Analysis: Cloudflare’s Earnings Call Signals the Rise of the AI Agent Economy, and On-Chain Settlement Could Be the Big Winner

The crypto market spends a lot of time watching token prices, ETF flows, and headline-friendly partnerships. But sometimes the most important signal comes from a place people do not expect: a traditional infrastructure company’s earnings call.

That was the argument made by ARK Invest research lead Lorenzo Valente after Cloudflare’s latest results. His core thesis is simple: if AI agents are becoming a major source of internet traffic, then the internet’s payment model has to change with them. And once that happens, blockchain infrastructure, stablecoins, and crypto settlement rails move from “interesting alternatives” to potential core infrastructure.

The real takeaway from Cloudflare’s numbers

Cloudflare is not a niche startup. It sits in the middle of global internet traffic, which makes its telemetry unusually valuable as a macro signal for how the web is changing. In its latest disclosure, the company highlighted explosive growth in AI agent requests, with one of the most striking data points being year-over-year growth measured in the thousands of percent.

That matters because AI agents do not behave like human users.

A person may visit a site, see an ad, compare a few products, and maybe subscribe. An AI agent does not browse in that way. It queries, verifies, transacts, and leaves. If that becomes the dominant pattern, the old monetization stack of advertising and subscription software starts to look incomplete.

You can review Cloudflare’s latest investor materials here.

Why AI agents break the old internet business model

The traditional internet is built around attention.

Ads work because a human can be persuaded or distracted. Subscriptions work because a human values a service enough to pay for access. But AI agents compress the entire user journey. They do not need to be “reached” in the emotional sense, and they do not consume content for entertainment. They access data, APIs, and services as part of a workflow.

That creates a new monetization question:

How do you charge software that behaves like a machine, not a person?

The answer is unlikely to be another banner ad. More likely, it is some form of machine-to-machine micro-payment, usage-based access, or automated pre-authorization for trusted requests. In other words, the internet may need a native settlement layer for agents.

This is where crypto becomes more than a speculative asset class. It becomes a candidate payment rail.

Why stablecoins fit the agent economy better than legacy rails

If AI agents are going to pay for compute, data, inference, identity checks, or API access in small increments, the payment system needs to support:

• high frequency
• very low fees
• near-instant settlement
• global availability
• programmable rules

That is a poor fit for legacy card rails and a much better fit for blockchain-based value transfer, especially stablecoins.

Stablecoins are already one of the clearest product-market fits in crypto. They are used for cross-border payments, treasury operations, and on-chain liquidity. As the stablecoin supply continues to expand, the infrastructure around them is becoming more important for both crypto-native and non-crypto-native use cases. A useful view of the sector’s growth can be tracked through DeFiLlama’s stablecoin dashboard.

For AI agents, the appeal is obvious: a stablecoin is a programmable dollar that can move globally without needing to wait for the banking system to wake up.

That is especially important if the unit of payment is tiny. Once transactions move from “monthly subscription” to “per request” or “per action,” cost and latency become decisive.

The throughput problem is bigger than most people realize

Valente’s most provocative point was not just that AI agents may pay differently. It was that the scale could be enormous.

If only a small percentage of internet traffic becomes commercially monetizable machine activity, the resulting transaction count could still be far beyond what conventional financial networks are built to handle. The important insight is not the exact number of transactions per second. It is the order of magnitude.

That is why this debate is really about infrastructure throughput.

The current financial stack was designed for humans, businesses, and batch settlement. The AI agent economy demands something closer to instant, continuous, internet-native clearing. That is a very different engineering problem.

For context, blockchain roadmaps have been moving in this direction for years, with scaling upgrades and rollup-centric designs pushing settlement layers to support far more activity than early crypto networks could handle. Ethereum’s long-term scaling direction is outlined on its official roadmap.

If the next wave of internet value flows consists of tiny, frequent, automated payments, then network capacity, finality, and fees stop being abstract blockchain talking points. They become product requirements.

What this means for Layer 1s, Layer 2s, and crypto infrastructure

This is where the “L1 throughput debate” gets a new framing.

For years, the crypto industry has argued over whether a base chain should optimize for security, decentralization, or raw throughput. The AI agent economy adds a practical test: can a settlement network support millions of small-value events without friction?

That does not mean one chain will do everything. In fact, the likely outcome is a layered stack:

• application-specific logic at the edge
• fast execution on L2s or specialized networks
• final settlement on robust base layers
• stablecoins as the unit of account
• wallets and key management as the user’s trust anchor

In that world, the winning infrastructure is the one that makes machine payments feel invisible.

This is also why the next phase of crypto adoption may look less like retail speculation and more like middleware for software payments, identity, and authorization.

The user question: how do you hold value in an agent-native world?

If the future is full of automated payments, then self-custody becomes more important, not less.

A user interacting with stablecoins, on-chain apps, and agent-driven payment flows will need secure signing, clear transaction verification, and reliable key protection. The more automated the financial layer becomes, the more dangerous blind approvals and unsafe signing habits become.

That is where a hardware wallet still makes sense. For users who expect to hold stablecoins or interact with on-chain payment apps in a more automated future, a self-custody setup like OneKey can be a practical base layer for protecting private keys and reviewing transactions before they are signed.

Bottom line

Cloudflare’s earnings call was not just an internet infrastructure story. It was a glimpse into what happens when AI agents become a primary economic actor on the web.

If that trend continues, the internet will need a new payment model built for machines, not humans. And the most likely winners are not the ad networks of the past, but the blockchain and stablecoin systems that can settle value instantly, globally, and at scale.

In that sense, the rise of AI agents may do more than reshape web traffic. It may force the next major upgrade in internet money.