Should Crypto Investors Watch?

Transaction volume will be important, but it is unlikely to capture the full picture.

Most interesting to investors will be the patterns of behavior that emerge among autonomous agents.

For crypto, this could include a number of different metrics.

• How much stablecoin volume is being spent on agents? How many x402 payments are being processed? How many APIs are being opened up to machine access?
• What is the average transaction size? Will agents pay repeatedly for the same good or service?
• What blockchains are being used for agent payments? What fees are being paid?
• How many agent wallets are being created? How is the spending distributed among them?

The value of AI agent spending as an economic metric will ultimately depend on how much it reflects actual value.

One million free API calls by an agent is an interesting sign, but it does not have nearly the value of one agent making repeated payments for the same service.

So, while agent activity may be an interesting metric to watch, AI agent spending has the potential to be much more valuable to investors.

Related: XRP 2016 Pattern Returns: EGRAG Crypto Says $50 Could Be in Play

Could AI Agents Influence the Winning Blockchains?

If agents are choosing what to spend their time and money on, it could potentially reshape the entire industry.

Humans tend to favor certain products and services because of reputation, brand image, or availability.

An autonomous agent, on the other hand, can optimize its spending based on raw value.

It can favor one cloud infrastructure over another due to reliability or price, or pick one database over another due to superior features.

This means that an agent can be much more picky about the services that it uses, potentially shifting the balance of power between different providers.

A blockchain, for example, must offer better value to an agent than its competitors, or the agent will use another blockchain.

That is why AI agent spending could transform the competitive landscape for any industry that attracts autonomous agents.

Is AI Agent Spending a Useful Signal For Now?

Not yet.

There is clearly potential down the road. AI agent spending is still early and fragmented, but it is no longer hypothetical

Many of the techniques that ARK envisions will dramatically increase AI agent spending will still require further development.

Meanwhile, much of the spending that does happen will be controlled by developers, as opposed to autonomous agents.

If an agent makes a purchase because the developer hard-coded it into the software, it may not reflect true value as closely as one might hope.

Why Does Cathie Wood’s “Follow the Agents” Matter To Crypto Investors?

The real appeal is in a third category: potential value discovery.

If software agents begin to adopt various technologies as their own, the increased adoption could represent a potential source of value to the firms that provide those technologies.

That could apply to virtually any type of business technology, from databases to applications to models to blockchain networks.

For crypto, the opportunity is particularly compelling.

Stablecoins and blockchain networks still overwhelmingly serve human-directed activity, but autonomous agents are already beginning to use them for machine payments. If agent payments sharply grow in volume, that could make blockchains far more valuable.

Related: BitGo CEO Mike Belshe Warns US Crypto Is at Risk After CLARITY Act Failure

Whether that happens is impossible to say with certainty at this point, but if it does, it will not be due to marketing or narrative — it will be due to value.

If Cathie Wood is right, investors can follow the money to find it.

FAQ

What Is AI Agent Spending?

AI agent spending refers to payments made or initiated by autonomous AI systems while completing tasks, including payments for products, APIs, data, computing resources, and other digital services.

Why Does Cathie Wood Think AI Agent Spending Matters?

Wood believes investors may increasingly be able to “follow the agents” to see which technologies autonomous systems actually choose and pay for.

How Could AI Agent Spending Benefit Crypto?

Stablecoins and blockchain networks can support automated, programmable machine-to-machine payments without requiring human approval for every transaction.

Will AI Agents Use Stablecoins Instead of Cards?

Not necessarily. Visa, Stripe, and other traditional payment companies are also developing payment infrastructure specifically for autonomous agents.

Is AI Agent Spending Already a Useful Investment Signal?

Not yet. Much of today’s agent spending is still controlled by developers, so it does not always represent an autonomous choice between competing products or services.

The post Cathie Wood Says AI Agent Spending Could Reveal Where Demand Is Heading appeared first on Bitcoin Foundation.