Long COIN — The Stablecoin Flywheel Is Doing The Work The Street Isn't Modeling

The lazy read on COIN is that it's a leveraged bitcoin call. It was, three years ago. It isn't now, and the price behavior versus BTC over the last two quarters is telling you the market is starting to figure that out. I'm long.

The setup is simple: a fee business that is no longer the only story, a stablecoin annuity that pays whether crypto goes up or down, a base-layer L2 (Base) that is now a real fee machine, and an institutional custody book that grows every time a new ETF issuer needs a US-regulated custodian. Underneath all of it is a float — client cash and USDC reserves — that COIN earns on directly or through the Circle revenue share.

This is the piece that says why the flywheel works, where I'm long from, what breaks the trade, and how it ends.

Retail trading is still the highest-beta line in the P&L, but it is no longer the dominant one. Subscription and services — which is where stablecoin revenue, staking, custody, interest income, and Base sit — has been the majority of gross profit for several quarters running. That is the whole trade in one sentence.

Once your non-transaction revenue stack out-earns your transaction stack, you rerate. You stop trading on BTC beta and you start trading on the stability of the float and the moat around the custody book. That is what happened to the exchanges in the '90s, that is what happened to Schwab in the 2010s when rate income became the anchor, and that is what is happening here.

The Street model I keep seeing still runs COIN as a bitcoin ADR with some extra businesses attached. That is the mismodel. The correct frame is: it's a rates-plus-fees fintech with a crypto-native exchange as its highest-beta segment.

The live tape is at the top of this page — COIN, BTC, ETH, HOOD, and CRCL update in real time so the levels below stay honest against the print, not against a stale screenshot.

Structural levels I'm playing off, top down:

• Prior all-time high shelf — the ceiling from the last leg. First rejection is a fade, second reclaim is a launch. Trade the second touch, not the first.

• Post-earnings volume node — the fat part of the last gap. This is where mean-reversion algos anchor. Loses this on volume and the tape hunts the shelf below.

• Consolidation shelf (my entry zone) — the range low that has been defended three times since the last print. Layered long from here with adds on flushes.

• Swing low / mental stop — one full ATR below the shelf. Break on a daily close with volume is the invalidation.

• Hard stop line — two consecutive daily closes below the swing low. That is the tape telling me the flywheel narrative isn't sticking and I flatten without argument.

Correlation read: 30-day rolling correlation to BTC is running mid-band, well off the 0.85+ prints that defined 2021-2023. Correlation to HOOD and to a basket of rate-sensitive fintechs is climbing. That is the rerate showing up in cross-asset behavior before it shows up in the multiple.

The line items I underwrite this trade against. These are the levers — if any one of them breaks the direction I model, the thesis weakens; if two break, I flatten.

• Subscription & Services share of gross profit — running majority. Track this every print. As long as it holds above 50% and grows sequentially, the fintech rerate story is intact.

• USDC revenue-share line — the direct read on the Circle relationship. Grows with USDC market cap and with short-rate income on reserves. Watch reserve composition disclosures for duration and quality.

• Custody AUC (assets under custody) — measured in tens of billions across the spot BTC and ETH ETF complex. Basis-point fees on a base this large is annuity income. Growth here is a leading indicator for the next quarter's services line.

• Base sequencer revenue — the highest-growth line, smallest today, biggest optionality. If it gets broken out cleanly on a print, that is the SOTP rerate catalyst.

• Retail transaction take rate — the highest-beta line. Volatile, mean-reverting, and no longer the anchor. I want this line to surprise upside, but I don't need it to.

• Interest income on customer cash — direct function of the Fed path and average customer balances. Compresses on cuts, expands on volume.

• Adjusted EBITDA margin — the summary stat. Track quarter-over-quarter direction more than absolute level. Expanding into a mixed tape is the tell.

• Operating cash flow — the reality check. Non-GAAP moves around; cash conversion doesn't lie. Rising through the last several quarters.

• Buyback capacity — quiet lever. Every dollar of retained cash the company can deploy at a compressed multiple is accretive to per-share value. Watch the authorization line on the print.

These are the numbers you can verify off-chain analytics dashboards and ETF filings. They are not marketing — they are the substrate under the thesis.

• Spot BTC ETF AUC — aggregate US spot bitcoin ETF assets sit in the mid hundreds of billions. COIN custodies the majority of that book. Every basis point of custody fee on that AUC prints, quarter after quarter, whether BTC is at a new high or chopping.

• Spot ETH ETF AUC — smaller than BTC but growing on a higher slope since the staking-approved products came online. Custody wins on the ether side stack on top of the BTC book without cannibalizing it.

• USDC circulating supply — the top-line on the stablecoin annuity. Rising into a post-GENIUS-Act regulatory regime as banks, brokers, and payment networks route dollars onto compliant on-chain rails. Every dollar of new USDC float is a dollar earning short-rate income.

• USDC on-Coinbase share — the portion of USDC that lives inside the COIN app and on Base. Higher share equals higher net take from the Circle revenue split. Track this every quarter — it is the cleanest read on ecosystem lock-in.

• Base TVL and daily active addresses — Base has been running as a top-three L2 by both metrics, with sequencer revenue climbing on transaction volume rather than on token speculation. That is the boring, durable version of L2 economics.

• Base sequencer revenue — book-of-record fee income the market doesn't yet price. When this line gets its own row on a print, the SOTP crowd shows up.

• Institutional custody client count — the sticky book. Every new ETF sponsor, every new pension trustee, every new corporate treasury that names COIN as custodian is a client with high switching costs and long-dated fees.

• Staking AUC (regulated jurisdictions) — grows with ETH, SOL, and adjacent PoS assets under custody. Yield-split revenue is a second income stream on top of custody basis points.

• Insurance and cold-storage footprint — the moat under the moat. The operational infrastructure that makes COIN the default custodian is expensive to replicate and gets more expensive every year regulators tighten disclosure.

Cross-check these against Circle's transparency reports, the ETF sponsor 10-Qs, and the L2 analytics dashboards before every earnings print. If two or more roll over quarter-over-quarter without a clear macro explanation, the flywheel is decelerating and I trim.

The Circle relationship is the single most under-appreciated line on the P&L. COIN's revenue share on USDC reserves is a direct participation in the float income earned on the reserve assets. When rates are high, this line prints. When rates come down, it compresses on rate but expands on volume because lower rates and clearer stablecoin rules push more dollars onto the rails.

The GENIUS-era stablecoin regime in the US has done two things at once. One, it forced issuers to disclose reserves and hold them in the highest-quality short-dated paper — which is exactly what generates the float income. Two, it opened the door for banks, brokers, and payment networks to treat compliant stablecoins as settlement instruments rather than exotic instruments. Every additional dollar of USDC in float is a dollar earning short-rate income, and COIN's share of that income shows up as high-margin, low-capex revenue.

You do not have to believe crypto goes up to own this line. You have to believe that the number of dollars parked in compliant, dollar-denominated on-chain instruments keeps growing. That is happening.

Base was launched as a strategic move to own the developer surface. It has quietly turned into a real fee business — sequencer revenue, plus the derivative revenue that comes from being the default on-ramp between the COIN app and the on-chain economy.

The Street either ignores Base in the model or pencils in a token that doesn't exist. Both are wrong. The correct way to model Base is as an internal payments rail: every transaction that runs through it monetizes at the sequencer, and every user COIN routes onto Base is a user who is now inside the COIN stack rather than a competitor's.

It is not the biggest revenue line, but it is the highest-growth one and the one with the best long-duration option value. If the market ever puts a real multiple on Base the way it puts a multiple on payment rails, that is a rerate catalyst all by itself.

COIN custodies for the majority of US spot bitcoin ETFs and a growing share of the ether ETFs. That is a reflexive moat. Every new ETF issuer that comes to market has to pick a US-regulated custodian with the balance sheet, insurance, SOC audits, and operational track record to satisfy the trust sponsor and the SEC's disclosure regime. That short list is short, and COIN is on top of it.

Custody revenue is measured in basis points. Basis points on tens of billions of ETF AUM is a real number. Basis points on hundreds of billions — which is where the aggregate spot crypto ETF complex is going — is a business that on its own is worth a real chunk of COIN's current market cap.

This is the segment I like most because it is the least tied to price. ETF AUM can grow while BTC chops. Custody fees are AUM-based, not volume-based. That is annuity income in a name the market still treats like a beta.

Common pushback: 'if the Fed cuts, USDC float income compresses.' True on rate, false on the total. Two offsets. One, float grows into a lower-rate regime because stablecoins are more attractive relative to bank deposits when banks pull back on yield. Two, retail transaction revenue and Base activity both benefit from lower real rates and a broader risk-on tape.

The 2024-2025 tape was the stress test. Rates high, retail cool, and COIN still expanded gross profit because the annuity lines carried it. Now bring rates lower into 2026 with retail waking up and the annuity lines still in place. That is operating leverage the model doesn't have penciled in.

The regime that kills this trade is not lower rates. It is a rate regime plus a real regulatory reversal that unwinds the stablecoin framework. That is a tail risk, not a base case, and I'll trade it if it shows up.

Entry: I'm layering long on any close that holds the prior consolidation shelf, adding on a reclaim of the volume node from the last earnings gap. I'm not chasing green — I'm building into red days and letting the tape come to me.

Invalidation: a weekly close back inside the pre-breakout range, on volume, with no reclaim inside five sessions. That would tell me the flywheel narrative isn't sticking and the tape wants to trade COIN as a BTC proxy again. I don't fight that; I flatten and wait.

Stop: mental, below the last swing low that defined the shelf. Hard stop only if the tape rips through it and prints two consecutive daily closes below. Opening prints in COIN can be noisy on ETF flow days and I don't want to donate to a bad tick.

Sizing: swing size, not concentration size. COIN can move ten percent on a headline and I want to be able to add on the flush, not defend on it.

Stablecoin volume prints. Every quarter of expanding USDC float is a direct read-through to the revenue-share line. This is the number I care about most on the print.

Base fee take rate. If Base sequencer revenue is disclosed cleanly and grows quarter-over-quarter, that is the rerate catalyst that pulls the sum-of-the-parts crowd off the sidelines.

Institutional custody wins. Every new ETF that names COIN as custodian is a data point the annuity book is compounding. Watch the ether-linked and spot-solana ETF pipeline; those custodian slots get announced before the funds trade.

Rate cuts with a still-warm retail tape. That is the combination that lights both sides of the P&L at the same time.

Trend end: the market fully rerates COIN as a fintech, the annuity multiple gets priced, and I sell into strength as the multiple crosses into the range where the risk/reward flips. That is the boring, good ending.

Blow-off: a BTC parabolic pulls COIN with it into a vertical run that decouples from the fundamentals. I take profit in tranches on the way up and don't argue with a gift.

Liquidation: a stablecoin regulatory reversal, a custody-side operational event, or a broader risk-off unwind that breaks the shelf. I'm flat before the second daily close below the stop. No hero trades.

COIN is a fintech with a crypto-native exchange bolted to the front and a stablecoin/custody/L2 annuity bolted to the back. The Street is still modeling it like it's 2021. That mismatch is the trade.

I'm long into weakness, stopped below the shelf, adding on stablecoin-print catalysts and ETF-custody wins. I don't need bitcoin at a new all-time high to make this work. I need the flywheel to keep spinning. It is.

Nothing but death stops the guy who trades the business instead of the ticker. See you at the next earnings print.

Everything on this page is my opinion based on publicly available disclosures and my read of the tape as of publication on July 20, 2026. It is not investment advice and not a recommendation to buy, sell, or short any security. I may be long COIN at time of publication and my position can change without notice. Prices and levels discussed can invalidate the same session. Verify independently before risking capital.