Sui, a Layer-1 blockchain focused on high transaction throughput, has introduced programmable transaction block technology designed to make blockchain operations more efficient for artificial intelligence agents. The technology will be demonstrated during the project’s Basecamp event, according to reports, as blockchain networks increasingly seek to provide infrastructure for automated AI applications.
The development centers on Sui’s programmable transaction blocks, or PTBs, which allow multiple blockchain operations to be grouped into a single transaction. The technology can execute as many as 1,024 Move function calls within one atomic transaction, allowing AI agents to complete complex sequences of operations without relying on multiple separate transactions.
A PTB treats the bundled operations as one unit. If one part of the transaction fails, the entire operation is reversed, helping prevent incomplete execution. This feature could be particularly important for applications involving financial transactions, automated decision-making and other processes where partial completion could create errors or unexpected outcomes.
Faster execution for AI-driven applications
AI agents can require several blockchain interactions to complete a single task. These may include authentication, retrieving information, applying financial logic, and processing payments. Executing each action separately can add latency and increase transaction overhead.
Sui’s PTB architecture is intended to combine those steps into one atomic process. By reducing the number of individual transactions required, the approach could improve responsiveness while potentially lowering costs for developers operating AI-powered applications on the blockchain.
Sui reports that its network can achieve transaction finality in approximately 400 milliseconds. In controlled testing involving AI agents, the blockchain reportedly reached throughput of up to 6.08 million transactions per second. Such figures represent test conditions rather than guaranteed performance under normal network activity, but they indicate the level of scalability the project is targeting.
High throughput could become increasingly important as AI agents begin handling large numbers of automated requests and micro-transactions. Applications could involve decentralized finance, automated trading, supply chain systems, and other services requiring frequent interaction with blockchain infrastructure.
Atomic execution targets reliability
The atomic design of PTBs is also intended to address reliability concerns. An AI agent executing several dependent actions may produce undesirable results if one operation succeeds while another fails. By treating the entire sequence as a single transaction, Sui’s architecture is designed to ensure that either all required steps are completed or the transaction is reverted.
For developers, the approach could simplify the construction of blockchain-based AI agents by allowing authentication, data retrieval, financial calculations, and payments to be coordinated within a single transaction framework.
As @EmanAbio said, one-txn-at-a-time sequencing doesn’t work for agents.
Agents take the fastest path across the internet, and execute it atomically.
Sui runs that path as one txn. Up to 1,024 operations.
Sui Basecamp is where teams building the rails reveal what’s next. pic.twitter.com/4NmAPUdif7
— Sui (@SuiNetwork) August 20, 2026
That capability could also be relevant to enterprises considering blockchain-based AI systems. Financial applications, in particular, can require strict consistency because an incomplete transaction could expose users or businesses to financial losses.
Blockchain competition for AI applications
The announcement comes as blockchain developers increasingly focus on supporting AI-related workloads. Networks across the industry are seeking to improve transaction speed, scalability and programmability as developers explore autonomous software capable of interacting directly with decentralized applications.
Sui’s approach emphasizes the combination of atomic execution, fast finality and high throughput. While other major blockchain ecosystems are also developing infrastructure for AI applications, the project is positioning PTBs as a way to reduce the technical complexity associated with executing multi-step operations.
The reported test results will need to be evaluated against real-world network conditions, where transaction demand, infrastructure limitations, and application complexity can affect performance. Internal benchmark results therefore may not directly translate into sustained production throughput.
For businesses and developers, however, the technology could provide a more streamlined framework for deploying AI agents capable of interacting with blockchain systems. Sui’s combination of programmable multi-step transactions and atomic execution could help create faster, more reliable, and potentially more cost-efficient infrastructure for the next generation of AI-powered decentralized applications.
The development highlights a broader shift toward blockchain networks becoming execution layers for increasingly autonomous software. As AI agents take on more complex tasks, the ability to process multiple dependent operations quickly and consistently could become an important factor in determining which blockchain platforms gain adoption.