The headline figure emerged from a live demonstration at Sui Basecamp in Singapore on October 7, where more than 10,000 tunnels handled activity across payments, games and chat applications. Sui said in its official announcement that the test was designed to examine how its infrastructure could support high-frequency interactions between autonomous software agents.
The result was substantially higher than Sui’s previous test. In July, the network reported 6,086,766 transactions per second during an AI-agent livestream experiment. That earlier demonstration also used programmable tunnels, which process transactions away from the main chain before settling activity back to Sui when a tunnel closes, according to Sui’s July explanation.
The distinction matters for investors and developers evaluating what the number represents. Sui’s base layer did not individually execute 40 million transactions per second under ordinary public-network conditions. Instead, the test measured the volume that could move through application-specific channels connected to Sui’s settlement layer.
That makes the result less a conventional layer-1 performance claim than a demonstration of how liquidity and activity might be organized around the blockchain. If autonomous agents begin making large numbers of small payments, requesting services, exchanging data or taking actions inside games, pushing every interaction directly onto a public chain could be expensive and inefficient. Offchain tunnels offer a way to keep those interactions active without requiring every event to compete for inclusion in the base layer.
Throughput is not the same as economic settlement
A raw transaction count is useful for measuring system capacity, but it does not by itself show how much economic value a network can process. The more important question is what happens when activity moves from a benchmark environment into real applications involving money, disputes and security guarantees.
In Sui’s model, tunnels can carry activity during a period of operation and settle the resulting state to mainnet when they close. That architecture could reduce the cost of frequent interactions, particularly when individual actions have low value but occur at high volume. A machine making thousands of small service payments, for example, may not need each payment to be confirmed independently on the base layer if the channel can preserve the relevant state and settle the final result.
The trade-off is that users and applications must trust the tunnel’s rules, operators and settlement process. A channel that handles payments or game assets needs clear guarantees about balances, authorization and what happens if participants disconnect or dispute an outcome. Programmability can make tunnels adaptable, but it also introduces more design choices than a simple transaction sent directly to the mainnet.
That shifts the scaling discussion from a single throughput number to a broader question of where execution, liquidity and risk should sit. The base chain remains important because it provides the eventual settlement path. Yet the user experience may increasingly take place elsewhere, inside channels that are optimized for a particular application or group of agents.
AI agents create a different demand profile
Human users are unlikely to generate millions of blockchain transactions per second. Autonomous software could create a very different workload. Agents may communicate continuously, negotiate with other agents, purchase access to data, trigger automated services or perform actions in digital environments without waiting for a person to approve every step.
The economic significance of that activity would depend on whether agents can control funds and whether the value of each interaction is high enough to justify settlement costs. A high transaction count does not automatically translate into high fees or meaningful capital flows. In some cases, millions of events could represent only small amounts of value. In other cases, rapid machine-to-machine settlement could become a new source of demand for blockchain infrastructure.
Sui is positioning tunnels around that possibility rather than around the needs of conventional retail users. The network’s July experiment established an earlier benchmark, while the October test attempted to show that the architecture could scale as the number of parallel channels increased. Sui’s official Basecamp page describes the October 7-8 Singapore event and the live record attempt using Sui tunnels. The event page provides the broader setting for the demonstration.
For capital allocators, the key signal is therefore not simply that Sui produced a larger number than in July. It is that the project is directing development toward infrastructure capable of absorbing activity that may never be practical on a single shared execution layer. If the AI-agent economy expands, protocols that can combine fast local execution with credible global settlement may compete for that flow.
Independent review will determine the weight of the claim
CertiK participated in the test as an independent auditor and is reviewing the underlying data. A report is expected after the demonstration, according to Sui’s announcement. That review will be important because benchmark results depend heavily on their assumptions, including the transaction format, workload distribution, tunnel configuration and definition of a completed transaction.
Independent reproduction could strengthen the case that Sui’s tunnel design is suitable for large-scale agent workloads. It could also clarify how much of the measured activity involved meaningful state changes, payments or application actions, rather than lightweight events optimized for the test.
Until that information is available, the 40.6 million TPS figure should be read as an architectural demonstration, not as evidence that Sui’s base blockchain can permanently process that volume in normal conditions. The more durable takeaway is that scaling may increasingly involve dividing responsibilities. Offchain channels can handle rapid application activity, while the mainnet provides a final settlement venue and shared security anchor.
That approach may prove useful if autonomous agents begin moving capital through digital markets at machine speed. It will also face competition from other channel systems, rollups and specialized networks. The winners will not necessarily be the platforms with the largest laboratory number. They will be the ones that can turn high throughput into reliable settlement, usable applications and sustained economic activity.
Ethan Brooks is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Ethan Brooks's usual length and in Ethan Brooks's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
Without human check
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.