# The Flywheel Hidden Inside Coinbase's Income Statement

> A bottom-up look at seven business lines, one accounting rule change, and the single contract renewal that determines whether the margin story holds in 2026

- Author: Siddhant Shah
- Published: 2026-03-02
- Canonical URL: https://siddhants.com/blog/coin-research-202603/
- Series: Coinbase: The Infrastructure Flywheel

## Illustrative exercise

This is a **demonstrative modelling exercise**, written to show a method: how the numbers are built, which assumptions carry the answer, and where the disclosure gaps are. It is **not a price target, not a recommendation, and not a view on the security**. Any valuation range shown is an output of the stated assumptions, not a judgement about what the shares are worth.

## Thesis

Coinbase is not an exchange with side projects. It is a vertically integrated financial infrastructure company — seven distinct layers of the crypto financial stack, each generating standalone revenue and each feeding the others. The FY2025 numbers confirm what the strategy has been building toward: $7.2B in total revenue, 44.4% Adjusted EBITDA margin, $11.3B in cash, and $1.2B in buybacks executed. More than 39% of revenue now comes from subscription and services — sources that don't require a single trade to be executed.

The compounding is real, and it's already in the financials. But so is the primary risk. One contractual renegotiation — the `USDC` revenue-sharing agreement with Circle, up for renewal in 2026 — has the potential to remove $220-300M from the P&L in a single event. A company executing one of the most sophisticated vertical integration strategies in financial services has a single-counterparty concentration risk at the center of its recurring revenue base.

That's the tension 2026 will resolve.

This series analyzes all seven layers. What follows is the analytical case for each one, the financial evidence behind it, and the specific risks most models I've come across have not yet priced.

## Financial Snapshot — FY2025 Actuals

*[Figure: The Seven Layers of the Coinbase Flywheel]*

| Metric | FY2025 |
| --- | --- |
| Total Revenue | \$$7.2$B |
| Subscription & Services Revenue | \$$2.83$B ($39\%$ of total) |
| Adjusted EBITDA | \$$3,193.7$M |
| Adjusted EBITDA Margin | $44.4\%$ |
| Adjusted EPS | \$$6.95$ |
| GAAP Net Income / (Loss) — Q4 2025 | (\$$667$M) |
| Cash & Cash Equivalents | \$$11.3$B |
| Buybacks Executed | \$$1.2$B |
| Assets on Platform (Q3 2025) | \$$516$B |
| Monthly Transacting Users | $9.2$M |
| Total Trading Volume | \$$5.2$T ($+156\%$ YoY) |

*Adjusted EBITDA of \$$3,193.7$M is Coinbase's company-reported figure, which adds back all non-cash and non-recurring items including crypto mark-to-market, investment gains/losses, and restructuring charges. My model uses a narrower definition (GAAP EBIT + SBC + acquisition-related amortization + D&A) that produces \$$2,530$M. The difference is entirely definitional — the operating inputs are identical.*

*Note: The Q4 2025 GAAP net loss is driven almost entirely by non-cash ASC 350-60 fair-value accounting on the crypto portfolio and Circle equity stake — not operating deterioration. Adjusted metrics are the economically meaningful read. Full explanation in [later](#digital-asset-portfolio-&-asc-350-60-—-the-accounting-story-behind-the-$667m-loss).*

Download: [COIN Model.xlsx](https://siddhants.com/COIN%20Model.xlsx)

## USDC — The \$1.35 Billion Revenue Stream Coinbase Does Not Operate

**FY2025 Revenue: \$1,348.8M | % of Total: 18.8% | YoY Growth: +48%**

Coinbase does not issue `USDC`, does not hold its reserves, and does not set its yield. What Coinbase holds is a revenue-sharing agreement — confirmed in Circle's S-1 — entitling it to $100\%$ of reserve interest on `USDC` balances held on-platform and $50\%$ on `USDC` held off-platform. At an average on-platform balance of \$$17.8$B in FY2025 and an approx. $3.9\%$ effective net yield to Coinbase (after the on-platform/off-platform sharing split), that contract produced \$$1,348.8$M — the single largest recurring revenue line in the business, growing $48\%$ year-over-year on balance growth rather than rate movement.

The agreement renews in 2026. Circle completed its IPO in June 2025 (ticker: CRCL) — meaning its public shareholders can now see exactly what it pays Coinbase — and Circle has already paid Binance \$$60.25$M upfront plus monthly distribution fees to carry `USDC`. That deal is the market precedent. A shift to $80\%/40\%$ sharing terms removes \$$220$-$300$M from Coinbase's FY2026 P&L depending on the rate path — \$$259$M at the modeled $3.5\%$ Fed Funds Rate.

The `USDC Scenarios` tab in my model maps the full $3\times 3$ matrix of sharing-rate and rate-path outcomes — a \$$910$M spread across nine cells on a single revenue line.

**Primary risk: `USDC` renewal terms. Secondary risk: Fed rate sensitivity.**

Deep dive: [Coinbase's Stablecoin Business Is Not What You Think It Is](https://siddhants.com/blog/coin-research-202603/usdc-stablecoin/)

## Base Chain — 30x Growth, Zero Consensus Coverage

**FY2025 Gross Sequencer Revenue: \$75.4M | YoY Growth: approx. 30x | L2 Fee Market Share: 62%**

Base generated \$$75.4$M in gross sequencer revenue in FY2025 — 30x year-over-year — while commanding 46% of total L2 DeFi TVL at \$$4.63$B and 62% of the L2 fee market by revenue. As of February 2026, it processes approx. $10.7$ million transactions per day at roughly \$$147,000$ in daily fees.

Most models I've come across bury the entire Base revenue stream inside "Other Transaction" with no sub-breakout and no forward estimate for sequencer economics specifically. Here's what makes the gap more significant than it looks: on February 18, 2026, Coinbase announced Base is departing the OP Stack, eliminating the approx. $12\%$ Optimism Protocol revenue share beginning Q2 2026. Coinbase now retains $100\%$ of sequencer fees. That event has not been incorporated into consensus models I've reviewed.

**Primary risk: DeFi activity cyclicality. Upside catalyst: CDP enterprise client growth driving Base transaction volume.**

Deep dive: [Base Is Coinbase's Best-Kept Secret, and the Numbers Are Starting to Show It](https://siddhants.com/blog/coin-research-202603/base-chain/)

## Deribit — Transforming the Institutional Revenue Profile

**FY2025 Partial-Year Contribution: approx. \$152M | Q4 2025 Institutional Revenue: \$185M (+37% QoQ) | Acquisition Price: \$2.9B**

On August 14, 2025, Coinbase closed Deribit — the world's largest crypto derivatives exchange by open interest — for \$$2.9$B (\$$700$M cash plus approx. $11$ million shares). The partial-year contribution was approx. \$$152$M and was the primary driver of the $39\%$ YoY growth in institutional transaction revenue. Q4 2025 institutional revenue hit \$$185$M, up $37\%$ sequentially.

FY2026 is the first full year. At the \$$33$-$35$M monthly run-rate implied by Q4 2025, annualized Deribit revenue alone approaches \$$400$-$420$M — roughly $7$x forward institutional revenue on the acquisition price, defensible given the franchise position. The GAAP cost is real: the Deribit PPA Schedule tab estimates approx. \$$194$M in annual amortization beginning FY2026, a \$$127$M step-up from the partial-year FY2025 figure of \$$67.1$M.

CME Group's 24/7 crypto futures and options launch on May 29, 2026 is the first credible institutional competitor to materialize since the acquisition closed. The consensus FY2026E for institutional revenue of \$$695.9$M (+45% YoY) carries no adjustment for it. My model estimates \$$950$M for FY2026E institutional transaction revenue, inclusive of a \$$400$M first-full-year Deribit contribution — materially above consensus.

**Primary risk: CME competitive entry in H2 2026 and PPA amortization drag on GAAP EBIT. Strategic case intact on medium-term horizon.**

Deep dive: [The Deribit Bet: What Coinbase Actually Bought for $2.9 Billion](https://siddhants.com/blog/coin-research-202603/deribit/)

## Ethereum Staking — A Math Problem Disguised as a Growth Story

**FY2025 Revenue: \$677.4M | Q4 2025: \$152M (-18% QoQ) | Consensus FY2026E: \$595.2M (-12% YoY) | Model FY2026E: \$249.8M (-63% YoY)**

Coinbase earns a commission of approx. $29$-$35\%$ on gross `ETH` staking rewards — high-margin, recurring, zero customer acquisition cost. In FY2025 that produced \$$677.4$M. The problem is that the same mechanics that made this a growth driver in 2023-2024 are now working in reverse.

ETH staking participation has crossed 30% of total supply, with $36$M+ `ETH` staked. Block rewards are fixed in aggregate — divided across a growing validator set, per-validator yields decline. From approx. $5\%+$ APY in early 2024 to $3.5$-$4.2\%$ APY by early 2026, with approx. $20$-$30$ basis points of further annual compression expected. Q4 2025 staking revenue was \$$152$M, down $18\%$ sequentially, in a quarter where `ETH` prices were relatively stable — consistent with yield compression driving the decline rather than price movement alone.

The consensus FY2026E of \$$595.2$M is in the right direction but it requires `ETH` prices to hold roughly flat to materialize.

My model estimates \$$249.8$M — substantially below consensus — driven by formula: $2.7$M ETH in custody x \$$2,500$ average ETH price x $3.7\%$ net user yield. The gap versus consensus reflects the compounding effect of yield compression applied to the full balance, which most consensus estimates do not fully capture.

**Primary risk: `ETH` price appreciation required to offset structural yield dilution.**

Deep dive: [Why Coinbase's Staking Business Is Structurally Getting Smaller](https://siddhants.com/blog/coin-research-202603/eth-staking/)

## Digital Asset Portfolio & ASC 350-60 — The Accounting Story Behind the \$667M Loss

**Q4 2025 GAAP Net Loss: (\$667M) | Unrealized Crypto Loss: \$718M | Adjusted EBITDA Margin: 44.4%**

Coinbase reported a (\$$667$M) GAAP net loss in Q4 2025. The cause: \$$718$M in unrealized losses on its corporate crypto portfolio and \$$395$M in fair-value markdown on its Circle equity stake, both flowing through net income under ASC 350-60 — adopted January 1, 2025. No cash changed hands. No position was sold.

Under ASC 350-60, digital assets classified as held for investment are marked to fair value each quarter, with gains and losses flowing directly to the income statement. Coinbase held $15,389$ `BTC` (approx. \$$1.34$B at year-end $87{,}000$/BTC) and $151,175$ `ETH` (approx. \$$297$M) at year-end 2025. Every $10\%$ move in those assets produces a approx. \$$101$M GAAP income impact with zero connection to the operating business. GAAP EPS for Coinbase is now structurally a hybrid: part operating earnings, part crypto mark-to-market.

Adjusted EPS of \$$6.95$ is the economically meaningful number. Adjusted EBITDA of \$$3,193.7$M at a $44.4\%$ margin is the operating story. Any screen or model filtering by GAAP earnings will systematically misread every quarterly result for as long as Coinbase holds a material crypto portfolio.

**Key implication: Use Adjusted metrics exclusively. GAAP headlines will continue to distort in both directions every quarter.**

Deep dive: [Why Coinbase Reported a $667 Million Loss Last Quarter — And Why It Doesn't Mean What You Think](https://siddhants.com/blog/coin-research-202603/asc350-gaap-loss/)

## Developer Platform & Institutional Infrastructure — The Business with No Revenue Line

**Enterprise Clients: 270+ | Government Clients: 150+ | "Other Transaction" + "Other Sub & Services": \$807.7M combined**

JPMorgan. Citi. BlackRock. Stripe. PayPal. Revolut. Shopify. Plus $150+$ government agencies. Brian Armstrong has used the AWS analogy explicitly — *an infrastructure layer that financial services companies plug into*. FY2025 platform metrics suggest the thesis is already materializing: \$$5.2$T in trading volume ($+156\%$ YoY), $9.2$M monthly transacting users, \$$516$B in assets on platform.

The problem for investors is that none of the developer platform revenue has its own line. CDP economics are embedded across "Other Transaction" (\$$252.9$M, $+20\%$ YoY) and "Other Sub & Services" (\$$554.8$M, $+30\%$ YoY) — two catch-all segments with no sub-breakout in any public filing. The models I've come across have no dedicated sub-model for either. At institutional custody fee rates of $0.5$-$1.5$bps on \$$516$B in platform assets, implied custody revenue alone is \$$258$M-\$$774$M annually — a range that substantially overlaps the combined catch-all lines. This is deliberately opaque for competitive reasons.

AWS was also buried inside Amazon's consolidated North America revenue for years before the disclosure that repriced the entire company. That precedent is the relevant analogy — and that disclosure event, when it comes for Coinbase, will be a revaluation catalyst.

**Key catalyst: Separate CDP revenue disclosure. Watch for client-side disclosures from Citi, JPMorgan, BlackRock on digital asset infrastructure spending.**

Deep dive: [Coinbase's Invisible Business: The Developer Platform That 270 Enterprises Are Already Using](https://siddhants.com/blog/coin-research-202603/developer-platform/)

## Coinbase One — Building a Subscriber Base and Cannibalizing Transaction Revenue Simultaneously

**Subscribers: approx. 971K (ATH, 3x in 3 years) | Estimated Annual Subscription Revenue: \$60-350M | Estimated Fee Revenue Foregone: approx. \$81M/year**

Coinbase One reached approx. $971$K subscribers by end of 2025 — all-time high, $3$x growth in three years. The product offers zero-commission trading, enhanced support, and staking rate boosts at \$$4.99$/month base. Revenue is inside "Other Sub & Services" with no public breakout. Estimated annual subscription revenue: \$$60$-$350$M depending on tier mix. The range is wide because the premium tier pricing is undisclosed.

The more important number is what Coinbase One costs. At approx. $971$K subscribers trading approx. \$$500$/month and a $139$bps blended take rate, the annual transaction fee revenue Coinbase does not collect is approximately \$$81$M. That sum grows with the subscriber base. The net P&L effect is positive only if subscriber lifetime value — lower churn, higher on-platform balances, incremental staking, deeper engagement — exceeds the cumulative fee sacrifice. Coinbase has not disclosed the data needed to verify that calculation externally.

Most model coverage I've seen applies a approx. $30\%$ growth rate to the full Other Sub & Services aggregation as a placeholder. That approach can't capture whether subscription growth is accelerating relative to cannibalization, or whether the flywheel retention effects are materializing. The disclosure that would close this gap: separate Coinbase One subscriber count, average revenue per subscriber, and churn rate. Until then, the net P&L analysis above is the best available external approximation.

**Key catalyst: Coinbase One revenue disclosed as standalone line (as previously done with custodial fees) — that disclosure reprices the recurring revenue multiple.**

Deep dive: [Coinbase One Has 971,000 Subscribers: What That Actually Means for the P&L](https://siddhants.com/blog/coin-research-202603/coinbase-one/)

## The Financial Model

My financial model is the analytical foundation for every financial claim above. Built on Bloomberg Terminal standardized data for FY2019A-FY2025A, it extends through FY2030E across a full three-statement architecture.

Download: [COIN Model.xlsx](https://siddhants.com/COIN%20Model.xlsx)

The model contains ten tabs:

- The `Income Statement` covers FY2019A through FY2030E with analyst-added sub-rows for Base chain sequencer revenue inside Other Transaction, Deribit contribution inside Institutional Revenue, and Coinbase One ARR inside Other Subscription & Services. All historical line items are reconciled to Bloomberg standardized data.

- The `Balance Sheet` and `Cash Flow Statement` complete the three-statement model, with cash computed as a roll-forward from CFS operating, investing, and financing activities.

- The `DCF Model` contains a full CAPM-based WACC build (Blume-adjusted beta = 2.80x, WACC = 14.3%), FCFF projections through FY2030E, terminal value via both Gordon Growth (g = 6%) and Exit EV/EBITDA Multiple (18x) methods, two 5x5 sensitivity tables, and a peer multiple calibration panel (CME, ICE, NDAQ, HOOD).

- The `Assumptions` tab is the single source of truth for all driver inputs: macro/market assumptions (FFR path, ETH yield curve, BTC price), revenue line-item drivers (volumes, take rates, USDC balances, staking custody), OpEx margin assumptions, D&A schedules, balance sheet items, and cash flow assumptions. Every forecast cell in the three statements references this tab.

- The `USDC Scenarios` tab maps three revenue-sharing scenarios (Base: current 100%/50% terms, Bear: renegotiated 80%/40%, Stress: no off-platform share) across three Fed Funds rate paths (4.0%, 3.5%, 3.0%). Range of outcomes: \$$570$M to \$$1,480$M — a \$$910$M spread on a single revenue line.

- The `Deribit Acquisition` tab contains the acquisition summary, purchase price allocation, and annual amortization schedule FY2025A through FY2032E — including the \$$194$M annual GAAP drag beginning FY2026E.

- `Supporting Schedules` consolidate the Debt Schedule, PP&E Schedule, Share Count, Working Capital, and Tax Schedule into a single reference tab.

Deep dive: [Inside the Model: A Walk-Through of the Coinbase FY2019-FY2030 Equity Research Framework](https://siddhants.com/blog/coin-research-202603/model-walkthrough/)

## Key Risks

1. **`USDC` contract renewal:** A renegotiation to 80%/40% sharing terms removes \$$220$-$300$M from FY2026 revenue in a single event, depending on the rate path. This is the only risk in the model that is fully within a counterparty's control rather than Coinbase's.

2. **CME Group institutional competition:** The May 29, 2026 24/7 crypto options launch introduces direct competition to Deribit's institutional franchise. A 10-15% institutional flow shift to CME in H2 2026 implies \$$40$-$65$M of revenue downside versus current estimates.

3. **`ETH` staking yield compression:** 20-30bps of annual yield dilution as the validator set grows. Requires `ETH` price appreciation to offset on a dollar-revenue basis. Bear case: compound compression plus flat `ETH` prices.

4. **ASC 350-60 GAAP volatility:** Crypto price moves will continue to produce large GAAP swings that bear no relation to operating performance. The risk is misinterpretation, not operational deterioration.

## Disclosure

The views expressed here are my own personal opinions. This is **not investment advice** and should not be relied upon as such. Nothing here is connected to, endorsed by, or written on behalf of Rosenblatt Securities.

---

Source: https://siddhants.com/blog/coin-research-202603/
About the author: https://siddhants.com/about.md
Site index for agents: https://siddhants.com/llms.txt