Sui hits 40 million TPS through Pilotfish and tunnels

Sui hits 40 million TPS

Sui tunnels processed 40,614,180 transactions per second during a high-stress speed test at the Marina Bay Sands Basecamp event in October 2026. This result exceeds the 6,086,766 TPS record Sui set on July 4, 2026. The test, conducted by Adeniyi Abiodun, Chief Product Officer and Co-Founder at Mysten Labs, proved the network resilience for AI agents. CertiK will examine the execution logs and supporting evidence to verify the results. Pilotfish, a prototype extension, increases throughput in direct proportion to the number of machines used. If you monitor validator costs, the Pilotfish architecture allows the use of smaller, cheaper servers on demand. Pilotfish partitions the validation function into the Primary, which sequences transactions using high-throughput consensus, the SequencingWorkers, which store and dispatch transactions, and the ExecutionWorkers, which store blockchain state and execute transactions. This approach allowed Sui to reach 65.8 million transaction executions in a single day following its mainnet launch. During the Basecamp event, more than 10,000 tunnels opened on Sui mainnet within seconds of each other. Users and AI agents used these tunnels to conduct activities in categories ranging from chat to games.

Comparison of execution engines

Aptos implements Block-STM to execute transactions. The engine uses Software Transactional Memory, optimistic execution, and multi-version memory. Block-STM v2 became the default on the Aptos mainnet on October 15, 2026. While Aptos can execute over 160,000 non-trivial Move transactions per second, Sui’s tunnel throughput reached much higher levels using its programmable, offchain execution model that settles to the mainnet upon closure. Aptos engineers use the Block-STM algorithm to manage all conflicts internally and automatically adapt to the workload. Aptos relies on the Rayon, Dashmap, and ArcSwap crates for concurrency. Under low contention, Block-STM achieves a 16x speedup with 32 threads. Under high contention, it achieves over 8x speedup. Block-STM uses a collaborative scheduler and atomic counters to manage tasks. This avoids the synchronization overhead found in priority queues. The engine uses a multi-version data structure to avoid write-write conflicts. All writes to the same location include versions containing transaction IDs and the number of re-executions. When a transaction reads a memory location, it obtains the value from the highest transaction that appears before it in the preset order. Block-STM uses a counting-based approach to manage tasks by leveraging the preset order of transactions. This method allows threads to grab tasks by fetch-and-incrementing the counter with a lower index. Block-STM also uses two additional atomic counters to verify that execution and validation indices reach the block size. If these conditions are met and no ongoing tasks exist, the entire block is safely committed.

Stablecoin volume and market trends

Sui implemented 0-fee transfers for seven stablecoins including USDC, USDY, and FDUSD. The total stablecoin capitalization reached $460 million, where USDC accounts for approximately 65%. This equals less than $400 million of total USDC market capitalization. Since August 2025, the network facilitated over $1 trillion in stablecoin volume. Within five days of launching gasless stablecoin transfers, the volume for these transfers reached $65 billion, though some activity came from arbitrage bots. The total ecosystem TVL exceeds $1 billion. SUI trades around $0.75, with resistance at $0.84 and support at $0.66. SUI dominance decreased from 0.1406% to 0.1388% over the month. The Fear and Greed Index rose from 22 to 34 points. How long can Sui maintain these massive throughput spikes during sustained agentic commerce?