Maybe we all misunderstood the original meaning of x402

Author: Charlie Liu

If you’re involved in fintech, crypto, or AI circles, you’ve certainly seen the term “x402” flooding your feeds.

A slew of internet giants have entered the scene in the agentic payments “intelligent agent payment” track: Coinbase and Cloudflare jointly launched the x402 Open Foundation, Google integrated encryption extensions into its AP2 standard, and even the traditionally steady Adyen rolled up its sleeves to compete. Visa’s TAP promises to achieve interoperability, while Stripe and OpenAI have teamed up to push ACP and “instant checkout” as industry standards.

An almost forgotten internet status code — 402 “Payment Required” — has unexpectedly become the key to determining the next-generation AI agent business model.

In fact, within the payments industry, this undercurrent has been flowing for over half a year.

Back in February, I wrote on Substack “AI Agents: The Next Frontier Reshaping the Future of Payments,” noting that star projects like Skyfire and Payman—seed-stage ventures raising tens of millions of dollars—are backed by Web2 and Web3 payment giants like Visa and Coinbase, and are actively positioning themselves in this space.

Moreover, companies like Stripe, Visa, and PayPal mentioned the prospects of AI agents combined with payments in their annual reports, albeit briefly.

Recently, however, the term x402 has truly come into the spotlight, thanks to major moves by AI giants and fintech heavyweights. The x402 itself and the broader trend of agent business models have finally caught the attention of the wider tech and investment communities.

Many comments point to its origin from the HTTP 402 status code, analyzing how emerging AI agents and mainstream crypto tech are fueling this topic.

But they largely overlook a fundamental driving force: What is the real motivation behind this AI + crypto wave?

The Starting Point: AI Companies Facing Survival Crises

On the surface, everyone talks about “AI agents finally being able to pay autonomously,” imagining what new skills future AI agents might unlock.

But from a closed-door discussion with a prominent Silicon Valley investor, I learned that the root cause—what truly drives this—is the survival pressure on large models vendors like OpenAI and Anthropic.

While they tout a future worth trillions to capital markets, they face mounting copyright lawsuits and rising public concerns over the ethics of their training data. Without a scalable, automated way to compensate for the data and resources they scrape, future funding and massive capital expenditures will be under threat.

One story I heard is that OpenAI’s Sam Altman approached Coinbase’s Brian Armstrong—both alumni of YC. This might explain why Coinbase briefly launched AgentKit earlier this year, a developer toolset supporting the OpenAI SDK, enabling agents to perform trades.

But AgentKit didn’t last long—soon it was offline. Not because intuition was wrong, but because their strategic paths diverged, their company focuses quickly went in opposite directions.

Perhaps under pressure from collaborations like Perplexity and Shopify, OpenAI shifted its growth strategy toward consumer e-commerce—becoming the interface for discovering and purchasing products—favoring a checkout method that merchants can adopt without major upheaval.

Meanwhile, Coinbase, influenced by the GENIUS Act and a series of crypto enforcement actions in the US, gradually turned its ambitions toward internet monetary infrastructure—pushing toward “machine-to-machine” traffic, where no accounts are needed, and bots pay directly via interfaces or pages.

Through First Principles: The Four Major Camps Formed by Giants

Although AgentKit was short-lived, it laid the groundwork for Coinbase’s subsequent upgraded version—x402—and opened new avenues for other major players.

It’s precisely because of the core driving forces mentioned above that understanding the formation of these camps through first principles offers better insight into how the giants’ alliances and competitions are unfolding:

Stripe × OpenAI: Capture consumer scenarios, perform “seamless upgrades”

OpenAI and Stripe jointly built the ACP protocol and introduced instant checkout in ChatGPT. Essentially, this is an “elegant upgrade” to existing payment networks.

Stripe issues shared payment tokens for specific orders and merchants, which agents carry out transactions with, while merchants continue using familiar fraud prevention, refunds, and tax processes—without restructuring teams or retraining staff.

Etsy has already integrated this, Shopify is in line, and the core of this strategy is to leverage existing channels and habits: seamlessly embed agent transactions into established merchant payment pipelines.

Coinbase × Cloudflare: Target the machine economy, do “underlying reconstruction”

Coinbase’s upgraded x402 aims to be more radical. It makes the HTTP “402 Payment Required” status code executable: servers declare prices and accepted tokens, clients complete payments (preferably using their own USDC on Base chain), and continue operations with proof—no accounts, login, or monthly bills needed.

This scheme excels in micro-payments of $0.01–$0.1—for API calls, article context retrieval, or tool usage—similar to the early demo by Skyfire, invested in by Coinbase.

Cloudflare’s previous “pay-per-crawl” model paved the way, and Coinbase’s strong entry has united key players like AWS, Circle/USDC, NEAR. Notably, Anthropic—considered a perfect stand-in for OpenAI—is part of this alliance.

Compared to OpenAI’s consumer shopping direction, Anthropic’s core MCP framework focuses on enabling agents to discover and invoke tools per request. Its economic model aligns naturally with x402’s “pay-per-request,” far better suited than traditional store-front billing models.

Google × Adyen: Set standards, ensure compliance and auditability

Google’s AP2 standard addresses the most challenging aspect of payment systems—authorization. It uses a set of signed authorization instructions, binding user intents (like price caps, frequency limits, category ranges) with specific operations. This design is inherently channel-agnostic and includes encryption extensions, allowing agents to handle micro-payments seamlessly via the x402 link while maintaining complete audit trails.

Additionally, Adyen’s involvement means this system is equipped from day one with enterprise-grade dispute resolution and compliance capabilities.

Visa TAP: No new track, but inclusive and compatible

Visa’s TAP doesn’t try to reinvent payment rails. Instead, it plays a key compatibility role—helping issuing banks and acquiring banks recognize agent transactions and apply unified risk and dispute rules.

Its core positioning is interoperability—TAP explicitly aligns with the Stripe+OpenAI camp’s ACP and connects with Coinbase+Cloudflare’s x402, truly implementing its Network of Networks concept, ensuring a dominant position.

With Mastercard’s “Agent Pay,” card brands are guiding standard convergence rather than erecting barriers around agent business.

Though these schemes seem independent, in practice they are forming a layered, collaborative, mutually reinforcing ecosystem:

1. Authorization Layer: Google+Adyen’s AP2 standard provides verifiable user intent credentials—serving as “passports” for all operations.
2. Execution Layer:
   • For consumer shopping, use OpenAI+Stripe’s ACP + traditional card networks for a smooth experience.
   • For machine-to-machine micro-payments, use Coinbase+Cloudflare’s x402 + cryptocurrencies for ultra-low-cost transactions.
3. Coordination Layer: Visa’s TAP and others behind the scenes inform banks “this is an agent transaction, not fraud,” ensuring smooth processes.

A typical scenario might be: your travel agent first uses x402 to query multiple airlines for prices (authorized via AP2), then upon confirmation switches to the ACP channel to purchase tickets, with TAP informing the bank. Same task, two payment tracks, one audit log.

In this future system, interoperability is the core—this is why at last week’s Fed Payments Tech conference, industry participants emphasized interoperability as the most critical future payment system feature.

Who Will Win the Big Race of Agent Payments?

Existing giants have won the first round—since identity verification, fraud prevention, refunds, taxation, and dispute resolution are all scale games. ACP/TAP/AP2 directly connect to mature systems that reassure CFOs.

But returning to the original motivation: OpenAI aims to pay content owners. In fact, a fully interconnected agent economy will also benefit the “long tail” participants.

Long-tail players can succeed in scenarios where tiny amounts, previously hampered by cumbersome processes, now thrive: per-request APIs, article-based context retrieval, per-use tools.

Many startups are working in value-added layers beyond the three basic levels—such as developing cross-protocol identity + payment solutions that enable small teams to implement metered billing without establishing a dedicated billing department (this sounds very much like Stripe’s original value proposition, doesn’t it?).

Further, stepping outside the online space—when we realize that AI’s biggest limitations are rooted in the real world—I believe there are at least three domains where agent + x402/AP2/ACP/TAP will significantly change economic models:

1. Prediction markets and oracle revolution: If DeFi oracles adopt AP2 authorization + x402 payments, data sources will be incentivized to provide more real-time, higher-quality data—fundamentally boosting market efficiency and liquidity, opening larger possibilities for hot prediction markets.
2. AI data centers and grid resilience: Fluctuations in training compute demand are intense. Edge storage devices and EV fleets could sell excess power to data centers per second via x402, transforming rough “power rationing” into precise “auction markets.”
3. Climate data and DePIN: Thousands of sensors could become data centers, selling air quality, temperature, flood levels, etc., via x402 per event to municipal agencies or insurers. AP2 authorization ensures data use is within scope, and all transactions are auditable.

Summary: Victory of Division of Labor

Looking back, OpenAI and others initially just wanted to solve the “copyright original sin”—pay fair prices for data used.

But the ultimate answer is not a single technological breakthrough, but a clever division of industry roles:

• OpenAI + Stripe’s ACP makes large-scale, compliant consumer transactions possible.
• Coinbase + Cloudflare’s x402 provides infrastructure for fragmented machine economy payments.
• Google + Adyen’s AP2 establishes trusted authorization and auditing standards.
• Visa’s TAP alleviates concerns from traditional finance.

Coinbase’s deep tie-in with Anthropic is because MCP and x402’s microeconomic models are inherently compatible; OpenAI’s focus on consumer scenarios is driven by the potential for rapid growth and trust-building. Both start from the same pain point but choose different battlegrounds, ultimately weaving a more resilient ecosystem.

For startups, value-added layers and services beyond this “authorization–execution–coordination” core framework are excellent entry points.

Even outside online domains, there are extraordinary opportunities—predictive markets and oracles reflecting real-world data, AI data center energy demands, weather data for climate disaster risk—all are significant and promising fields.