Bitcoin is losing ground to stablecoins in the near term for artificial intelligence agent payments, according to recent research from the Bitcoin Policy Institute and analysis in a Blackrock report on machine-native economies. However, the longer-term winner in this emerging market remains uncertain.
Blackrock's latest paper on "The Machine-Native Economy" focuses primarily on stablecoins as the leading candidates for transactional use in AI agent systems, with only limited mention of bitcoin's potential as a savings technology. The report highlights protocols such as x402, MPP, ACP, AP2, and TAP—backed by companies including Coinbase, Stripe, OpenAI, Google, and Visa—as the infrastructure layer for these payments.
Shifting AI Preferences
Updated research from the Bitcoin Policy Institute shows a significant shift in how AI models view different assets for payments. In March 2024, AI models chose stablecoins for payments in 53% of responses, compared to 36% for bitcoin. By September, stablecoin preference rose to 77%, while bitcoin support dropped to 20%. Bitcoin retained its advantage as a savings technology, though by a narrower margin than before.
The institute tested nine newer AI models in the updated research, with only Grok 4.6 favoring bitcoin as the top payment choice. The different sets of models tested in each study make direct comparisons difficult, but the trend toward stablecoins is clear.
Infrastructure Gap
Data from 402index.io, a directory of paid services for AI agents, shows bitcoin's L402 protocol accounts for only 0.4% of verified payment services compared to 98% for x402. However, these figures may understate L402 adoption, as Coinbase's x402 directory registers services automatically while L402 relies on community lists. Additionally, the Lightning Network can operate under multiple payment protocols, making full adoption figures unclear.
Structural Challenges
Bitcoin advocates face multiple obstacles in the emerging agentic economy. In the Bitcoin Policy Institute's latest research, fiat led as the unit of account for AI payments at 41.4%, compared to 13.8% for bitcoin. Transaction sizes present another challenge: median agentic transactions fall between $0.01 and $0.10, a range where Lightning's cost advantages matter most for payments below one cent, yet stablecoin fees on Base and Tempo are already fractions of a cent.
Institutional preference for permissioned systems and the dominance of centralized marketing from traditional finance and established crypto projects create additional headwinds for bitcoin adoption in AI payments.
Potential Pathways Forward
Bitcoin development continues despite the current disadvantage. Lightning Labs released Wavelength in alpha in July, enabling developers and AI agents to add self-custodial bitcoin payments without operating their own Lightning infrastructure. Lightning solutions provider Amboss Technologies argues that at scale, Lightning would offer advantages in scalability and cost efficiency.
The adoption of a Lightning wallet by a major AI lab or popular open-source project as a default option could significantly shift bitcoin's competitive position. Meanwhile, the agentic economy itself remains in its infancy, currently operating as an extension of existing economic systems rather than creating fundamentally new structures.
While bitcoin appears to be losing the medium-term battle for AI agent payments, the longer-term outcome of the competition among payment rails remains open. Multiple scenarios could still reshape the landscape as the agentic economy develops.
