The proliferation of AI agents could bring blockchain scalability back into the spotlight as a key technical issue. Previously, network overloads led to high fees and failures: in 2021, the average fee on Ethereum exceeded $50, and Solana faced significant loads, including millions of requests per second. The situation has since changed thanks to the development of Layer 2 and higher throughput networks. So much so that earlier this year, Vitalik Buterin stated that Ethereum's own blockchain trilemma—the problem of balancing decentralization, security, and scalability—was effectively "solved." However, Bart Smith, CEO of Avalanche Treasury Company, believes that the mass use of autonomous agents in financial operations could once again increase the demands on blockchain infrastructure. He stated this in an interview with The Block at the Avalanche Summit in New York last week.
The first signs of such a change are already noticeable. According to Elliptic data cited by Bloomberg, the number of cryptocurrency transactions involving agents increased by 500% over three months. Separate reports from McKinsey, Mastercard, and JPMorgan independently forecast that consumer spending facilitated by AI agents could reach $5 trillion by 2030. This figure reflects only a part of the potential machine economy, excluding B2B payments, autonomous trading, and low-cost machine-to-machine transactions.
Agent operations differ from regular user transfers primarily in scale and payment size. An AI agent can send thousands of microtransactions to pay for API calls, obtain data, or access computing resources. Visa's panel shows that over the past six months, around $7.5 million in agent payments were processed through the x402 and MPP protocols, distributed across approximately 100 million transactions. The average transfer was $0.0746, and the median was just $0.0086. In comparison, over the previous 12 months, the total stablecoin activity accounted for 18.1 billion transactions worth $85.8 trillion, or about $4740 per transaction.
Related articles
