The impact of artificial intelligence on the digital assets sector is becoming increasingly tangible, but the distribution of benefits from this process will be extremely uneven. My data analysis, based on research work at Fidelity Digital Assets, shows that AI agents can become a new catalyst for activity—from automated payments to complex trading operations and lending. But the key question is not whether they will bring benefits, but who exactly will end up among the beneficiaries.
Development Productivity: A Double Effect
The empirical data I examined, based on telemetry from more than 100,000 developers on GitHub, demonstrates impressive productivity growth. The use of digital assistants for writing code increased the number of commits by up to 180%, and release frequency by 30%. This is critically important for small teams, which can now bring blockchain applications to market faster. However, in the financial sector and other critical systems, manual code review remains mandatory—automation will not replace human oversight here.
Notably, a similar trend is observed within the crypto industry itself. In 2026, despite falling prices, the reduction in the number of developers was accompanied by a slower decline in commit volume. As a result, the average number of changes per developer continued to grow, indicating increased labor efficiency even in a bear market.
On-Chain Activity: A New Frontier for AI
I consider the entry of autonomous AI agents into the on-chain space to be the most promising direction. These algorithms can independently conduct payments, trade assets, provide liquidity, and manage loans. Blockchains are ideally suited for such operations thanks to 24/7 operation, programmable settlements, and support for microtransactions without human involvement.
The market is already beginning to take shape. According to my calculations, by May, AI agents had conducted more than 176 million transactions worth over $73 million, with 98.6% of operations involving the USDC stablecoin. Infrastructure is actively developing: players such as Coinbase have already launched the x402 protocol for automatic payments and the Coinbase for Agents toolkit.
Reality vs. Expectations
Despite the optimism, I do not expect a complete transition of agent operations to public blockchains. Traditional banks and fintech companies, with their established client bases and access to credit products, will develop their own solutions. AI agents will most likely become multiplatform, choosing between systems based on cost and convenience.
A critically important point: the growth in transaction volume does not guarantee a proportional increase in network revenues. My analysis shows that payments, despite their scale, generate significantly less revenue than capital operations. Over the past 180 days, trading at the base layer of Ethereum generated 49 times more revenue per dollar of volume than payments. MEV provides additional margin.
Therefore, I see the greatest potential in AI agents that will trade, lend, and provide liquidity. In the case of mass adoption of automatic payments, the main beneficiaries will be stablecoin issuers and infrastructure providers, rather than holders of native blockchain tokens.
My conclusion: the market is on the verge of structural changes, but investors should be selective. Infrastructure projects and stablecoin issuers look more attractive than universal L1 solutions, unless the latter can secure a dominant position in high-margin segments such as trading and lending.