A study by the Bitcoin Policy Institute found that AI models, when given economic autonomy, overwhelmingly prefer Bitcoin and digital assets over traditional fiat currency, a finding with direct consequences for how companies build financial infrastructure around autonomous systems.
The research tested 36 models from six providers, including Google, Anthropic, and OpenAI, across 9,072 neutral monetary scenarios. Bitcoin emerged as the top choice in 48.3 percent of all responses. More telling: not a single model out of the 36 selected fiat currency as its top preference, with over 90 percent of responses favouring digitally-native money.
A Two-Tier Logic
The models did not treat all digital assets the same. Without any prompting, they defaulted to a clear functional split between saving and spending. For long-term value preservation, Bitcoin dominated at 79.1 percent. For everyday transactions, stablecoins captured 53.2 percent of preferences, finishing second overall across all scenarios at 33.2 percent.
The pattern reflects a rational internal logic. Bitcoin serves as a hedge against long-term debasement and counterparty risk. Stablecoins handle the operational layer, enabling instant, programmatic payments that sidestep the settlement delays and currency conversion fees embedded in traditional banking rails.
A supply chain agent paying international freight vendors illustrates the point. On legacy fiat infrastructure, weekend settlement gaps and conversion costs create friction. On stablecoin rails, the same payments execute instantly. The treasury holding the capital base, meanwhile, stores value in Bitcoin.
Provider Choice Shapes Financial Behavior
Bitcoin preference varied sharply by model provider. Anthropic’s Claude Opus 4.5 selected Bitcoin in 91.3 percent of relevant responses. OpenAI’s GPT-5.2 did so in only 18.3 percent. The gap is not trivial. An organisation deploying an AI model for automated portfolio management is also, implicitly, deploying the financial biases baked into that model’s training and alignment methodology.
That variance forces a more deliberate vendor selection process. Choosing an AI provider is no longer purely a capability decision. It directly shapes how autonomous agents assess risk and allocate capital.
Compute as Currency
One finding sits outside conventional finance entirely. Across 86 separate responses, models independently proposed using compute units or energy, specifically GPU-hours and kilowatt-hours, as a method to price goods and services. No prompt suggested this. The models arrived there on their own.
Managing that kind of abstract value exchange requires a level of data maturity most organisations have not yet built. It is early, but the signal is worth tracking.
Infrastructure Implications
The research points to concrete steps for technology leaders. Stablecoin settlement integrations for lower-risk vendor payments represent a practical starting point. Beyond that, the study identifies growing demand for AI agent-native Bitcoin payment infrastructure, self-custody solutions, and Lightning Network integration.
Because these models consistently prefer open, permissionless networks, organisations that rely solely on traditional banking infrastructure will constrain what their autonomous systems can actually do. Building compliant gateways to digital asset networks now positions those platforms to support the next generation of AI-driven commerce without retrofitting later.
This article is a curated summary based on third-party sources. Source: Read the original article
