Coinbase Renewed Through 2029, USDC Circulation Shrank, and Every Dollar of the Guidance Raise Is Arc Token Revenue

Results vs. Consensus

The consensus range matters more than any midpoint here. Providers entering the print carried revenue anywhere from $717M to $742M, a $25M spread on a single quarter, because neither the reserve return rate nor average float is guided. On EPS the dispersion was narrower and the actual landed at the top of it. A company can miss revenue by 5% and hit EPS exactly when 95% of the revenue line is reserve income and the cost base below it is largely contractual.

Year-over-Year Comparison

The year-over-year comparison flatters Q2 for a reason that has nothing to do with the business. Q2 2025 was the IPO quarter, carrying $435.0M of stock-based compensation and a $167.7M fair-value charge on convertible debt and related instruments. That is what turns a $482.1M prior-year net loss into a $530M "improvement" and a 56% decline in GAAP operating expenses. Neither tells you anything about operating momentum. The lines that do are the adjusted ones, and there the picture is that revenue grew 7%, adjusted operating expenses grew 23%, and adjusted EBITDA grew 8% while its margin fell 329bps. Also worth noting: the diluted share count is up 150% against the pre-IPO base, so per-share progress lags the aggregate.

Sequential (Q/Q) Comparison

Quality of Beat/Miss

Revenue (missed, and the composition of the miss is worse than the size of it). Total revenue grew 1.0% sequentially, which sounds like stabilization until you look at what moved. Reserve income rose $15.2M on an all-time-high average float and a reserve return rate that finally stopped falling, holding at roughly 3.5% after a year of compression that took it down 66bps Y/Y. That is the good half. The bad half is that other revenue, the line carrying the entire platform-diversification thesis, fell $8.0M or 19.3% sequentially. Management attributed the decline to two things: moderating blockchain-partnership revenue in weak digital-asset markets, and a deliberate choice to redirect resources toward Arc rather than sign additional chain partnerships. The first is cyclical. The second is a strategic bet that trades a proven, if lumpy, revenue line for an unproven one. Both are defensible. Neither changes the fact that the diversification line went backwards in the quarter it was supposed to prove itself.

Margins (the headline is a small decline, the composition is genuinely encouraging, and the forward setup is the problem). RLDC margin slipped 21bps to 41.2%, ending a three-quarter expansion streak. But underneath it, net reserve margin, which measures reserve income less distribution and transaction costs as a share of reserve income, rose to 39% from 38% and has climbed from 36% to 39% across the five quarters, with the last two the strongest consecutive run. In other words the core reserve economics improved; the RLDC margin fell only because other revenue, which carries essentially no distribution cost, shrank. That distinction is the single most useful thing in the release for anyone modelling the business, and it argues the on-platform strategy is still working. The forward problem is separate and unrelated to Q2: the Hyperliquid migration onto Coinbase's platform ramped too late in the quarter to matter, and management said explicitly that it starts showing up in Q3.

EPS (in line, but the quality is thin). Diluted EPS of $0.18 landed at the top of the consensus band, yet operating income from continuing operations was only $34.4M, down 23.7% sequentially. Net income of $48.2M got there with help from $17.9M of other income (of which $14.5M is interest on corporate cash) and an income tax charge of just $4.1M on $52.3M of pre-tax income, an effective rate under 8%. Strip the below-the-line contribution and the operating engine produced roughly $34M this quarter against $45M last quarter. The adjusted EBITDA figure of $143.5M is the more honest read of the operating business, and it declined 5.2% sequentially while its margin hit a post-IPO low of 50%.

Revenue Composition and the Five-Quarter Trend

Reserve income stabilized, and that is the quarter's most underappreciated fact

Reserve income peaked at $733M in Q4 2025, fell to $653M in Q1 as the reserve return rate compressed 30bps, and has now risen to $668M. Two things drove the recovery: average circulation hit an all-time high of $76.5B, and the reserve return rate held at roughly 3.5% rather than stepping down again. Circle has taken 66bps of year-over-year yield compression and, at this point, appears to have absorbed most of what the current easing cycle is going to deliver. If the rate holds here, the reserve income line becomes a pure function of float, which is a cleaner (if more uncomfortable) thing to model.

Assessment: The rate story that dominated Circle's first three quarters as a public company has largely run its course. That removes a headwind but also removes an excuse. From here, reserve income growth requires float growth, and float shrank this quarter.

Other revenue went backwards, and the guide says it stays there

Other revenue at $33.6M is up 41% year-over-year but down 19.3% sequentially, the first decline since Circle went public. Subscription and services revenue fell $7M on fewer blockchain integrations and transaction revenue fell $1M on lower validator awards. The critical detail is in the guidance rather than the print: management's ex-Arc full-year figure of $130-150M, against $75.2M already booked in the first half, implies $55-75M in the second. This line is guided to be flat at best and down 27% at worst.

"Those contracts have both an upfront component and a recurring component. And as we said, we've been aggressively working through a pipeline of those over the first few quarters of this year and at the back end of last year." — Jeremy Fox-Geen, CFO

Assessment: The honest reading is that the blockchain-partnership revenue stream was substantially a pipeline of upfront integration fees, and that pipeline has been worked through. Circle is explicitly choosing to reallocate the engineering and business-development capacity behind it toward Arc. That may prove the right trade. It also means the "platform diversification" pillar we have tracked since Q4 2025 is now, in practice, an Arc pillar rather than a diversified one.

The operating expense base is what turned a flat revenue quarter into an earnings decline

Adjusted operating expenses of $146.4M rose 7.9% sequentially and 23% year-over-year, driven by Arc marketing spend, infrastructure expansion and general and administrative investment. Over five quarters adjusted opex has gone from $119.4M to $146.4M, a 23% increase, while RLDC has gone from $251M to $289M, a 15% increase. Costs are compounding faster than gross profit, which is the arithmetic behind adjusted EBITDA margin falling from 53% to 50%.

Assessment: Management guided full-year adjusted opex unchanged at $570-585M but said it expects the high end. First-half adjusted opex was $282.1M, so the high end implies roughly $303M in the second half, another 7% step. This is a deliberate investment cycle and we would not want Circle underspending five weeks before a Mainnet launch. But the bull case has been arguing since Q1 that this ramp decelerates in the second half, and the guide says it does not.

Key Operating Indicators

The two on-platform numbers point in opposite directions, and only one of them is economic

This is the subtlety most likely to be misread. On-platform USDC at the period-end snapshot fell from $13.7B to $12.4B, and as a share of circulation from 17.8% to 16.9%. But the daily weighted average share, which is what actually determines how much reserve income Circle keeps across the ninety-one days of the quarter, rose from 17.2% to 19.5%, a 230bps improvement and up 1,204bps year-over-year. The economics follow the daily average, not the snapshot, which is exactly why net reserve margin improved to 39% while the end-of-period on-platform figure deteriorated.

Assessment: The on-platform strategy is working better than the headline suggests, and a reader who tracks only the period-end figure will draw the wrong conclusion about margin durability. That said, the gap between a 19.5% daily average and a 16.9% period-end exit means the quarter finished weaker than it ran, which is a poor handoff into Q3 and compounds with the Hyperliquid migration.

Float, wallets and share all declined together, which is harder to explain away than any one of them

At Q1 the float was flat and management attributed it to a digital-asset market correction and hack-driven deleveraging. Both were credible one-quarter explanations. This quarter the float actually declined, redemptions exceeded mints by $4B, meaningful wallets fell for the first time, and circulation-based market share slipped a point. Management's counter is that USDC circulation grew 19% year-over-year while the broader digital-asset market capitalization fell roughly 40%, which is a genuine relative-strength argument.

"While the broader digital asset market capitalization declined approximately 40% year-over-year, leading to reductions in trading activity, DeFi, collateral demand and associated market maker balances, USDC circulation grew 19% over the same period, underscoring the decoupling of USDC usage from the vagaries of the digital asset markets, the resilience of USDC through that market cycle and signposting the underlying growth in non-crypto market adoption and usage." — Jeremy Fox-Geen, CFO

The decoupling claim is partly supported and partly not. Supported: real-world payment volume grew 84% year-over-year, USDC's share of stablecoin transaction volume reached roughly 70% in June per Visa's data (a record, against 36% a year earlier), and USDC reached 40% of open interest collateral on Binance and Hyperliquid. Not supported: onchain transaction volume fell 31% sequentially, meaningful wallets declined, and circulation-based market share fell. Usage intensity is genuinely decoupling from the crypto cycle. Supply is not.

Assessment: This is now a two-quarter pattern, not a one-quarter event, and the composition changed for the worse. A flat float with growing wallets is consolidation. A declining float with declining wallets and declining share is contraction. Management has not yet offered a supply-side reacceleration mechanism beyond Arc and CPN, both of which are second-half and 2027 stories.

The non-USDC asset stack keeps compounding off a small base

EURC grew 2.2x year-over-year and remains the largest digital euro. USYC, Circle's tokenized money market fund, grew 10x year-over-year to over $3B and is the largest tokenized money market fund globally. Neither is material to the model today. USYC is the more strategically interesting of the two because tokenized money market funds are precisely the competitive threat one bulge-bracket desk cited when cutting its Circle target, and owning the largest one is a hedge against the category cannibalizing USDC balances.

Assessment: Small, cheap and directionally right. We do not model meaningful contribution before 2027, but the USYC position specifically converts a competitive threat into partial optionality, which is worth more than its current revenue.

Key Topics & Management Commentary

Overall Management Tone: Front-footed and structured around competitive rebuttal, a marked change from Q1's expansive vision-setting. The CEO opened by addressing the competitive threat head-on, without ever naming the competitor, and spending the first several minutes of prepared remarks enumerating moats (licenses, chains, banks, distribution partners) before reaching any financial result, which is the posture of a team that knew what the call was about. Where management was least convincing was on the second-half revenue bridge: the guidance raise was presented as a straightforward Arc success story, and the offsetting reduction to the organic line was disclosed in a single sentence and never revisited, including in Q&A.

1. The Coinbase Renewal on Existing Terms

Circle's distribution agreement with Coinbase is the most economically consequential contract the company has. Under it, Coinbase earns 100% of the reserve interest on USDC held on Coinbase's platform and 50% of the interest on USDC held anywhere else in the world. Circle's FY2025 payments under that arrangement were approximately $908M. The agreement carries automatic three-year renewals where both parties continue meeting their obligations, and the renewal window fell in August 2026. Coinbase joining the rival Open USD consortium as a launch partner on July 4 made the renewal terms an open question. The sharpest single-session damage came earlier, on June 30, when the consortium was unveiled and the stock fell 17.6%; the first session after Coinbase joined actually rose 6.2%, which suggests the market read the consortium itself, rather than Coinbase's participation in it, as the threat.

"That network was built with partners, including a strategic partnership with Coinbase that we have grown over many years, and I'm pleased to share today that our agreement with Coinbase has renewed on its existing terms, ensuring that USDC remains central across all of Coinbase's products." — Jeremy Allaire, Co-Founder, CEO and Chairman

"Existing terms" is the operative phrase. In a quarter where a 140-member consortium launched explicitly around sharing nearly all reserve income with distributors, Circle's largest distributor chose not to extract worse terms. That is a stronger signal about Circle's bargaining position than anything management said in rebuttal.

Assessment: This removes the single largest quantifiable contract risk for three years and is the reason a revenue miss closed flat. It does not resolve the underlying question, which is what happens at the 2029 renewal if a reserve-income-sharing model has become the market standard by then. But three years is a long runway, and Circle now has it.

2. Open USD and the Consortium Attack on Reserve-Income Economics

The competitive event of the quarter happened outside it. On June 30, Open Standard unveiled Open USD, a stablecoin backed by a consortium of more than 140 firms including Visa, Mastercard, Stripe and BlackRock, designed so that nearly all reserve interest flows to distribution partners after a management fee rather than being retained by the issuer. Coinbase joined as a launch partner on July 4. Visa launched its Stablecoin Platform on July 16 with Open USD as the first supported asset. The model targets precisely the line that produces 95% of Circle's revenue.

Management never used the name on the call. The CEO addressed it obliquely and pivoted to network scale.

"That strength is evident even in recently announced purported consortium projects. Approximately 70% of the companies that have expressed interest are already participants on our network. Whatever role they may ultimately play in those projects, the more important fact is that they are already building on, distributing and supporting USDC today." — Jeremy Allaire, Co-Founder, CEO and Chairman

The 70% statistic is a real argument and also an incomplete one. It establishes that the consortium's members are not abandoning USDC. It does not establish that they will not route incremental volume to an asset whose economics pay them more. The relevant test is not whether these firms continue supporting USDC, but what they do with new flows once Open USD exists as a live alternative. As of this print, Open USD has not launched.

Assessment: This is the most serious competitive challenge Circle has faced, and it is aimed at the business model rather than the product. Two things temper it. First, the consortium is a coalition of firms that have historically struggled to ship shared infrastructure quickly, and the asset does not yet exist. Second, several of its most prominent members are simultaneously anchoring Arc, which suggests hedging rather than conviction. We treat this as a live, unresolved structural risk that justifies a lower multiple, not as a thesis-ending event.

3. USDC Circulation Declined, and the Plateau Became a Drawdown

Ending circulation fell to $73.3B from $77.0B. Circle minted $83B and redeemed $87B during the quarter, a net outflow of $4B, against a net inflow of $1B in Q1. The gross figures are worth pausing on: $170B of combined mint and redeem volume, with daily primary-market activity averaging $1.9B and up 105% year-over-year. The plumbing is working at genuinely impressive scale. The net direction reversed anyway.

Management's framing on the call and in the release was that the softness is external.

"Our quarterly financial results reflect the current rate environment and a crypto market that has slowed" — Jeremy Allaire, Co-Founder, CEO and Chairman

At Q1 we accepted an equivalent framing and modelled a soft Q2 followed by second-half reacceleration. The soft Q2 arrived and was softer than modelled. What we did not get is any mechanism for the reacceleration: no supply-side catalyst was named for the second half other than Arc Mainnet and CPN monetization, neither of which has a demonstrated float-conversion rate.

Assessment: We are marking down our float assumptions materially, to roughly $76B average for 2026 from the $84B we carried after Q1. Two consecutive quarters without supply growth, with the second one negative, is enough to move this from a watch item to a base-case revision. The offsetting point is real and we do not want to lose it: average circulation still set an all-time high, so the reported quarter benefited from a float profile that the exit rate no longer reflects.

4. The Guidance Raise Is Entirely Arc, and the Organic Line Was Cut

This is the most important thing in the print and the easiest to miss. Full-year other revenue guidance roughly doubled.

"We are raising our other revenue guidance range to $310 million to $330 million, up from $150 million to $170 million. The increase is driven by Arc." — Jeremy Fox-Geen, CFO

Of the new range, $180M is Arc token presale revenue, recognized as product milestones are achieved. Management expects roughly 75% of those milestones to be hit in 2026. What sits underneath is the part that did not make the headlines.

"The remainder of our products are expected to deliver between $130 million and $150 million. This reflects our strategic decision to focus resources on the development of Arc instead of additional blockchain partnerships as well as moderation in commercial opportunities with new blockchain partners, reflecting softer digital asset markets." — Jeremy Fox-Geen, CFO

The original all-in guide was $150-170M. The new ex-Arc guide is $130-150M. The organic line was cut by $20M at both ends of the range, inside an announcement framed as a doubling. The same structure runs through the margin guide: RLDC margin goes to 41.7-43.7% from 38-40%, but excluding Arc, management expects the full year "near the midpoint of the prior range," which is roughly 39%. Against a first half that printed 41.3%, a full-year ex-Arc average near 39% requires the second half to run around 37% if second-half revenue approximates the first half's.

Assessment: Strip the token recognition and Circle guided the second half of 2026 to a material earnings decline. That is a legitimate outcome of a deliberate strategic reallocation, and the $242M of presale cash is already banked, so this is not a paper exercise. But $180M of one-time milestone recognition is not a run rate, and a valuation anchored to 2026 adjusted EBITDA including it will anchor to the wrong number. Our clean 2026 figure is roughly $495M, not $675M.

5. The Hyperliquid Arrangement and the Q3 Margin Step-Down

Circle and Coinbase jointly structured a revenue-sharing arrangement to bring Hyperliquid, one of the largest decentralized perpetual futures venues, firmly onto USDC. Strategically it is a win: Hyperliquid evaluated backing a different stablecoin and chose USDC, and USDC now accounts for 40% of open interest collateral across Binance and Hyperliquid. Economically it is a cost, because of where the balances sit.

"At quarter end, approximately 90% of Hyperliquid's total USDC was held within Coinbase's platform and about 10% of Hyperliquid's total USDC was within Circle's platform." — Jeremy Fox-Geen, CFO

Under the distribution agreement, USDC on Coinbase's platform carries 100% of its reserve interest to Coinbase. A 90/10 split therefore means the overwhelming majority of the Hyperliquid balance generates little or no retained reserve income for Circle, while still counting toward circulation and toward the transaction-volume and collateral-share statistics management leads with. The timing is the only reason it stayed out of the reported quarter: the migration ramped in the final weeks of June.

"The new Hyperliquid arrangement for USDC had minimal impact on Q2 results […] We expect that impact to be reflected beginning in Q3." — Jeremy Fox-Geen, CFO

Assessment: This is the mechanical explanation for why the ex-Arc second-half margin guide steps down so sharply from a first half that printed 41.3%, and it is why we would not extrapolate Q2's 41.2% forward. It also illustrates a structural tension the sell-side has started pricing: Circle can buy circulation growth and headline share statistics, but the incremental balances increasingly arrive at low or zero retained margin. Volume and revenue are decoupling, and not in the direction the bull case needs.

6. Arc Mainnet Gets a Date and a Validator Cohort Without Precedent

Arc public Mainnet launches September 16. Testnet has processed more than half a billion transactions across roughly three million wallets, and over 100 partners are active on private Mainnet. After five quarters of "coming soon," the date is the disclosure that matters most, and it removes the deferral concern we flagged at Q1.

"Today, we are announcing the initial cohort of firms that will operate the Arc blockchain network alongside Circle as network validators. This includes the world's leading asset manager, the world's leading equities and securities clearing firm, leading digital asset firms, the largest exchange group in the world, the 2 largest retail payments networks in the world, leading banks from around the world and leading payment processors and remittance companies." — Jeremy Allaire, Co-Founder, CEO and Chairman

Named in the release: BlackRock, DTCC, Galaxy, Global Payments, ICE, Mastercard, MoneyGram, SBI Group, Standard Chartered, Sumitomo Corporation and Visa. Two anchor partnerships came with it. DTCC will enable tokenization of DTC-custodied assets on Arc, extending over time to tokenized repo, collateral mobility, corporate actions, securities lending and dividend distribution. BlackRock is expected to deploy BUIDL on Arc with native USDC integration, letting institutional investors subscribe, redeem and deploy fund assets in a single onchain environment.

"DTCC underpins so much of our equities and securities markets today, and DTCC is collaborating with Circle to bring tokenized securities to Arc, focusing first on enabling the tokenization of DTC-custodied assets on Arc." — Jeremy Allaire, Co-Founder, CEO and Chairman

Assessment: No competing chain has assembled a validator set that includes the US securities depository, the largest asset manager, the largest exchange group and both major card networks. If Arc works, the institutional distribution problem is solved before launch, which is not how new networks usually start. Two cautions. First, Visa, Mastercard and BlackRock appear on both this list and the rival consortium, so their participation is a hedge across outcomes rather than an endorsement of Circle's economics. Second, "collaborating," "exploring" and "expected to deploy" are not volume commitments, and the September launch converts this from a story into a measurable one within five weeks.

7. Arc Token Economics and the Milestone Question

The Arc presale closed at $242M in Q2, up from the $222M announced at Q1, at a $3B fully-diluted network value. Circle retains 25% of the token genesis. The accounting path management described at Q1 is now producing guidance: recognized as milestones are achieved, into other revenue, then straight through RLDC to adjusted EBITDA at full margin. The balance sheet carries the trace, with other current liabilities rising from $18.4M at year-end 2025 to $256.0M at June 30, a $237.6M step consistent with the presale proceeds sitting as a deferred obligation.

What is missing is any specification of the milestones themselves. Management guided $180M as "approximately 75% of the milestones," which defines the fraction without defining the set. The risk-factor language in the release is newly explicit on why that matters, disclosing "possible repayment obligations if key launch milestones are not achieved."

Assessment: A $180M revenue item representing 56% of the guided full-year other revenue and roughly 27% of our estimate for full-year adjusted EBITDA, tied to unenumerated milestones with a repayment obligation attached if they are missed, is a disclosure gap rather than a rounding issue. The September 16 launch presumably satisfies the largest of them. We would treat the $180M as high-probability but not yet earned, and we would want the milestone schedule at the Q3 call.

8. The OCC National Trust Charter Goes Final

Circle received final approval from the Office of the Comptroller of the Currency to establish Circle National Trust, making it one of the first stablecoin issuers to hold a federal bank charter. The approval authorizes federally regulated digital asset custody and enables future management of the USDC reserve itself. A separate New York Department of Financial Services approval established Circle New York Trust as a limited purpose trust company.

"With final OCC approval, we have established Circle National Trust, an infrastructure bank for the Internet financial system. This is about confidence." — Jeremy Allaire, Co-Founder, CEO and Chairman

The conditional approval was in hand at Q4 2025. Final approval converts a contingent regulatory asset into an operating one, and the ability to eventually bring reserve management in-house has direct economic consequences that management did not quantify.

Assessment: This is the most durable part of the moat and the hardest for a consortium to replicate quickly. A federal charter takes years and cannot be assembled by committee. It is also the clearest counterweight to the competitive threat: whatever reserve-income split a rival offers, it will be offering it without a national trust bank behind it. Management underplayed the reserve-management economics, which we read as conservatism rather than absence of opportunity.

9. CPN Reaches Scale and Monetization Begins

Circle Payments Network reached $14.7B in annualized transaction volume on a trailing thirty-day basis at quarter end, up 76% sequentially, with 175 enrolled financial institutions, up 29% sequentially, across more than 58 countries. The more striking figure is the intra-quarter update.

"Sitting here today, as of July 31, annualized total payment volume on a trailing 30-day basis has already reached $23 billion, representing 130% growth since our last earnings report." — Jeremy Allaire, Co-Founder, CEO and Chairman

Management stated that monetization begins in the second half of 2026, having deliberately run the network unpriced to build scale first. Supporting institutional integrations landed during the quarter: BNY added USDC minting and redemption inside its Digital Asset Custody platform, Standard Chartered launched bank-led institutional mint and redeem, Nium connected USDC settlement across more than 190 countries, and Marex executed what the company describes as the first stablecoin-powered initial margin transaction in CFTC-regulated derivatives clearing.

Assessment: CPN is the cleanest growth asset in the portfolio and the one closest to producing revenue. At $23B annualized volume, even a 5-10bp take rate is only $12-23M annually, so it will not move 2026. But the growth rate is the point, and the enrollment of BNY and Standard Chartered as operating participants rather than pilots is the kind of institutional validation the CPN thesis needed. We model first meaningful CPN revenue in 2027.

10. Agentic Finance: Overwhelming Share of a Market That Is Still Tiny

USDC settles 99.3% of x402 agentic payment volume, down slightly from 99.8% at Q1, and the agent marketplace now hosts more than 900 paid services. Management plans to publish an agentic roadmap and white paper shortly, moving from agents that can pay to agents that can earn and be discovered, with identity and reputation layers.

The honest counterweight came from outside the call: one bulge-bracket downgrade published two days before the print put daily agentic transaction volume at approximately $41,900. Against $163B of daily onchain USDC transaction volume, that is a rounding error several orders of magnitude below materiality.

Assessment: Dominant share of a market that does not yet exist is worth having and worth almost nothing today. The strategic logic is sound, the cost is embedded in an opex base already growing 23%, and the optionality is asymmetric. But the small decline in share from 99.8% to 99.3% is the first evidence that this position is contestable, and it arrived in the same quarter competitors announced intentions to enter. We assign no value to agentic revenue before 2028.

11. Internal AI Adoption Moves From Slideware to Operating Model

86% of Circle employees are weekly active users of AI tools, and employees shipped more than 1,100 AI applications year to date, most of them during Q2 and most built by non-technical staff. Management describes product development velocity as up several hundred percent over the first half and frames the second half as moving "from adoption at scale to orchestration at scale," with hybrid human and agent teams, a shared memory layer and self-service agent authoring.

At Q1 we asked for this to be quantified in a way the buy side could model, specifically output per engineer or a headcount growth rate. That did not arrive. What arrived instead is a larger claim (several hundred percent velocity improvement) alongside adjusted operating expenses accelerating to 23% year-over-year growth.

Assessment: Those two facts sit awkwardly together. If product velocity is up several hundred percent, the cost base should at some point show operating leverage rather than acceleration. Circle is building Arc, CPN, the agent stack and a national trust bank simultaneously, so the absence of leverage is explicable. But the AI productivity narrative has now been advanced for two consecutive quarters without a single quantified cost or headcount consequence, and until one appears we treat it as capability commentary rather than a margin input.

12. Capital Allocation With the Stock Down 59%

Circle ended the quarter with $1.73B of corporate cash and equivalents, up from $1.53B at year-end, plus $889M segregated for corporate-held stablecoins. Convertible debt went to zero from $36.8M. Asked directly about instituting a dividend, the CFO declined and framed the company as a growth asset.

"Fundamentally, we are a massive future market growth stock versus a stock that returns capital to shareholders today." — Jeremy Fox-Geen, CFO

We agree with the conclusion and note what it left unaddressed. A dividend from a company at this stage would be a strategic error. A repurchase authorization, with the shares down roughly 59% over twelve months and management repeatedly describing the platform as undervalued relative to its opportunity, is a different question, and it was neither asked nor volunteered.

Assessment: Retaining capital ahead of a Mainnet launch and an unresolved competitive fight is defensible. But management is asking shareholders to accept a 32% ex-Arc second-half earnings decline as the price of investment while holding $2.6B of corporate liquidity and declining to address buybacks at a 52-week-low-adjacent share price. That combination will keep drawing questions.

Guidance & Outlook

Management also defended the multi-year 40% circulation CAGR framework at length, pre-empting the obvious question about how a 40% target survives a shrinking float. The defence rests on third-party projections putting the 2030 stablecoin market between roughly $1T and $4T, implying 27% to 77% compound growth, with Circle's 40% sitting inside that band, plus a total addressable market for money of about $120 trillion of which roughly half is non-interest-earning.

Implied second-half ramp: The full-year adjusted opex high end of $585M against $282.1M booked in the first half implies roughly $303M in the second, a 7% step. On the revenue side, the ex-Arc other revenue guide of $130-150M against $75.2M in the first half implies $55-75M in the second. Combining these with the roughly 37% implied ex-Arc second-half RLDC margin produces ex-Arc second-half adjusted EBITDA around $200M against $294.9M in the first half. The $180M of Arc recognition brings the reported second half back to roughly $380M, which is why the guide reads as strength.

Street at: Published targets on the name span roughly $38 to $96, a 2.5x spread, which is unusually wide and reflects genuine disagreement about whether the reserve-income model survives a income-sharing competitor rather than disagreement about near-term estimates. Consensus revenue estimates were revised down repeatedly over the three months into the print with no EPS upgrades. We expect full-year adjusted EBITDA estimates to move up on the Arc recognition and 2027 estimates to move down on the float and Hyperliquid revisions, which will make the reported 2026 number look better and the forward multiple look worse.

Guidance style: Circle continues to guide only what it controls, excluding the reserve rate and total revenue entirely, and continues to structure the controllable items conservatively. The new element this quarter is a guidance raise whose headline and whose substance point in different directions. That is not misleading, since management disclosed the ex-Arc components clearly enough for anyone doing the arithmetic. But it is the first quarter where reading the guide at face value would have produced a materially wrong conclusion about the underlying business, and we now assume investors will need to decompose every future guide the same way.

Analyst Q&A Highlights

Competing for Distribution When Reserve-Income Sharing Becomes Table Stakes

The dominant topic of the call, raised first and framed most precisely. The question put the competitive threat in structural terms: if a rival model shares reserve income equally with all distributors, is Circle constrained from matching by the economics of its existing Coinbase relationship? It also surfaced the joint Circle-Coinbase revenue-share arrangement with Hyperliquid as evidence that Circle can deploy distribution capacity alongside Coinbase rather than being boxed in by it. Management's answer emphasized existing scale (over 150 distribution partnership agreements carrying economic incentives) and pivoted to the announcement that both major card networks were joining Arc.

Q: "The market seems to be framing this potentially as a binary issue, Open Standards model of equally sharing reserve income with all distribution partners, including Visa, potentially as a structural advantage that Circle can't replicate given its existing economics with Coinbase. But what caught our attention, I think, this quarter was the joint Hyperliquid announcement, which you spoke about earlier, where Circle and Coinbase collaborated on a share revenue agreement together."
— Pete Christiansen, Citi

A: "Just today, we announced expanded collaboration with Visa and with Mastercard, who are becoming key infrastructure partners in Arc, which is USDC native in terms of its transaction infrastructure and settlement infrastructure. And so we see around 70%, in fact, of companies involved in these kind of consortium efforts already building with us."
— Jeremy Allaire, Co-Founder, CEO and Chairman

Assessment: The question was better than the answer. It correctly identified that the Hyperliquid structure is the template for how Circle competes for distribution without unilaterally repricing its whole book, which is genuinely the most important strategic mechanism disclosed this quarter. Management confirmed the capability exists but declined to characterise the economics, and redirected to Arc participation. Reading the two halves together: Circle can match competitive distribution terms deal by deal, and each time it does, retained margin falls. That is the trade the Hyperliquid arrangement makes explicit and it is why we cut our forward margin assumptions.

Where the Hyperliquid Balances Actually Sit

A follow-up sought specifics on the Arc presale and on the Hyperliquid deal structure. Audio problems garbled much of the question, but the response was the most economically revealing disclosure of the call, quantifying the split of Hyperliquid's USDC between the two platforms and therefore the retained-margin consequence, while declining to detail the revenue-share terms.

A: "You can see on chain the exact location of the funds within Hyperliquid's platform in relation to where they're held within either Circle or Coinbase's platform. At quarter end, approximately 90% of Hyperliquid's total USDC was held within Coinbase's platform and about 10% of Hyperliquid's total USDC was within Circle's platform. As for the specifics of how that revenue share is detailed, we're not commenting in more detail on the precise nature of that between Circle and Coinbase."
— Jeremy Fox-Geen, CFO

Assessment: Volunteering the 90/10 split while withholding the revenue-share terms is a deliberate trade, and a defensible one, since the split is observable onchain and the terms are not. The number itself is the takeaway. Under a contract that gives the distributor 100% of reserve interest on balances held on its platform, a 90/10 split means a strategically important circulation win converts to very little retained revenue. This is the concrete instance of the general problem: Circle is winning volume battles on terms that do not flow to the income statement.

Trading Blockchain Partnership Revenue for Arc

A direct question on the disclosed decision to deprioritize blockchain partnerships in favour of Arc, and specifically what that does to the subscription and services component of other revenue going forward. Management answered the strategic half expansively and the modelling half by pointing at the guide.

Q: "Jeremy, you spoke to a deliberate decision to prioritize Arc over other blockchains and that could have an impact on other revenue. I was hoping you might just expand a little bit on this decision, what this means going forward and how this changes the subscription services component of other revenue going forward."
— James Yaro, Goldman Sachs

A: "I think I maybe said on the last earnings call, we look at Arc as potentially as bigger than an opportunity than USDC itself. And so this is the birth of a new operating system layer for economic activity in the world. […] And so from our perspective, that as a source of major other revenue going forward is very attractive. The margin characteristics are very attractive."
— Jeremy Allaire, Co-Founder, CEO and Chairman

Assessment: "Bigger than USDC itself" is an extraordinary claim for a network that has not launched, and management has now made it twice. It clarifies the capital allocation: this is not a portfolio of platform bets, it is a concentrated wager on Arc funded partly by running down an existing revenue line. The CFO's follow-on, that the deferred question about 2027 subscription revenue can be inferred from the second-half implication of the guide, was the closest anyone came to acknowledging the organic cut. Nobody followed up on it.

Distribution Costs, On-Platform Mix, and Why Margin Is Unforecastable

The most technically precise question of the call, probing whether costs specific to on-platform USDC fell between the first and second quarters and how mix shifts, including CPN, are affecting what Circle pays to win on-platform balances. Management declined the disclosure and instead offered the most candid statement of the quarter about the limits of margin forecasting.

Q: "Can you talk about distribution and transaction costs associated with USDC on Circle platform? I can't tell for sure, but it looks like costs specific to Circle on Platform USDC came down a lot from 1Q to 2Q. Did I get that right? And if so, how did the mix of USDC on Circle Platform change?"
— Kenneth Worthington, JPMorgan

A: "There are pockets of heavily incentivized USDC. There are large pockets and very large dispersed surface area of USDC that carries very little incentive at all. And so within any one quarter and over any period of time, there's movements in every single one of those pockets. Now I appreciate that makes it very, very hard to forecast margin quarter-on-quarter, and that's the world we're inhabiting in."
— Jeremy Fox-Geen, CFO

Assessment: This is the honest answer and it is also the problem. A business whose gross margin depends on the quarter-to-quarter migration of balances between incentive pockets that are never disclosed cannot be modelled to better than a few hundred basis points, and a few hundred basis points on $2.9B of revenue is most of the earnings. That opacity was tolerable while the margin trend went one way for four quarters. Now that it has turned, and with a known Hyperliquid step-down coming, the same opacity works against the multiple. This is a disclosure Circle should get ahead of.

When Agentic Commerce Becomes Material

A question on the second-half agentic product roadmap and, more pointedly, when agentic commerce contributes revenue that matters. Management described the architecture at length (identity, discovery, reputation, monetization, all on open standards) and located the revenue not in the agentic protocols themselves but downstream in stablecoin balances and Arc network activity.

Q: "Could you please add more color on the road map of agentic product in the second half of this year? When should we expect to see more revenue contribution from agentic commerce? I know it's still small. It's still growing, but we just want to understand when it can become more material to Circle longer term."
— Owen Lau, Clear Street

A: "So we see the agentic stack as driving some revenue around the protocols themselves, but fundamentally driving stablecoin adoption, which drives revenue and driving Arc infrastructure adoption, which drives revenue as well."
— Jeremy Allaire, Co-Founder, CEO and Chairman

Assessment: The answer contained no timeline, which is the correct answer to an unanswerable question but leaves the modelling exercise where it was. The useful content is the admission that agentic protocols are not themselves the monetization vector; they are a demand driver for float and for Arc. That reframes the agent stack from a revenue line into a customer acquisition channel for the two things that do monetize, which is a more defensible framing and a less valuable one in the near term.

Whether the x402 Lead Is Defensible

A question on whether Circle's overwhelming share of agentic payment settlement constitutes a durable right to win as competitors launch stablecoins targeting the same standard. Management grounded the answer in network effects and made an interesting argument about the intelligence layer itself reinforcing incumbency.

Q: "I understand USDC pretty much dominates that right now at 99% plus. There are some other competitors who are looking to launch stablecoins and operate within that. Is there sort of almost an inherent right to win for you guys within that x402 stack or even a benefit you could get from just more adoption within those payment channels?"
— John Todaro, Needham

A: "And so we absolutely have a right to win, and we are clearly winning. […] And so even the LLMs as they work with the agents that are deployed on them and discover like what can I use and what's available and what are other agents using, that actually is a network effect. The intelligence layer itself is informed by the utility that exists and the scale of that, that exists."
— Jeremy Allaire, Co-Founder, CEO and Chairman

Assessment: The argument that model training and retrieval bake in the incumbent default is genuinely novel and, we think, partly right. If agents learn which rail to use from a corpus in which USDC is the overwhelming answer, that is a real and self-reinforcing advantage. The caveat is scale: the category is currently measured in tens of thousands of dollars a day, so the network effect is being established over a base too small to defend anything yet. Ask again in eight quarters.

Capital Return With the Shares Down Sharply

A shareholder-submitted question asked whether Circle plans to introduce a quarterly dividend. The answer was an unqualified no, delivered with the clearest articulation of the capital allocation philosophy management has given on any call, and notable for arriving in a quarter when the shares sit roughly 59% below year-ago levels. The CFO's reasoning was that "the returns available to our shareholders on investing in the platform are far greater than those from sort of paying out quarterly dividends."

Q: "The second question comes from Sean, and he wants to know if we plan to roll out quarterly dividends in the near future."
— Shareholder question submitted via Say, read by Scott Blair, Head of Strategic Finance

A: "The short answer is no, we don't. But let me put that in context. We have a massive opportunity ahead of us to be the leading Internet platform company as the whole world evolves from traditional technologies and rails into new Internet-based financial services built on blockchain technology. […] Fundamentally, we are a massive future market growth stock versus a stock that returns capital to shareholders today."
— Jeremy Fox-Geen, CFO

Assessment: The right answer to the question asked. What makes the exchange worth surfacing is the question that followed it, which is none. With $2.6B of corporate liquidity, a share price down 59% year-over-year, and management's own view that the platform is worth far more than the market credits, the absence of any repurchase discussion is conspicuous. A dividend would signal the growth story is over. A buyback would signal management thinks the stock is mispriced. Declining the first while never addressing the second leaves the capital allocation question genuinely open.

What They're NOT Saying

1. The renewed Coinbase economics, in any quantified form: "Existing terms" was the entire disclosure. No restated cost curve, no term length stated in the release, and no clarification of whether the Hyperliquid revenue-share sits inside or alongside the renewed contract. Given that Coinbase-related distribution payments ran approximately $908M in FY2025 and $330.6M in Q1 2026 alone, this is the largest single line item in Circle's cost structure and it now runs three more years on undisclosed-but-unchanged terms.
2. The competitor's name, or any quantification of the distribution-cost risk: Management addressed "recently announced purported consortium projects" without naming Open USD, and offered a share-of-network statistic in place of an economic analysis. Nobody quantified what matching a reserve-income-sharing model across the top ten distributors would cost, which is the number every model needs.
3. The size of the Hyperliquid margin impact they explicitly flagged for Q3: Management volunteered that the effect begins in Q3 and volunteered the 90/10 platform split, then stopped short of sizing it. The ex-Arc second-half margin guide implies it is large. Investors are left to back into it from a full-year range.
4. The Arc milestone schedule: $180M of guided revenue, roughly a quarter of guided full-year adjusted EBITDA, is tied to achieving "approximately 75% of the milestones." The milestones are enumerated nowhere, and the release newly discloses possible repayment obligations if key launch milestones are not achieved.
5. Any rate sensitivity disclosure, now for a sixth consecutive quarter: The reserve return rate finally stabilized this quarter, which would have been a natural moment to frame forward sensitivity. It passed without comment.
6. Why redemptions exceeded mints by $4B: Circle disclosed $83B minted and $87B redeemed and never addressed the net figure on the call. The circulation decline was framed entirely through market-capitalization weakness rather than through primary-market flow.
7. The decline in meaningful wallets: The wallet count fell from 7.2M to 7.0M, the first sequential decline on record, and was mentioned nowhere in the prepared remarks or the Q&A. For a company arguing that usage is decoupling from the crypto cycle, a shrinking user base is the datapoint most in tension with that claim.
8. Share repurchase: Dividends were addressed and rejected. Buybacks were not raised by management or by any analyst, despite $1.73B of corporate cash and a share price down roughly 59% over twelve months.
9. Quarterly guidance of any kind: With a known margin step-down starting in Q3 and a Mainnet launch mid-quarter, Circle continues to guide only annually. The absence is defensible policy and unhelpful timing.

Market Reaction

• Pre-print setup: Closed at $63.25 on August 4. Down 20.2% year to date against the S&P 500 up 13.0%, down 58.9% over trailing twelve months, and down 7.9% over the trailing thirty days. The 52-week closing range entering the print was $50.23 to $168.10, so the stock came in at roughly 26% off its low and 62% below its high. The setup was heavily de-risked: a bulge-bracket downgrade two sessions earlier had cut a published target by roughly two-thirds, and the stock fell 3.6% that day to $60.35 before recovering 4.8% into the print.
• Print-day session (August 5, before the open): Gapped down 2.0% to open at $61.96 and traded as low as $59.12, down 6.5% from the prior close, on the revenue miss and the sequential circulation decline. Recovered through the session to close at $63.28, up $0.03 or effectively unchanged, against the S&P 500 down 0.2%. Volume of 20.1M shares was 1.5x the thirty-day average.
• Following sessions: August 6 closed at $63.28 again, unchanged, after trading as high as $65.91. August 7 gapped up and closed at $66.67, up 5.4%. Cumulative move from the pre-print close through August 7: up 5.4%, against the S&P 500 up 0.3% and Coinbase up 1.9% over the same three sessions.
• Sell-side response: One desk that had carried an Underperform rating through the interquarter downgrade cycle upgraded to Neutral on August 7, coinciding with the strongest session. Published targets across the name remain unusually dispersed.

The intraday round trip on print day is the whole story of the quarter compressed into six and a half hours. The market opened on the numbers, which were weak: a revenue miss, a shrinking float, a margin at a post-IPO low. It closed on the contract, which was strong: Coinbase renewed on existing terms through 2029, removing the risk that had driven a 17.6% single-session decline on the consortium's unveiling five weeks earlier and had anchored two downgrades since. When a stock gaps down 6.5% on the print and finishes flat, the recovery is being paid for by something the numbers did not contain.

The follow-through is more informative than the print day. Circle outperformed the index by roughly five percentage points over three sessions while Coinbase, the most direct read-across, added under two percent. That is Circle-specific repricing rather than a sector move, and it is consistent with the market working through the Arc validator list and the September date over a couple of days rather than pricing them instantly. It is also a very small repricing relative to what was lost: the stock is up 5.4% off a base that is 52% below where it closed after the Q1 print.

What the tape has not yet priced, in our view, is the ex-Arc second-half guide. The doubled other-revenue headline is easy to read and easy to price. The $20M cut to the organic line, the roughly 37% implied ex-Arc second-half RLDC margin and the resulting 32% ex-Arc earnings decline require decomposing three separate disclosures. That work will get done in estimate revisions over the coming weeks, and we would expect 2027 numbers to come down even as 2026 numbers go up.

Street Perspective

Debate: Does a Reserve-Income-Sharing Competitor Actually Break Circle's Model?

Bull view: The consortium is a coalition of 140-plus firms attempting to ship shared financial infrastructure, an exercise with a poor historical completion rate, and the asset does not yet exist. Roughly 70% of the participating companies already build on Circle's network. Circle holds a federal trust charter that no consortium can assemble on any relevant timeframe, 55-plus licenses, presence on 35 chains, and a distribution book of over 150 partners. Most tellingly, Circle's largest distributor just renewed on unchanged terms rather than extracting the improved economics a credible alternative should have enabled.

Bear view: The threat is not that Open USD wins; it is that its existence resets the price of distribution. Every renewal from here gets negotiated against a published alternative that pays distributors nearly all of the reserve income. Circle does not need to lose partners to lose margin, it only needs to keep them at worse terms. The Hyperliquid arrangement is the template and it already shows what that looks like: a headline circulation win where 90% of the balances sit where Circle retains nothing. Cutting 2027 and 2028 USDC supply forecasts by a third and by 44% respectively, as one desk has, follows directly from that logic.

Our take: The bear framing is the more rigorous one and we have moved toward it, but it overshoots on timing. The margin-reset mechanism is real, observable in the Hyperliquid structure, and already embedded in management's own ex-Arc second-half guide. What the bear case cannot yet demonstrate is velocity, because the competing asset has not launched and the anchor distributor is contractually settled for three years. Our resolution is to lower the forward margin assumption and lower the multiple, rather than to model share loss. The disconfirming test is specific: if Open USD launches and any top-ten Circle distributor publicly routes new volume to it, the bear case moves from structural risk to realized loss and the rating goes to Underperform.

Debate: Is Arc a Business or a One-Time Revenue Event?

Bull view: The validator cohort settles it. The US securities depository, the largest asset manager, the largest exchange group and both major card networks do not commit operational resources to running validator infrastructure for a chain they expect to fail. DTCC tokenizing DTC-custodied assets and BlackRock deploying BUIDL are institutional flows with real balance-sheet weight behind them, and they arrive at launch rather than being courted afterward. Circle's 25% token stake is worth roughly $750M at the presale mark and scales with network value. The $180M of 2026 recognition is the smallest part of the story.

Bear view: Every dollar of Arc revenue in the 2026 guide comes from selling tokens, not from network activity, and it is recognized against milestones the company will not enumerate, with repayment obligations attached if they are missed. Circle is funding this by explicitly running down the blockchain-partnership revenue that was the platform-diversification proof point. Strip the token recognition and the second half declines 32%. "Collaborating," "exploring" and "expected to deploy" are not volume commitments, and the validators most often cited as validation are simultaneously backing the competing consortium.

Our take: Both are right about different time horizons and the resolution date is unusually near. The bears are correct that 2026 Arc revenue is a financing event dressed as revenue and should not be capitalized at a business multiple, which is why we value Circle on a clean $495M rather than the reported ~$675M. The bulls are correct that the validator set is without precedent and cannot be dismissed as a press release. September 16 is five weeks away, and by the Q3 call there will be observable transaction data. That is an unusually short wait for an unusually large question, and it is a substantial part of why we are at Hold rather than Underperform.

Debate: Is the Shrinking Float a Cycle Problem or a Share Problem?

Bull view: The digital-asset market capitalization fell roughly 40% year-over-year and USDC circulation still grew 19%, which is the definition of relative strength. Average circulation set an all-time high. USDC took stablecoin transaction share to roughly 70% in June from 36% a year earlier, real-world payment volume grew 84%, and USDC reached 40% of open interest collateral on the two largest perpetuals venues. The float is a lagging indicator of a cycle; the usage metrics are a leading indicator of the franchise, and they are accelerating.

Bear view: Circulation-based market share fell to 27% from 28%, meaningful wallets fell for the first time, redemptions exceeded mints by $4B, and onchain transaction volume fell 31% sequentially. That is not one cyclical metric, it is the supply side, the user side and the activity side all turning together. Tokenized money market funds and tokenized deposits now offer yield that a non-interest-bearing stablecoin structurally cannot, which is a permanent reason for large balances to sit elsewhere. Two consecutive quarters without growth, the second one negative, is a trend.

Our take: The bears win the quarter and the bulls have the better long-run argument, which is an uncomfortable place for a twelve-month rating. The usage-versus-supply divergence is genuine and the transaction-share data is not manufactured. But we set an explicit test at Q1, that another non-growth quarter combined with decelerating other revenue would strengthen the bear case, and both legs tripped with room to spare. Circle's own answer to the yield-bearing-alternatives problem is USYC, and owning the largest tokenized money market fund is a real hedge. We now model roughly $76B average circulation for 2026 against the $84B we carried, and we will not model reacceleration until we see a quarter of it.

Model Update

The single most important change is the introduction of a clean adjusted EBITDA line that excludes Arc token presale recognition. Reported 2026 adjusted EBITDA of roughly $675M includes $180M of milestone recognition that will not repeat at that scale in 2027 absent further token sales. Capitalizing the reported figure at a business multiple double-counts a financing event. Our clean 2026 estimate is $495M and our 2027 estimate is $500M, which assumes modest float recovery, the start of CPN monetization, early Arc network fees and staking revenue, offset by the full-year effect of the Hyperliquid margin step-down and adjusted opex growing roughly 10%.

The 2027 range of $450-550M is wide because the inputs are. A bear path of float declining another 10% with the reserve rate at 3.2% and no CPN or Arc monetization produces something closer to $300M. A bull path of float recovering to $85B average, ex-token other revenue reaching $300M on Arc network fees and CPN take rates, and RLDC margin holding at 41% produces something above $600M. We are not confident enough in either to carry it as a base case, and that dispersion is itself part of the rating.

Valuation: At the August 7 close of $66.67 on 268.6M diluted shares, market capitalization is approximately $17.9B. Deducting $1.73B of corporate cash gives an enterprise value near $16.2B; deducting the $750M carrying reference for Circle's 25% Arc token stake at the presale-implied network value leaves roughly $15.4B attributable to the operating business. That $15.4B against our clean FY2027E adjusted EBITDA of $500M is about 31x. On the $16.2B enterprise value, which still carries the Arc stake, reported FY2026E adjusted EBITDA of $675M including token recognition is about 24x. Thirty-one times forward EBITDA is a growth multiple, and this quarter the float shrank, wallets shrank, market share slipped and ex-token second-half earnings are guided down 32%. The multiple is being paid for the charter, the transaction-share position and the Arc option, not for current growth.

Fair value: $50-75 over 12 months. The low anchor is 24x the bottom of our clean FY2027E adjusted EBITDA range ($450M), giving $10.8B of enterprise value, plus $1.73B of corporate cash and the Arc stake at its $750M presale mark, or roughly $49 per share on 268.6M diluted. The high anchor is 31x the top of that range ($550M), giving $17.1B, plus the same cash and the Arc stake marked at $1.5B on the assumption that a successful Mainnet roughly doubles the network's presale-implied value, or roughly $75. The range brackets the current price with the stock trading in its upper half: the $62.50 midpoint sits about 6% below the last close and the high end about 13% above it. Bull case $95-115 (roughly +42% to +72%) requires Arc Mainnet producing measurable institutional settlement volume by the Q4 call, float reaccelerating, and the consortium failing to launch a competitive asset. Bear case $30-40 (roughly -55% to -40%) requires Open USD launching with live top-tier distribution and the float continuing to contract into 2027.

Thesis Scorecard Post-Earnings

Overall: Thesis materially weakened. Of the five bull pillars, one deepened (regulatory moat), one advanced (Arc), and three were challenged (adoption, diversification, margins). Of the six bear points, two improved (rate dependency, the Coinbase contract), one reopened after we prematurely closed it (competition), and two escalated (supply, opex). The Q1 thesis rested on a four-quarter margin regime plus a broadening platform plus a closing competitive threat. All three of those legs weakened in a single quarter. What replaced them is a single concentrated bet on Arc with a date on it and an institutional cohort behind it, plus a federal charter that is genuinely hard to replicate.

Action: Downgrade to Hold from Outperform. We were wrong to carry Outperform through a quarter in which the stock fell 52%, and the specific error was closing the competitive bear point at Q1 on consumer-platform adoption evidence when the actual threat was to the economics rather than the product. Fair value $50-75 brackets the current $66.67 with the stock in the upper half of the range, so we see no cushion and no reason to add. We are not going to Underperform five weeks before a dated Mainnet launch with an unprecedented validator cohort, on a stock already down 58.9% over twelve months, in the same quarter its largest contract risk resolved favourably for three years. Monitor items for Q3 2026: (1) the September 16 Arc Mainnet launch and observable transaction volume; (2) the size of the Hyperliquid margin impact management flagged for Q3; (3) whether USDC circulation stabilizes above $73B; (4) the Arc milestone schedule and the pace of the $180M recognition; (5) whether Open USD launches and which distributors carry it; (6) CPN monetization terms and first revenue; (7) any repurchase authorization.