TL;DR
Solana payment channels are being tested as infrastructure for a growing class of AI-driven applications that need to make frequent, low-value payments without requiring a blockchain settlement for every individual request.
The new system allows an agent to authorize a spending ceiling once, process multiple payments through signed messages, and settle the final amount onchain. This approach addresses a key limitation of per-request payment models, particularly when autonomous software makes hundreds or thousands of paid calls during a single task.
1,000,000 payments per second. Live on Solana today with Payment Channels.
Agentic payments have a friction problem. Every call an agent makes needs a signature and a settlement. It works well enough for a single payment, but it's broken for an agent making hundreds of metered… pic.twitter.com/QetxKZnllY
— Solana Foundation (@SolanaFndn) September 3, 2026
According to the benchmark, developers processed more than 1 million payments per second using 100,000 unique wallets through a payment channel proxy. At that rate, the infrastructure could theoretically handle more than 80 billion payments over 24 hours.
The reported processing cost was approximately $0.000000000776 per payment, although the figure comes from testing rather than ordinary production usage. The benchmark is designed to demonstrate the potential capacity of payment channels under heavy workloads.
The architecture uses an onchain escrow deposit as a spending ceiling. Funds remain non-custodial while an agent conducts transactions through signed messages. Once the activity ends, the consumed amount is settled onchain and unused funds return to the primary wallet.
This structure can be particularly useful for services with unpredictable or highly granular pricing, such as AI inference billed by tokens, computing resources charged by usage, or streaming applications that deliver many small units of value.
Payment channels are designed to work with Solana-based implementations of x402 and the Machine Payments Protocol, commonly known as MPP. Instead of requiring a separate blockchain transaction for every interaction, these systems can use cumulative payment authorizations and settle multiple deliveries together.
Alibaba Cloud is also launching API endpoints with Payment Channels, targeting AI inference as an early use case. The integration allows an agent to authorize spending once and consume services repeatedly without requiring manual approval for each individual request.
The development reflects a broader push to make blockchain payments more compatible with autonomous software. For crypto, the opportunity extends beyond conventional transfers because agents can potentially hold controlled spending limits and interact directly with paid digital services.
