Solana × Google Cloud: Analysis of the Pay-Per-Use AI Agent API Payment Layer Based on Stablecoins and the Implementation of the x402 Protocol

On May 5, 2026, the Solana Foundation and Google Cloud jointly announced Pay.sh, a payment gateway for AI agents. This platform allows AI agents to pay per request using stablecoins on the Solana network to access core services like Google Cloud’s Gemini, BigQuery, Vertex AI, and more than 50 community API providers, all without registering accounts, API keys, or subscriptions.

This is the industry’s first case of transforming AI agent payments from a “theoretically feasible protocol” into a “directly callable commercial product.” Its significance is comparable to when Stripe embedded credit card payments into the internet—except this time, the initiator of payments is no longer a human.

HTTP 402: The Return After Twenty-Six Years

To understand the origins of Pay.sh, one must revisit a status code that lay dormant for twenty-six years.

In the HTTP standard, status code 402—meaning “Payment Required”—has been reserved since 1997 but was never actually implemented. Prior attempts to settle payments via this status code on websites or APIs all failed.

The turning point came when several structural conditions matured. First, daily settlement volume of stablecoins surpassed $30 billion, and high-performance public blockchains reduced transaction costs to fractions of a cent. Second, AI agents gained the ability to hold private keys autonomously and initiate payments.

The timeline is as follows:

Deconstructing Pay.sh’s Four-Layer Architecture

Pay.sh’s technical architecture can be understood through four layers.

Access Layer: Solana Wallets as Identity. Traditional API calls require developers to register accounts, authenticate, manage API keys, and set billing cycles—processes designed for humans. For AI agents, this is inefficient and costly. Pay.sh replaces the entire authentication system with Solana wallet addresses, so agents do not need Google accounts or rotating credentials.

Routing Layer: API Proxy on Google Cloud. Pay.sh runs as an API proxy on Google Cloud infrastructure, intercepting requests from AI agents, authenticating, billing, and forwarding them. This architecture avoids exposing each API provider’s billing system directly to the proxy, centralizing complexity within the gateway.

Settlement Layer: Open Standard Based on x402 Protocol. Pay.sh adopts the x402 protocol, incubated by Coinbase and maintained by the Linux Foundation, and also supports the Machine Payment Protocol (MPP) developed jointly by Tempo and Stripe. The x402 protocol fully implements the state machine logic of HTTP 402—clients request a resource, the server responds with 402 “Payment Required” along with structured metadata such as price, accepted token types, target wallet address, and settlement network; clients sign payments and reissue requests, with on-chain mediators verifying and delivering content.

Provider Layer: Ecosystem of Over 50 Community API Vendors. Beyond Google Cloud’s core services, Pay.sh has integrated over 50 community API providers across blockchain infrastructure, data services, e-commerce, and communication tools. Confirmed providers include Helius, Alchemy, Dune Analytics, Nansen, The Graph, as well as AI tools like Anthropic Claude Code, OpenAI Codex, OpenClaw, Hermes, and others.

As of March 2026, the x402 protocol has processed over 119 million transactions on the Base network and over 35 million on Solana, with an annualized transaction volume of about $600 million. The protocol itself charges no fees. The Solana network has previously handled 15 million on-chain AI agent payments, accounting for roughly 49% of the agent-to-agent payment market share as of February 2026.

A Tripartite Battle Over AI Payment Infrastructure

Industry discussions about Pay.sh and the x402 protocol focus on three competing narratives.

Machine Payments Are Essential for Scaling AI Agents. Coinbase’s developer lead Erik Reppel explicitly states that AI agents bypass internet advertising—a core business relying on traffic conversion—while protocols like x402 will reshape the economic incentives of the internet. Traditional finance cannot support autonomous transactions: no automated channels, and no way to process sub-dollar payments. Rishin Sharma, head of AI growth at Solana Foundation, emphasized this point in March: “Agents cannot transact via traditional card networks,” and they need a payment system tailored for autonomous software.

The Competition Among Four Major Proxy Payment Protocols Is Not Zero-Sum. Currently, four main proxy payment protocols coexist: x402 (Coinbase + Cloudflare), MPP (Stripe + Tempo), ACP (OpenAI + Stripe), and AP2 (Google). Industry analysis suggests these protocols address different layers of the AI payment stack; they are not mutually exclusive but complementary.

Among these, MPP and x402 are the most directly comparable, both based on the HTTP 402 model. The main difference is that x402 focuses on on-chain stablecoin settlement, while MPP offers a plug-and-play architecture supporting both stablecoins and fiat card payments.

On-Chain Transaction Volumes for x402 Have Experienced Volatility; Real Market Demand Is Still to Be Validated. In December 2025, x402’s daily transaction volume peaked at about 731,000, but by February 2026, it plummeted to around 57,000—a drop of over 92%. This suggests early adoption may have been driven by speculation or testing, with large-scale organic demand still in early stages.

What Exactly Has Pay.sh Changed?

Current reports on Pay.sh tend to over-attribute its significance. The joint announcement itself does not alter the underlying logic of the x402 protocol. The real incremental achievement is that Pay.sh has implemented x402 as a commercial-grade product within Google Cloud’s production environment. It’s not inventing a new protocol but making the open-source protocol configurable, monitorable, and directly usable in production as a gateway.

Part of Its Success: API keys and subscriptions are indeed becoming bottlenecks for scaling AI agents. If agents want to freely call multiple API providers, maintaining individual subscription accounts is impractical. Micro-payments on demand directly address this pain point.

Part to Approach with Caution: While large-scale deployment of AI agents is currently concentrated in a few scenarios, whether API providers will embrace pay-per-use models and abandon subscription revenue remains to be seen, requiring real-world data.

Industry Impact: Why Does This Payment Infrastructure Matter for the Ecosystem?

For the Solana Ecosystem, Pay.sh Is a Foundational Step Toward AI Agent Economy. Solana Foundation’s Chief Product Officer Vibhu Norby stated at the March digital assets summit that 95% to 99% of on-chain transactions in the future will come from AI agents and LLM-based systems. “Agents are cold, precise, computational machines—they are not loyal to any crypto narrative,” and will evaluate the most optimal settlement paths. Solana’s high throughput and low latency give it an advantage in cost assessment. The ongoing expansion of on-chain stablecoins—surpassing $15.58 billion by February 2026—provides liquidity for high-frequency, small-value payments.

For Google Cloud, This Is Cutting-Edge Infrastructure for Monetizing AI Model Services. Traditional subscription models limit API usage because developers must sign high-value contracts beforehand. Pay.sh’s pay-as-you-go model lowers entry barriers and fixed costs, potentially boosting API utilization. In Q1 2026, Google’s general business agreement partners exceeded 100, indicating strategic investment in standardized proxy payments.

For the Industry Infrastructure, Pay.sh Marks the Transition from Protocol Debate to Productization. Machine-to-machine commerce—enabling AI systems to autonomously pay for data, compute, and services—is a trillion-dollar market. DWF Ventures’ April 2026 report estimates AI agents account for about 19% of all on-chain activity—a preliminary figure based on a single-party report, not a definitive statistic. Since 2025, roughly 17,000 agents have been launched, handling about $50 million in transactions.

Meanwhile, the competitive landscape is tightening. TRON DAO has expanded its dedicated AI fund from $100 million to $1 billion, and the upstream open API marketplace for Pay.sh is rapidly growing.

Conclusion

Pay.sh is not just another hastily assembled AI hype in the crypto bubble. Its core logic—using stablecoins and the HTTP 402 standard to challenge traditional API settlement—represents a deep penetration of ledger infrastructure into the software ecosystem. AI agents don’t need login screens or invoices, but they do need a settlement pathway. While Solana and Google Cloud’s investments do not guarantee success, they ensure this pathway is no longer just theoretical.