AI agents choose Bitcoin as savings: this is a structural change, not a speculative theme

Why Do AI Agents Keep Choosing Bitcoin?

The reason BPI’s research has gained so much attention is because the results are so consistent: 79.1% of models favor Bitcoin for “long-term store of value.” It’s not just about crypto assets in general, but Bitcoin itself. The related tweets have been viewed 437,000 times, and leading crypto accounts have shared them widely. The discussion focus has shifted—from “AI and crypto might be related”—to “AI agents are converging on Bitcoin’s scarcity.”

This happened on March 3, 2026, when the market fear index was only 13, and BTC was around $68,000. On-chain signals—NVT around 23.3, neutral funding rates—indicate the market hasn’t fully priced in this long-term implication.

• Market opinions are clearly divided: bulls see this as further validation of the “hard currency” paradigm; skeptics question whether the prompt design and research methods are biased.
• Jeff Park from Bitwise pointed out an interesting point: stablecoins have freezing risks, which explains why none of the 36 models prioritize fiat currency.
• Outlets like Cointelegraph and Forbes are following up, discussing whether infrastructure investments will accelerate. But honestly, the pace is exaggerated. A more realistic expectation is that developers will gradually build out AI-driven Bitcoin demand over the next few years, rather than a capital surge happening in a few months.

I suggest ignoring one thing: all the discussions about whether “this tweet can pump the price this week.” Current volatility mainly depends on macro liquidity, which no single study can influence. AI adoption is a multi-year process, not a short-term signal.

Overlooked Rotation: Using Bitcoin for Savings, Stablecoins for Payments

Looking deeper, this is a clear structural rotation: Bitcoin is associated with savings and reserves, while stablecoins are linked to payments and settlement. This aligns with human asset usage habits—just systematized and scaled through autonomous agents.

David Zell from BPI reminds us: don’t over-interpret market forecasts; models mainly reflect training data distribution. But the consistency is notable—Anthropic series shows about 68% preference for Bitcoin, OpenAI series about 26%. Despite differences, the overall direction is the same.

On-chain pricing signals also support this: MVRV around 1.25, NUPL around 0.2—these suggest hope rather than euphoria. If the narrative that “AI economy needs neutral reserves” continues to be validated, we might see accumulation from whales later. But don’t treat this as a short-term indicator; the core issue is that the market hasn’t fully digested the meaning of “Bitcoin as an AI-neutral reserve asset.”

Key points:

• AI’s preference for Bitcoin as a reserve is a slow-changing variable that will reshape capital allocation over years.
• Current on-chain signals point to “undervalued” rather than “overheated,” favoring patient builders and long-term holders.
• Funds are likely to flow first into infrastructure (Lightning, self-custody tools) and developer ecosystems, not immediately boosting secondary market prices.

Bottom line: If you treat this as a short-term theme, you’re missing the main point. The advantage lies with builders and patient holders who understand the “AI—Bitcoin neutral reserve” logic. On-chain metrics show valuation remains attractive; long-term allocation during volatility is more rational than chasing short-term catalysts.

Conclusion: You are still in the “early but not blank” phase. The biggest beneficiaries are infrastructure builders and long-term holders, followed by institutional funds able to add on dips; short-term traders have little edge.