Mizuho Warns Circle Faces Up to 20% Revenue Erosion From AI Agents

A Quiet Threat to USDC’s Reserve Income

Circle Internet Financial, the company behind the USDC stablecoin, is staring down a medium-term revenue problem that has nothing to do with regulation or market volatility. Mizuho, the Japanese investment bank, has flagged a 10-20% downside risk to Circle’s reserve revenue – the income generated by holding the fiat assets that back USDC in circulation. The culprit, according to Mizuho, is the accelerating adoption of AI agents that are increasingly capable of redirecting user funds toward higher-yielding alternatives.

Mizuho cut its price target for Circle in conjunction with the warning.

The concern is structural. Circle’s business model depends heavily on the spread between what it earns on the Treasury bills and cash equivalents backing USDC, and what it pays out – effectively nothing – to USDC holders. As long as billions of dollars sit idle in USDC without seeking yield, Circle captures that interest. AI agents change that dynamic by actively scanning and routing capital on behalf of users, eliminating the passive behavior that Circle’s reserve income relies on.

How AI Agents Undermine the Stablecoin Carry Trade

The mechanics are straightforward. When a person holds USDC, they typically do so for convenience – to trade, transfer, or park funds temporarily in a crypto-native format. That convenience comes at a cost: forfeited yield. In a higher interest rate environment, the gap between holding USDC and holding a money market fund or short-duration Treasury instrument is meaningful. Historically, most retail holders have accepted that tradeoff passively. AI agents are designed to close exactly that kind of gap, automatically and continuously.

Mizuho’s 10-20% medium-term revenue risk estimate reflects how much of Circle’s reserve income could evaporate if a significant portion of USDC’s circulating supply shifts toward yield-bearing fiat instruments. The figure isn’t catastrophic on its own, but it represents a durable compression – not a one-time hit. If AI agent adoption scales as expected across crypto wallets, DeFi interfaces, and financial applications, the share of USDC holdings that remain inert and yield-free could shrink steadily over time.

Circle filed for an IPO earlier in 2024, putting its financials under greater scrutiny than ever before. Reserve revenue is the dominant line in its income statement, which means any sustained pressure on that figure feeds directly into valuation models. Mizuho’s price target cut reflects that sensitivity – investors pricing Circle as a yield-capture business now have to discount the possibility that AI-driven capital efficiency will erode the supply of captive, non-yielding USDC over the medium term.

What Higher-Yielding Fiat Alternatives Actually Mean Here

The phrase “higher-yielding fiat alternatives” in Mizuho’s analysis points to instruments like money market funds, tokenized Treasury products, and short-duration bond vehicles – assets that offer interest income while maintaining dollar stability. Several of these products already exist in tokenized form on-chain, including offerings from BlackRock, Franklin Templeton, and a growing list of fintech entrants. AI agents plugged into DeFi infrastructure could, in theory, treat these products as default destinations for idle capital rather than letting it sit in non-yielding USDC.

This creates an ironic pressure on Circle specifically. USDC’s transparency, regulatory standing, and deep liquidity make it one of the most trusted rails in crypto – but those same qualities make it easy to use as a routing layer rather than a final resting place for capital. An AI agent doesn’t have brand loyalty. It has a yield function. If tokenized T-bills return 4-5% and USDC returns 0%, the optimization path is obvious, and agents running continuously on behalf of users will follow it without hesitation or friction.

The broader stablecoin market faces a version of this challenge, but Circle’s exposure is particularly acute because reserve income is so central to how it monetizes USDC. Tether, by contrast, operates with less transparency and retains a larger share of reserve yield as profit without distributing it to holders – a model that is arguably less exposed to AI-driven yield-seeking behavior precisely because there’s no disclosed gap for an algorithm to exploit. Circle’s cleaner, more auditable structure, built partly to attract institutional users, may also make its revenue model easier for AI agents to route around.

A Price Target Cut With Longer Implications

Mizuho’s decision to lower Circle’s price target isn’t just a reaction to near-term earnings pressure – it reflects a judgment that the structural tailwinds Circle has relied on during the post-2022 rate environment are facing a new headwind that compounds as AI adoption deepens. Circle built a business that thrives when dollar-denominated crypto activity is high and when holders remain indifferent to yield. The first condition still holds. The second is beginning to erode, and it isn’t clear that any product decision Circle makes can fully arrest that erosion once autonomous financial agents become standard infrastructure in the wallets and applications where USDC already circulates.

The 10-20% figure Mizuho cited sits in the medium-term, meaning the risk isn’t immediate – but it’s not speculative either. It’s a forecast grounded in the direction AI tooling is already moving across financial applications, and Circle’s IPO ambitions mean it will have to answer for that exposure in public markets sooner rather than later.