Authors: Sumanth Neppalli, Nishil Jain Source: decentralized Translation: Shan Ouba, Jinse Finance
There are two distinct schools of thought in the cryptocurrency space. As a media outlet, we have the privilege of observing these two perspectives closely. One side believes everything is a market, and pricing is key to achieving transparency; the other firmly believes that cryptocurrency is a superior fintech infrastructure. Our publishing plans flexibly adjust between these two viewpoints because, like all markets, there is no single truth—we are simply integrating all possible models.
In this issue, Sumanth will delve into how a new payment standard is evolving online. In short, the core question is: what would happen if you could pay per article to read? To find the answer, we will go back to the early 1990s to see what happened when AOL tried pricing internet access by the minute; explore how Microsoft priced its SaaS subscriptions; and finally focus on Claude's case of pricing conversations by the amount of text.
In this process, we will explain the nature of the x402 protocol, its key players, and its significance for platforms like Substack. Intelligent agent networks are a growing topic of interest within us. The disconnect between internet business models and user behavior: In 2009, Americans visited an average of over 100 websites per month; today, users open fewer than 30 applications per month, but spend significantly more time on them—from about half an hour a day to nearly five hours. The winners (Amazon, Spotify, Netflix, Google, and Meta) have become aggregators, pooling consumer demand, turning occasional usage into habitual behavior, and pricing these habits through subscription models. This model works because human attention follows a fixed pattern: we mostly watch Netflix at night and shop on Amazon weekly. Amazon Prime memberships, priced at $139 per year, bundle shipping, returns, and streaming services, eliminating the hassle of frequent payments. Now, Amazon is pushing ads to subscribers to boost profit margins, forcing users to either watch ads or pay even more. When aggregators can no longer justify subscriptions, they shift to advertising models, like Google, monetizing attention rather than user intent. The composition of web traffic has drastically changed: bots and automation now account for nearly half of web traffic, largely thanks to the rapid proliferation of artificial intelligence and large language models (LLMs), making bot creation easier and more scalable. Of the dynamic HTTP requests processed by Cloudflare, 60% come from API calls—in other words, machine-to-machine communication now consumes the majority of traffic. Our current pricing models are designed for purely human internet use, but current traffic is primarily machine-driven and sporadic. Subscriptions are based on habitual behavior (listening to Spotify on the way to work, using Slack at work, watching Netflix at night), while advertising models rely on the attention economy (someone scrolling, clicking, considering a purchase). But machines have neither habits nor attention—they only have triggers and task objectives. Content pricing is not only constrained by the market but also depends on the underlying distribution infrastructure. For decades, the music industry sold albums because physical media required bundled sales—the cost of burning one song or twelve songs onto the same CD was almost the same, retailers needed high profit margins, and shelf space was limited. In 2003, when distribution shifted to the internet, iTunes changed its pricing unit to single tracks: any song could be purchased from iTunes on a computer for $0.99 and then synced to an iPod. Single-track streaming improved music discovery efficiency, but it also eroded revenue—most fans only bought hit songs, not filler tracks, leading to a decline in per capita income for many artists. Then, with the advent of the iPhone, the distribution infrastructure changed again. Inexpensive cloud storage, 4G networks, and global content delivery networks (CDNs) made access to any song instant and seamless. With phones always online, users had instant access to a near-infinite music library. Streaming services restructured all music at the access level: all recorded music could be listened to for $9.99 per month. Today, music subscription revenue accounts for over 85% of the music industry's total revenue—a situation Taylor Swift is unhappy about, having been forced to return to Spotify. Enterprise software follows the same logic. Because the product is digital, vendors can charge based on the resources actually used. B2B SaaS vendors typically offer predictable service access on a "per seat" basis, monthly or yearly, and limit functionality through tiered packages (e.g., $50 per user per month, plus $0.001 per API call). Subscriptions cover predictable human usage, while metered billing handles sudden bursts of machine usage. When AWS Lambda runs your functions, you only pay for the resources actually consumed. B2B transactions typically involve bulk orders or high-value purchases, resulting in larger transaction volumes and substantial recurring revenue from a smaller but concentrated customer base. Last year, B2B SaaS revenue reached $500 billion, 20 times that of the music streaming industry. If most consumption today is machine-driven and sporadic, why are we still using the pricing model from 2013? Because our current infrastructure is designed for occasional human choices. Subscriptions became the default choice because a monthly decision was more convenient than a thousand micropayments. This wasn't because cryptocurrency created the underlying infrastructure that enabled micropayments (although that's true), but because the internet itself had grown into a behemoth requiring new usage-based pricing methods. Why Micropayments Failed The dream of paying a few cents for content is as old as the internet itself. In the 1990s, Digital Equipment's Millicent protocol promised sub-cent transactions; Chaum's DigiCash piloted banking; Rivest's PayWord solved the cryptographic problem. Every few years, someone rediscovers this ingenious idea: what if you paid $0.002 per article, $0.01 per song, exactly what the item's actual value would be? They all failed in the same way: humans hate measuring their pleasure. AOL paid a heavy price to understand this in 1995. They charged by the hour for dial-up internet. Objectively cheaper than a fixed subscription for most users, customers resented it because it created a psychological burden. Every minute online felt like a ticking timer, every click accompanied by a tiny cost. People unconsciously perceived every tiny cost as a "loss," even if the amount was small. Every click became a tiny decision: Is this link worth $0.03? In 1996, when AOL switched to unlimited data, usage tripled overnight. People preferred to pay more rather than think more. "Paying precisely based on actual usage" sounds efficient, but for humans, it often means anxiety associated with price tags. Odlyzko, in his 2003 paper "Reasons Against Micropayments," summarized this by saying that people are willing to pay more for fixed-rate plans not because of rationality, but because they crave predictability rather than efficiency. We'd rather pay an extra $30 per month for Netflix than optimize for a $0.99 subscription fee per item. Later attempts (such as Blendle and Google One Pass) to charge $0.25 to $0.99 per article ultimately failed. Unless a large proportion of readers become paying subscribers, the unit economics are not viable, and the user experience creates a cognitive burden.
Subscription Hell

Isn't life just endless trouble? Perhaps the gods also adopted a subscription model for human existence.
If we crave the simplicity of subscription models, why are we now complaining about "subscription hell"? A simple pricing reasoning is: how frequently does the trouble that the product solves occur?
Entertainment needs are infinite. The black line in the chart represents this persistent pain point—an ideal state for both users and companies: a smooth, predictable pain point curve.
This is why Netflix rose from an eccentric DVD mail-order service to the elite FAANG club—it offers an endless supply of content, eliminating billing fatigue. The simplicity of subscription models reshaped the entire entertainment industry. When Hollywood studios saw Netflix's stock price soar, they began reclaiming their own film libraries and building their own subscription empires: Disney+, HBO Max, Paramount+, Peacock, Apple TV+, Lionsgate, and more. This fragmentation of content libraries forced users to purchase more subscriptions: watching anime required a Crunchyroll subscription, watching Pixar movies required a Disney+ subscription, and content viewing became a matter of "portfolio building." Pricing depends on two factors: whether the underlying infrastructure can accurately measure and bill usage, and who must make the decisions each time value is consumed. One-time payments are suitable for rare, unexpected events: buying a book, renting a movie, paying for a consultation. Pain points typically manifest once and then disappear. This model is suitable for scenarios with infrequent tasks and clearly defined value; sometimes, the pain point itself is even desirable—we might yearn for the experience of going to the movies or buying a book at a bookstore. Precisely measuring usage allows pricing to be linked to the unit of work. This is why you wouldn't pay for half a movie (its value is ambiguous). Figma cannot extract a fixed percentage from your monthly output (the value of creation is difficult to quantify). Even if it's not the most profitable approach, monthly billing is easier to implement. Computing resources are different: the cloud allows you to observe usage down to the millisecond. Once AWS can measure execution time with such fine granularity, renting an entire server becomes impractical—the server only starts when needed, and you only pay for it while it's running. Twilio uses the same approach with telecommunications services: one API call, one SMS segment, one charge. Ironically, even in areas where we can perfectly measure usage, we still bill like cable TV. Usage is measured in milliseconds, but money flows through monthly credit card subscriptions, PDF invoices, or prepaid “credit lines.” To achieve this, every vendor makes you go through the same process: create an account, set up OAuth/SSO authentication, generate API key authorization, link a bank card, set a monthly limit, and then pray you don’t overcharge. Some tools require you to pre-load credit lines, while others (like Claude) limit you to a lower-tier model once you reach your limit. Most SaaS products fall into the green “predictable pain point” zone: too frequent for a one-time purchase, and too stable for precise per-use metering. The strategy is tiered plans—you choose a plan that suits your typical monthly usage, and upgrade when you exceed your limit. Microsoft's "1TB storage per user" limit is one example—it differentiates between light and heavy users without measuring every file operation. The CFO limits the number of users who need access to higher-tier plans by assigning permissions. A concise pricing model classification method is a two-dimensional chart: the X-axis represents usage frequency, and the Y-axis represents usage variance (i.e., the degree of fluctuation in usage patterns of an individual user over time). For example, watching two hours of Netflix most nights is low variance; while an AI agent sending 800 API calls in 10 seconds and then stopping is high volatility. The bottom left corner is the one-time payment area: a simple "buyout" pricing model works when tasks are rare and predictable, because you only incur the cost once to continue. The top left corner represents the chaotic "casual browsing network": irregular news binge reading, link jumps, and low willingness to pay. Subscriptions are too cumbersome, while micropay-per-click collapses due to decision-making and transaction friction. Advertising becomes the financing layer, aggregating millions of tiny, inconsistent views. Global advertising revenue has surpassed $1 trillion, with digital advertising accounting for 70%, indicating that a large portion of the internet lies in this low-commitment zone. The bottom right corner represents the ideal area for subscriptions: Slack, Netflix, and Spotify align with human daily habits. Most SaaS products are located here, differentiating heavy and light users through tiered plans. Most products offer freemium plans to encourage users to start using them, then gradually shift their usage patterns from the top left to the bottom right through consistent daily habits. Subscriptions generate approximately $500 billion in annual global revenue. The top right corner represents the heart of the modern internet: LLM queries, proxy operations, serverless burst traffic, API calls, cross-chain transactions, batch jobs, and IoT device communication. Usage is both continuous and volatile. Seat-based fixed fees don't accurately reflect this reality, but they lower the psychological barrier to paying for initial use—light users pay more, heavy users are subsidized, and revenue is disconnected from actual consumption. This is why seat-based products are gradually shifting towards metered systems: retaining basic plans for collaboration and support while charging for heavy use. For example, Dune offers a limited monthly credit limit, with small, simple queries costing little, while larger, longer-running queries consume more credit. Cloud services have made millisecond-level billing for computing, data, and API platforms the norm, and the credits they sell scale with actual workload—revenue is increasingly tied to the smallest unit observable on the network. In 2018, less than 30% of software used usage-based pricing; today, that figure is close to 50%, while subscriptions still dominate at 40%. If spending is gradually shifting towards a consumption-based model, the market is telling us: pricing needs to keep pace with the pace of work. Machines are rapidly becoming the largest consumers on the internet—half of consumers use AI-driven searches, and machines are already creating more content than humans. The problem is that our infrastructure still runs on annual accounts. Once you sign up with a software vendor, you get access to their dashboards, including API keys, prepaid credits, and end-of-month invoices. This works fine for accustomed humans, but it's cumbersome for sporadic software use. In theory, you could set up automatic monthly bills using ACH, UPI, or Venmo, but these methods require batch processing, and their fee structures are unsustainable in sub-cent and high-frequency trading scenarios. This is the significance of cryptocurrency for the internet economy. Stablecoins offer programmable, global, sub-cent-precision payments that can be settled in seconds, operate 24/7, and be held directly by agents rather than being tied to bank interfaces. If usage is event-driven, settlement should follow suit—and cryptocurrency is the first infrastructure capable of truly keeping pace. The essence of the x402 protocol: x402 is an HTTP-compatible payment standard that utilizes the 402 status code reserved for micropayments decades ago. Essentially, x402 is a way for sellers to verify whether a transaction is complete. Sellers who want to accept on-chain, gas-free payments via x402 must connect to service providers such as Coinbase and Thirdweb. Imagine Substack charging $0.50 for a paid article: When you click the "Pay to Read" button, Substack returns a 402 code containing the price, accepted assets (such as USDC), network (such as Base or Solana), and relevant policies, formatted as follows: Your Metamask wallet authorizes the $0.50 payment via a signed message and passes it to the service provider. Service providers record transaction information on the blockchain and notify Substack to unlock the article. Stablecoins simplify the accounting process, allowing for settlement based on network speed and small denominations, eliminating the need for separate accounts with each provider. With x402, you don't need to pre-load five credit accounts, rotate API keys across different environments, or discover quota triggers at 4 AM that cause task failures. Human billing can continue to use the most suitable credit card method, while all sporadic machine-to-machine interactions are handled automatically and cheaply in the background. You can experience this difference in the smart agent checkout process. Imagine you're trying out a new fashion style on the AI fashion chatbot Daydream: Nowadays, the shopping process redirects you to Amazon so you can pay using your saved bank card information; in the world of x402, the agent understands the context, retrieves the merchant's address, and pays directly from your Metamask wallet without leaving the chat interface. What's interesting about x402 is that it's not currently a single entity, but rather composed of layers common in real-world infrastructure. Anyone building an AI agent using the Cloudflare Agent Kit can create a bot that charges per action. Payment giants like Visa and PayPal are also adding x402 as supported infrastructure. QuickNode provides a hands-on guide on how to add an x402 paywall to any endpoint. The direction is clear: unify the "smart agent checkout" functionality at the SDK layer, making x402 a go-to method for agent payment APIs, tools, and even final retail purchases.

Integrating the x402 Protocol
Once the network supports native payments, an obvious question arises: in which areas will it first become widespread? The answer is high-frequency usage scenarios with transaction values below $1—in these scenarios, subscription models would be too expensive for light users (a minimum monthly subscription fee becomes a barrier). As long as blockchain fees are feasible, x402 can settle every request at machine speed with an accuracy of $0.01. Two forces make this transformation imperative: Supply side: The explosive growth of "tokenization" of work—LLM tokens, API calls, vector search, IoT signals. Every meaningful operation on the modern internet has been appended with a tiny, machine-readable unit.
Demand Side: SaaS Pricing Leads to Huge Waste – Approximately 40% of licenses are idle because finance teams prefer pay-per-seat billing (easier to monitor and predict). We measure work at the technical level, but bill humans at the seat level.

Event-native billing with limits is a way to align the two worlds without scaring away buyers. We can set soft limits and ultimately bill at the best price: news sites or developer APIs are billed per use, then automatically refunded to the published daily cap.
Event-native billing with limits is a way to align the two worlds without scaring away buyers. We can set soft limits and ultimately bill at the best price: news sites or developer APIs are billed per use, then automatically refunded to the published daily cap.
If The Economist were to set a price of $0.02 per article, capped at $2 per day, curious readers could browse 180 links without any mental calculation—the agreement would automatically settle to $2 at midnight. This model also applies to developer platforms: news organizations could charge for each LLM crawl to sustain future AI browser revenue; search APIs like Algolia could charge $0.0008 per query, totaling $3 per day. You can already see consumer AI moving in this direction: when you reach Claude's message limit, it doesn't just show "Limit reached, try again next week," but instead offers two options on the same screen: upgrade to a higher subscription plan, or pay per message to complete the current action. What's currently missing is a programmable infrastructure that allows the agent to automatically make the second choice—pay per request, without UI pop-ups, bank cards, or manual upgrades. For most B2B tools, the actual end result is: "Subscription Bottom Line + x402 Burst Billing": Teams retain a base plan tied to the number of users for collaboration, support, and daily backend use; occasional heavy computational needs (build minutes, vector search, image generation) are billed via x402, without requiring a mandatory upgrade to a higher plan. Better network services are also available: Double Zero aims to provide faster, cleaner internet service via dedicated fiber optic cables—route proxy traffic to their network, priced per gigabyte via x402, with clear Service Level Agreements (SLAs) and limits. Proxy traffic requiring low latency for transactions, rendering, or model switching can temporarily switch to the fast lane, paying for specific bursts of demand before switching back to the normal lane. The SaaS industry will accelerate its transition to a usage-based pricing model, but with safeguards in place: Lower customer acquisition and activation costs: Revenue can be generated on the first use, and temporary developers who have never completed the OAuth or card binding process can still pay $0.03 to use the service; agents are more inclined to choose providers that offer immediate payment. Revenue will grow in tandem with actual usage, rather than relying on seat expansion: This will solve the problem of 30%-50% of seats being idle in most enterprises, with core billing shifting to limited, bursty usage scenarios. Pricing will become a competitive advantage at the product level: "Pay an extra $0.002 per request for the fast track," "Half price for bulk mode"—startups can increase revenue through such flexible pricing experiments. Reduced lock-in effect: Vendors can be tried without complex integration and time investment, reducing switching costs. A World Without Ads: Micropayments will not completely eliminate advertising, but rather narrow its application as the only viable model. Advertising will still perform well in "casual intent" scenarios, while x402 will price for scenarios that advertising cannot cover—occasionally, a user might be willing to pay for a high-quality article without subscribing to a monthly plan.
x402 reduces payment friction and may change the industry landscape once it reaches a certain scale:

Substack has 50 million users and a conversion rate of 10%, meaning 5 million subscribers pay approximately $7 per month. When paid subscribers double to 10 million, Substack may generate more revenue from micropayments—lower friction will encourage more casual readers to pay per article, accelerating the revenue growth curve.
Substack has 50 million users and a conversion rate of 10%, meaning 5 million subscribers pay approximately $7 per month. When paid subscribers double to 10 million, Substack may generate more revenue from micropayments—lower friction will encourage more casual readers to pay per article, accelerating the revenue growth curve.
... This logic applies to all sellers of "high-variance, low-frequency" sales: when people use a product occasionally rather than forming a habit, pay-per-use is more natural than a long-term subscription. This is somewhat like my experience playing badminton at my local court: I play two or three times a week, usually with different friends at different courts. Most courts offer monthly memberships, but I prefer not to be tied to a particular court—I like the freedom to choose which court to go to, how often, and to skip courts when I'm tired. Of course, I know this varies from person to person: some people like to consistently go to the nearest court, some like the habitual reinforcement of a subscription, and some might want to share a membership with friends. I can't comment on offline payments, but with x402, this personalized need can be reflected in the digital world. Users can set their own payment preferences through policies, and companies can provide flexible pricing models to adapt to everyone's habits and choices. The real shining scenario for x402 is intelligent agent workflows. If the past decade was about turning humans into logged-in users, the next decade will be about turning agents into paying customers. We're already halfway there: AI routers like Huggingface let you choose from multiple LLMs; OpenAI's Atlas is an AI browser that uses LLMs to perform tasks for you; and x402 fills the missing payment infrastructure gap in this ecosystem—it allows software to settle small bills with other software the instant a task is completed. However, infrastructure alone is not enough to constitute a market. Web2 has built a complete support system around bank card networks: KYC verification by banks, PCI compliance for merchants, dispute resolution by PayPal, card freezing for fraudulent transactions, and refund mechanisms in case of problems. Intelligent agent commerce currently lacks these safeguards. Stablecoins + HTTP 402 allow agents to make payments, but they also remove the built-in recourse that people are used to. What do you do when your shopping agent buys the wrong flight, or your research bot exceeds its data budget? This is precisely the question we'll delve into next: how developers can use x402 without worrying about potential future glitches.