Artificial intelligence is increasingly penetrating the digital assets space, and its impact is already extending far beyond mere hype. The question is no longer whether AI will be a growth driver, but rather who exactly among market participants will come out ahead. My data analysis, based on research conducted at Fidelity Digital Assets, shows that the distribution of benefits will be highly uneven.
Development is accelerating, but not for everyone
Having examined the activity of more than 100,000 developers on GitHub, I found an impressive effect from the adoption of AI assistants. The number of code commits grew by up to 180%, and release frequency by 30%. This opens doors for small teams, allowing them to launch blockchain applications faster. However, in critical sectors such as finance, manual code review will remain mandatory—automation here will not replace human oversight.
Interestingly, even during the bear market of 2026, when the number of developers declined, productivity per specialist continued to grow. This points to a structural shift, not a temporary trend. But it is important to understand: an increase in the number of applications does not guarantee their success. Users, liquidity, and trust will remain key factors, and as development becomes cheaper, they will become the main battleground.
On-chain activity: a new market is already taking shape
Autonomous AI agents are not just a theory. They are already making payments, trading assets, and providing liquidity. Blockchains are ideally suited for this task thanks to 24/7 operation and the ability to handle microtransactions. According to my data, by May, agents had conducted over 176 million transactions totaling more than $73 million, with 98.6% of them in the USDC stablecoin. This is a clear signal that stablecoins are becoming the primary settlement medium for machines, not just for people.
Infrastructure is actively developing. Coinbase, for example, has already launched the x402 protocol for automatic payments and a set of tools called Coinbase for Agents. However, I am skeptical that all operations will move to public blockchains. Traditional banks and fintech companies, which have a client base and access to credit products, will develop their own solutions. AI agents will most likely be multiplatform, choosing the optimal route based on cost and convenience.
Transactions ≠ revenue: let's look at capital
A key point that many miss: an increase in the number of transactions does not guarantee growth in network revenue. Payments are a low-margin business. They can be bundled, moved off-chain, or shifted to cheap L2 solutions. Far more significant revenue comes from capital operations. My calculation based on data from the last 180 days shows that trading at the base layer of Ethereum generates 49 times more revenue per dollar of volume than payments. MEV also adds additional value.
Ultimately, I see the greatest potential in AI agents that will engage in trading, lending, and liquidity management. At the same time, with the widespread adoption of automatic payments, the main beneficiaries will be stablecoin issuers and infrastructure providers, not holders of native tokens. This is an important signal for investors: not all AI-related projects will prove equally profitable.
My opinion: the market is moving toward a scenario where AI agents will become not just users, but full-fledged economic entities. However, investors should focus on projects with real capital turnover, rather than those that simply generate transactional activity. Otherwise, they may end up with assets that do not deliver the expected returns.