Average 21.2 million TPS, excluding signature verification in single-node Voyager test

Celestia Labs disclosed that its Voyager single-node test sustained an average of 21.2 million TPS and peaked at 21.6 million TPS over a five-hour randomized transfer trial, though transaction signature verification was excluded from the measurement. The test simulated random transfers among roughly 100 million accounts and, per the company, processed about 400 billion transactions in total. The disclosure follows Voyager's October 29 introduction as a payment architecture for AI-agent commerce, with the new technical note clarifying implementation status and the exact scope of what the throughput figures do and do not cover. The result matters because it positions Celestia's data-availability and settlement stack as a candidate backbone for high-volume machine-to-machine payments, but the omission of signature verification means real-world throughput will be lower than the headline number. What to watch next is whether Celestia publishes benchmarks that include signature verification and multi-node conditions, which would give a clearer picture of production-ready capacity.