Researchers have systematically reviewed blockchain transaction processing and scalability, proposing a five-layer framework to examine how blockchain can serve as a trusted infrastructure for AI agents. The study highlights key challenges in data management, networking, consensus, execution, and application-level mechanisms, while identifying future directions for building more reliable, auditable, and scalable AI agent ecosystems.
As large language models continue to drive the development of autonomous AI agents, these systems are increasingly capable of interacting with digital environments, calling external tools, managing resources, executing transactions, and collaborating with other agents. However, greater autonomy also introduces fundamental questions of trust: How can an agent’s actions be verified? Can its behavior be audited and reproduced? And who is responsible when a multi-step task fails?
Blockchain technology, with its capabilities for verifiable state changes, auditable records, and coordination among parties that may not fully trust one another, has the potential to become a trusted infrastructure for AI agents. However, this vision depends on whether blockchain systems can provide sufficient scalability, timely finality, reproducible execution, predictable costs, and reliable cross-domain communication under increasingly complex autonomous workloads.
Addressing these challenges, Professor Huawei Huang and his research team from Sun Yat-sen University have conducted a comprehensive survey of blockchain transaction processing and scalability, examining the technological foundations required for blockchain to support the emerging AI agent ecosystem.
“This work provides a cross-layer perspective on blockchain scalability and its role in supporting autonomous AI agents,” explains Professor Huawei Huang. “Rather than considering scalability solely as a performance problem, we examine how it relates to trust, accountability, verifiability, and economic reliability in agent-driven environments.”
The researchers organize existing blockchain scalability research into a five-layer technology framework covering data, network, consensus, execution, and application layers. This framework connects technical scalability challenges with different trust requirements of AI agents.
At the data layer, growing blockchain state can increase storage and historical verification costs. The survey examines approaches for state management, verification, and distributed auditing that can reduce verification overhead while preserving traceable evidence.
At the network layer, sharding and cross-domain communication can improve throughput but may introduce additional challenges, including communication latency, execution inconsistencies, and coordination failures. For AI agents executing complex multi-step workflows, reliable communication must also preserve atomicity, causal consistency, and verifiable failure handling.
At the consensus layer, the timeliness of transaction finality becomes particularly important for autonomous agents whose decisions depend on rapidly changing states. The researchers review recent approaches to consensus optimization and emphasize that throughput should be evaluated together with finality latency, verification costs, and transaction ordering.
At the execution layer, parallel execution can significantly improve performance but may also introduce conflicts, replay issues, and challenges in deterministic execution. The survey highlights the importance of maintaining reproducibility and verifiability as blockchain systems support increasingly concurrent workloads.
At the application layer, even a high-performance blockchain may not provide reliable services if transaction fees, liquidity, or resource allocation remain unpredictable. Mechanisms for transaction scheduling, liquidity provision, and fee management therefore become important for enabling AI agents to complete tasks reliably under changing network conditions.
Beyond organizing existing technologies, the study identifies several open research directions for building blockchain-based trust infrastructure for AI agents.
One key challenge is establishing long-term, verifiable records of agent identity and behavior. Conventional transaction records can verify state changes but may not fully capture an agent’s decision context, tool usage, task delegation, or responsibility boundaries. Future systems may need to combine decentralized identities, verifiable credentials, cryptographic proofs, and on-chain records while balancing auditability with privacy.
Another challenge is reliable cross-chain collaboration. AI agents may observe information on one blockchain, execute actions on another, and rely on off-chain services to complete a single task. Such workflows require stronger guarantees for atomicity, causal consistency, replay protection, and failure handling across different trust and finality models.
The researchers also emphasize the need to move from syntactic verification toward semantic verification. Existing blockchain mechanisms are effective at verifying deterministic execution, but they cannot necessarily determine whether an AI agent correctly understood a natural-language instruction, selected an appropriate tool, or achieved the user’s intended goal. Future verification frameworks will therefore need to connect on-chain execution evidence with higher-level semantic requirements.
Finally, the emergence of learning-based autonomous agents may fundamentally change blockchain economic mechanisms. Agents can rapidly adapt their strategies, coordinate across multiple identities, and interact with markets at high frequency. Designing incentive mechanisms that remain stable and resistant to manipulation in such environments represents an important direction for future research.
The study suggests that blockchain’s role in the AI era should extend beyond simply providing faster transaction processing. To serve as a reliable trust infrastructure for autonomous agents, blockchain systems will need coordinated advances across storage, communication, consensus, execution, and economic mechanisms.
“This survey highlights an important transition in blockchain research—from asking how fast transactions can be processed to asking whether blockchain systems can provide trustworthy foundations for autonomous digital actors,” says Professor Huawei Huang. “Addressing these challenges will require closer integration between blockchain systems, artificial intelligence, cryptography, and distributed computing.”
The study provides a comprehensive roadmap for future research on blockchain scalability and AI agents, offering researchers a unified framework for understanding how scalable blockchain technologies can support trustworthy, auditable, and reliable autonomous systems.
This paper, “A Survey of Blockchain Transaction Processing and Scalability: Toward a Trust Infrastructure for AI Agents,” was published in Blockchain .
Liao J, Chen Q, Zheng J, Tang X, Hu F, Huang H. A survey of blockchain transaction processing and scalability: toward a trust infrastructure for AI agents. Blockchain , 2026, 1(1): 0005. https://www.elspub.com/doi/10.55092/blockchain20260005
Blockchain
Survey
Not applicable
A survey of blockchain transaction processing and scalability: toward a trust infrastructure for AI agents
28-Jul-2026