AI agents may generate significant digital activity without creating the same level of value for public blockchains, Fidelity Digital Assets warned in a new report.
Senior Research Analyst Max Wadington listed this among six risks to the growing AI-crypto thesis, arguing that closed systems operated by large technology and fintech companies could capture much of the activity instead.
Fidelity Questions Where AI Activity Will Go
Fidelity said centralized platforms could have an advantage in performance, costs, user experience and regulatory clarity.
If AI agents stay inside those systems, public blockchains may see less demand than some investors expect.
That would challenge the argument that autonomous software will increasingly rely on blockchain-based payments, identity and settlement.
Some networks are already positioning for that use case. The XRP Ledger has integrated with the x402 payment standard, allowing AI agents to pay for services using XRP and RLUSD.
Grayscale Backs ETH, SOL, WLD and TAO
Fidelity’s caution contrasts with Grayscale’s more bullish view.
Grayscale Head of Research Zach Pandl recently identified Ethereum, Solana, Worldcoin and Bittensor as networks that could benefit from AI adoption across agentic finance, verifiable records and decentralized AI.
The core idea is that autonomous agents may need programmable money and always-on infrastructure that traditional financial systems were not built to provide.
Fidelity does not reject that thesis outright. Instead, it warns that increased usage may not translate directly into higher value for native tokens.
Payments are one example. Stablecoin transaction volumes could grow while most of the economic benefit goes to stablecoin issuers and service providers rather than the base blockchain itself.
Fidelity also warned that AI could make software development easier and cheaper, reducing technical differentiation between networks. In that environment, liquidity, distribution, security and trust may matter more.
The report also highlighted security risks, since AI can lower both the cost of writing code and the cost of finding vulnerabilities.
The key issue is therefore not whether AI agents will use blockchains, but how much economic value public networks and their native tokens will actually capture.
