Mizuho cut its Circle Internet Group price target to $43 from $45 and kept its Underperform rating. The core of the note is not the modest cut itself but the risk that AI assistants steer users away from USDC, shrinking the reserve income that Circle earns on the balances it holds.
🔑 Key takeaways
- Price target cut from $45 to $43, with the Underperform rating maintained.
- Mizuho sees a 10-20% downside risk to Circle’s reserve revenue over the medium term.
- USDC market capitalization could fall by $4-9 billion under faster AI adoption.
- Shares fell about 7% on October 26, ahead of third-quarter results due November 4.
What Mizuho is modeling: AI agents redirecting cash
Mizuho simulated how AI agents, including the Muse agent, would answer a simple question: where should a user keep idle cash? In the scenarios tested, the agents most often recommended moving away from USDC toward fiat-denominated products that offer higher yields and carry Federal Deposit Insurance Corporation (FDIC) coverage, such as savings accounts at SoFi. The more striking finding is that even users who stay within crypto were steered elsewhere, with the simulated agents “appearing to prefer other platforms” over keeping USDC inside Circle’s ecosystem.
A quantified scenario
In a faster-adoption case, the bank estimates that roughly $7 billion would leave USDC for fiat products, and $2 billion for stablecoins offering higher yields. USDC market capitalization could fall by up to $9 billion over the medium term, and Circle’s reserve revenue by 18%. Taken together, these estimates point to a 10-20% downside risk to reserve income.
| Indicator | Mizuho estimate |
|---|---|
| Migration to fiat products | about $7B |
| Migration to higher-yield stablecoins | about $2B |
| Decline in USDC market capitalization | $4-9B |
| Decline in reserve revenue | up to 18% |
| Downside range for reserve revenue | 10-20% |
“the simulated agents appear to prefer other platforms over keeping USDC within Circle’s ecosystem”
Mizuho, research note dated October 26, 2026
Reserve income remains the core of the model
Circle earns most of its revenue from interest generated by the reserves backing USDC, which are invested mainly in U.S. Treasury bills. In the second quarter of 2026, 95.2% of its revenue came from that interest. The model is therefore sensitive to two variables: the average amount of USDC in circulation and the level of interest rates.
The distribution of holders matters just as much. About 17% of USDC was held on Circle’s own platform, where holders earn no yield. About 30% sat on Coinbase, and the remaining 53% on other platforms. A large share of the balances therefore already generates no income for Circle. Coinbase, the main distribution partner, also captures a portion of reserve income: according to TIKR, it received about $1.4 billion of Circle’s $2.6 billion in reserve revenue in 2025, or roughly 54%.
| Indicator | Before | After (Mizuho) |
|---|---|---|
| Price target | $45 | $43 |
| Average USDC circulation, generalized AI adoption scenario, 2027 | $108B | $101B |
| Reserve income, 2027 | $4.3B | $4.1B |
| Adjusted EBITDA, 2027 | $635M | $603M |
The revisions look measured: average USDC circulation falls by about 6.5%, reserve income by 5% and adjusted EBITDA by roughly 5%. Still, according to Investing.com, the new EBITDA estimate sits about 30% below analyst consensus.
A limited adjustment on top of an already bearish path
Mizuho presents the revision as a limited change: the target fell only $2 and 2027 EBITDA was trimmed by about 5%. The context matters, however. The bank had already downgraded Circle to Underperform in July 2026, when its 2027 EBITDA estimate sat roughly 25% below consensus. TIKR concludes that most of Mizuho’s bearish case predates its note on AI agents.
A correction is also needed for some coverage. One Investing.com headline described a target raise to $43. That wording is wrong: the article body and the other sources all report a cut from $45.
Usage data points in the opposite direction
Circle’s own usage metrics tell a different story. As of October 16, the company said USDC accounted for roughly 77% of global stablecoin transaction volume in June, up from 60% a year earlier. In the first half of 2026, more than $36 trillion of on-chain transactions settled in USDC. TIKR also notes that USDC represented 99.3% of settlement volume on the x402 protocol in July 2026.
Transactional adoption does not translate mechanically into balances. Payments in USDC and passive holding of cash are distinct uses, and only the second feeds reserve income directly. That is exactly the link Mizuho considers at risk. Circulation has also been broadly stable: about $73 billion on October 8, against $73.3 billion on June 30, a decline of $300 million.
“banks retain some influence over the pace of this adoption through the pricing of deposits”
Bank for International Settlements, working paper
The competitive debate therefore extends to bank deposits. Referrals by AI agents to FDIC-insured, interest-bearing accounts add a competitive channel that traditional stablecoin distribution models did not have.
Market reaction and the earnings calendar
The stock closed at $79.84 on October 26, down 7%, according to The Block. The percentage decline to the $43 target depends on the reference price each publication uses.
| Reference price | Price | Implied decline to $43 |
|---|---|---|
| Friday close (TIKR) | $85.91 | about -50% |
| Friday close (Stocktwits) | $85.00 | about -49.4% |
| October 26 close (The Block) | $79.84 | about -46% |
Over twelve months the stock lost more than 33% according to Stocktwits, after an adjusted closing high of $103 in early September. The forward earnings multiple moved from about 66 times to 62 times, against an annualized average of 75 times and a low of 45 times in August. On consensus numbers, the $43 target implied about 33 times forward earnings. Coinbase fell 5% the same day, since its share of reserve income moves directly with USDC balances.
Circle reports third-quarter results on November 4, before the market opens. Consensus expects revenue of $787.70 million, up from $701.32 million in the previous quarter, an increase of about 12%. Adjusted earnings are expected at $0.42 per share against $0.26, roughly 62% higher on the raw figures, although Stocktwits cites about 65%. The report will cover balances through September, before the October decline, so management’s comments on that month will be decisive. In the first half, Circle had reported about $1.4 billion in revenue and kept roughly 7 cents of net profit per dollar of revenue.
Outlook and scenarios
The debate comes down to one question: will a holder of non-yielding USDC stay put when an AI agent offers an interest-bearing, insured account? Mizuho answers with caution, while usage data point to strong adoption of USDC for payments.
Three scenarios stand out. In a status quo case, growth in on-chain payments offsets balance outflows and reserve income stays close to consensus. In an intermediate case, a partial migration of roughly $4 billion shrinks the revenue base without calling the model into question. In Mizuho’s downside case, faster AI adoption reduces market capitalization by up to $9 billion and reserve revenue by 18%, which keeps the downside range at 10-20%.
The November 4 results will be the first concrete test. The trajectory of USDC balances, the level of reserve yields and management’s comments on October will show whether the slowdown is a one-off or a structural shift in Circle’s model.
Sources
- The Block, Mizuho cuts Circle price target (October 27, 2026)
- Web3 News (Kordis Labs), Circle faces up to 20% revenue risk from AI agents
- TipRanks (TheFly), Mizuho cuts Circle Internet target
- Stocktwits, CRCL stock in spotlight after Mizuho cuts price target
- Investing.com, Mizuho Circle Internet price target and AI risk assessment
- TIKR, Circle sinks 7% as Mizuho warns of risk from AI agents
This article is published for information and educational purposes only. It does not constitute investment advice in any way. Do your own research (DYOR) before making any decision.
