Coinbase is the dominant US crypto exchange with real network effects and a 26% 10-year ROIC, but its earnings are tied to a volatile, regulation-sensitive asset class no one can underwrite for a decade. The IV gap is wide, but the underlying earnings stream is not.

Plain English

Coinbase runs the biggest US-regulated crypto exchange. People pay fees to buy and sell Bitcoin and other tokens. Coinbase also earns interest on customers' stablecoin balances and holds Bitcoin for the new Bitcoin ETFs. The business prints money in good crypto years and loses money in bad ones. Earnings swing 5x or more between cycles. Whether it makes money long-term depends on three things nobody can predict: will crypto stay popular, will regulators stay friendly, will competitors not crush prices. That's three guesses in a row, which Munger would say is too many. Real but fragile.

Thesis

Coinbase Global (COIN) is the largest US-regulated cryptocurrency exchange, deriving the majority of revenue from retail trading fees, with a fast-growing 'Subscription & Services' line anchored by USDC interest income, staking, and Coinbase One. The bull case rests on three pillars: (1) regulatory moat — COIN is the cleanest publicly-listed crypto venue, custodian to most US spot Bitcoin ETFs, with Trust company status in NY and a long compliance track record; (2) optionality — Base L2, USDC economics shared with Circle, and institutional prime brokerage; (3) operating leverage — owner-earnings TTM of $3.58B on a still-modest cost base.

The scorecard tells two stories at once. ROIC 10y avg of 26.2% and FCF conversion of 2.03x scream 'wonderful business.' Yet the scorer flags 'base CAGR clamped from 71.7% to 14.0%' and 'insufficient history' — meaning the 26% ROIC was earned across one full crypto cycle (2021 mania, 2022 collapse, 2024–25 reflation), not a durable through-cycle figure. P/E TTM of 20.2 looks cheap versus a 28.5 ten-year average, but TTM earnings sit near a cyclical peak fueled by ETF-driven trading volumes and elevated rates juicing USDC yield income.

The stated IV range is $310–$483 with base $448.76 against a $191.25 price (px/IV = 0.43). On the math alone, this is a 57% discount to base IV. But the IV is built on owner-earnings that may halve or worse in the next bear leg. Reverse-DCF implies only 2.4% growth — a low bar — yet earnings volatility, not the long-term mean, is what sinks an investor. At $191, owning COIN is not buying a $4B owner-earnings stream; it is buying a call option on crypto adoption with negative theta from share dilution (+31.6% over 10 years). Pass. Wait for either much lower price or 3+ years of demonstrable through-cycle earnings stability.

Moat

Coinbase has a real but fragile moat. I work the five sources, then issue a verdict.

1) Network effects — narrow and asymmetric. Coinbase has the largest US retail crypto user base (~100M+ verified users historically) and is the custodian to most US spot Bitcoin ETFs (BlackRock IBIT among them). On the institutional side, deeper books attract market makers, which tighten spreads, which attracts more order flow — a classic two-sided liquidity flywheel. But unlike the NYSE or CME, crypto liquidity is globally fungible and migrates quickly to whoever offers cheaper fees (Binance, Kraken, OKX, Robinhood, Hyperliquid for derivatives). When fees fell, Coinbase lost retail volume share. The network effect is real domestically but does not 'travel' the way Buffett describes Coke and AMEX traveling [1].

2) Switching costs — narrow. Custody of crypto creates real friction (private keys, tax lots, transaction history). Coinbase One subscribers (~$30/mo flat-fee trading) and Prime institutional clients have higher stickiness. But transferring crypto is one on-chain transaction; it is materially easier than moving a brokerage IRA. Power users routinely arbitrage across venues. Coinbase's 'switching cost' is mostly inertia plus tax friction, not contractual or workflow lock-in.

3) Intangibles (brand + regulatory) — this is the strongest moat element. Coinbase is the most-trusted crypto brand in the US and the only major venue without a serious regulatory blow-up. It carries NY BitLicense, MiFID II authorizations in Europe, MiCA registration, and was the lead survivor of the SEC enforcement era of 2022–24. After FTX, 'Coinbase the public, audited, US-regulated venue' became a flight-to-quality destination. This is the closest analogue to Buffett's American Express insight [1]: 'unquestioned financial trust.' But trust in a custodian of speculative assets is not the same as trust in a global payments network. One major hack, one SEC consent decree, or one Trust-Company bank-run-style event can break it overnight.

4) Cost advantages — none structural. Coinbase is not the lowest-cost venue. Binance and Robinhood undercut on fees. Coinbase's cost structure is heavy in compliance, security, and US labor — these are necessary expenses to maintain the regulatory moat above, not durable advantages. USDC reserve income is split with Circle; Coinbase does not own the rails.

5) Pricing power — narrow and eroding. Retail take rates have compressed from ~1.5%+ to closer to 1% as competition arrived. Institutional take rates are basis points. The flat-fee Coinbase One tier exists because the per-trade pricing was unsustainable. The fee compression here mirrors brokerage commissions post-Schwab 2019 — likely directional.

Competitor stress test. Give a competitor $10B and 5 years. Could a well-capitalized incumbent (Robinhood, Fidelity, Schwab, BlackRock-affiliated venue) seriously erode COIN's retail share? Yes — Robinhood already proved it, going from negligible crypto presence to >$50B quarterly volumes by late 2024. Could a foreign exchange (Binance.US, OKX, ByBit) take institutional flow if regulation eases? Yes. The moat survives only if regulatory clarity continues to favor incumbent compliant venues. That is a regulatory-moat-by-permission, not a moat-by-economics.

Erosion risk: HIGH on a 10-year basis. Crypto fee compression is the base case; the only question is speed. Coinbase's bet is that take-rate compression is offset by volume growth, USDC interest income (regime-dependent), staking revenue (regulation-dependent), and Base L2 sequencer fees (early). None of these are Coke selling syrup [4].

Munger would say: this passes the 'too-hard' filter only if you believe (a) crypto adoption sustains, (b) regulation stays favorable, and (c) Coinbase wins the regulated tier. That's three serial probabilities — a low joint outcome.

Moat verdict: NARROW

Management

Brian Armstrong (CEO, co-founder, ~50% voting control via supervoting Class B) has been clear, public, mission-driven, and combative on regulation — he sued the SEC and won meaningful concessions, then capitalized on the post-2024 regulatory thaw. He is unusually candid in shareholder letters about being 'crypto-first' and willing to absorb short-term volatility. That is closer to a founder-owner than a hired CEO, and on dimensions of communication and conviction he scores well.

The five capital-allocation choices:

1) Reinvestment. Coinbase reinvests heavily in product (Base L2, derivatives, international, Coinbase One). Reported ROIC 10y average is 26.2% — high — but the scorer flags 'insufficient history' and 'base CAGR clamped from 71.7% to 14.0%'. Translation: reinvestment looks brilliant in the up-cycle, looks like burning money in the down-cycle. ROIIC over 5 years is unavailable in the scorecard, which is itself telling — incremental returns are unstable. Buffett's lesson is that high ROIC matters only if it is durable [4]; here the durability is unproven.

2) Acquisitions. Coinbase has been an active acquirer (Bison Trails, FairXBV/derivatives exchange, BUX Europe, Deribit announced). The Deribit-class deals are large and price-sensitive — paying premium multiples for derivatives access at a cycle high is exactly the pattern Buffett warns about [3]. Track record is mixed; integration risk is real.

3) Debt. Net debt to EBITDA of −1.68x (net cash). Balance sheet is fortress-like (score 20/25 from scorecard). Convertible notes outstanding are manageable. Plus on this dimension.

4) Buybacks. Coinbase has authorized buybacks but has been a net issuer. Share count is up 31.6% over 10 years per the scorecard — that is significant dilution, predominantly stock-based compensation (SBC). Buffett's repurchase test is 'price-dependent — sensible at a discount to business-value, stupid at a premium' [1]. COIN's de-facto buyback (offsetting SBC) happens at all prices — not price-disciplined. Capital-allocator demerit.

5) Dividends. None. Appropriate for a high-reinvestment business but worth noting that without dividends, the entire return depends on terminal value — extending the duration of the bet.

Communication quality. Armstrong's quarterly letters are above average for transparency on segment economics (USDC income, staking, transaction take rates) and on long-term goals (economic freedom mission). He has been wrong publicly (e.g., aggressive 2021 hiring later reversed in painful layoffs) but has owned the mistakes. CFO Alesia Haas is credible. Communication is a B+.

The dilution problem dominates. A 31.6% share count increase over a decade — almost entirely SBC — means a sizable share of operating income leaks to employees rather than owners. At ~$3.58B owner earnings and ~250M shares, every 1% of dilution costs ~$36M of present-value owner earnings. SBC is a real expense regardless of cash flow optics; the FCF conversion of 2.03x partially reflects that working-capital and SBC adjustments flatter cash flow.

Net grade. Founder commitment, balance sheet, and communication are A-tier. Reinvestment quality is B at best (high ROIC, unproven durability). M&A discipline is C. Buyback discipline is C-D (chronic dilution). Weight to capital deployed favors the mediocre dimensions.

Capital allocator: B-

Industry

Porter's Five Forces applied to centralized cryptocurrency exchanges:

1) Rivalry — HIGH and intensifying. Coinbase competes domestically with Robinhood, Kraken, Gemini, Fidelity Digital, and increasingly with bank-affiliated venues (post-clarity). Globally, Binance dwarfs everyone in volume; OKX, Bybit, Bitget, and Hyperliquid (perps) are aggressive. Decentralized exchanges (Uniswap, dYdX, Hyperliquid) take ~25–30% of spot volume in some weeks. Fee compression is structural — retail take rates have already fallen materially and continue to.

2) Buyer power — MODERATE-HIGH. Retail buyers individually have low power but collectively are price-sensitive and route to the cheapest venue with adequate trust. Institutional buyers (market makers, hedge funds, asset managers) negotiate hard and operate across multiple venues. ETF issuers (BlackRock, Fidelity) have meaningful leverage as Coinbase custody clients — concentration risk on the custody line.

3) Supplier power — LOW-MODERATE. Liquidity providers (market makers like Jane Street, Cumberland, Wintermute) have some power because they can route flow elsewhere. Banking partners are critical (post-Silvergate/Signature), and the universe of crypto-friendly banks is small — that is genuine supplier power but improving with regulatory thaw.

4) Threat of new entrants — HIGH. Robinhood went from zero to a top-three US crypto venue in three years. Schwab, Fidelity, and Bank of America-class incumbents could enter at scale. Stablecoin issuers (Circle, Tether) and prime brokers (Galaxy, FalconX) encroach on Coinbase territory. The regulatory moat is the only meaningful entry barrier and it is permission-based.

5) Threat of substitutes — HIGH and underappreciated. Self-custody wallets (MetaMask, Phantom, Rabby, hardware wallets), DEXes, and L2 native trading reduce dependence on centralized exchanges for active users. For passive exposure, spot ETFs (which COIN custodies but does not own the fee economics of) substitute for direct exchange holdings. CME futures substitute for derivatives. Yield-bearing stablecoins encroach on USDC-on-Coinbase economics.

Value pool location and trajectory. The crypto value pool is large and growing in absolute terms, but the share captured by centralized US exchanges is shrinking. The most profitable pockets — derivatives, on-chain liquidity, stablecoin issuance, custody for ETFs — each have stronger competitors than spot retail trading does. COIN is best positioned in custody for ETFs and US-regulated derivatives (post-FairXBV), but those are lower-margin businesses than peak-2021 retail spot.

Cyclicality dominates structure. Even if Coinbase wins its share fights, the entire industry's revenue moves 3-5x peak-to-trough with crypto prices. Operating leverage cuts both ways: 2021 saw $7.8B revenue / $3.6B net income; 2022–23 saw <$3.2B revenue and a net loss. The industry does not have predictable customer demand the way Coke or AMEX do [1][4]; it has reflexive demand tied to asset price expectations.

Regulatory regime-change is force-of-nature. A change in administration, a single Supreme Court decision on securities classification, a stablecoin bill that disadvantages USDC, an OCC ruling on custody — any of these can permanently re-rate the industry's profit pool up or down by 30%+. This is not 'normal' industry risk; it is policy risk, and Buffett-Munger explicitly avoid businesses where 'predicting regulatory outcomes' is required (the methodology's circle-of-competence test).

Industry Verdict: Average — large pool, but high competition, high cyclicality, high regulatory dependence. The industry rewards survival more than steady compounding.

Inversion

I am now the short-seller. I have studied this company, and here is the bear case.

The single event that kills this. The kill-shot is a coordinated regulatory + competitive shift: a Trump-or-successor administration walks back crypto-friendly executive orders, the SEC re-asserts that staking is a security and that USDC interest sharing is an unregistered investment contract, and simultaneously Robinhood + a major bank consortium launch zero-fee crypto trading bundled with brokerage. Coinbase's retail take rate falls from ~80bps blended to <30bps, USDC interest income (which I estimate is ~25-30% of run-rate operating profit) is restructured downward, and staking revenue is curtailed. The stock would not need a 'kill event' — a slow regulatory tightening combined with fee compression in the next bear cycle is sufficient.

Why the moat is narrower than bulls think. Bulls cite 'regulated US incumbent' as a wide moat. It is not. (a) Foreign competitors are returning to US after enforcement settlements; Binance, Kraken, and OKX all have paths back. (b) Self-custody and DEX volumes are structurally rising — Hyperliquid did >$300B in monthly volumes in 2024 with no regulatory entity. (c) Bank-affiliated venues post-stablecoin-bill will out-distribute Coinbase by leveraging existing customers. (d) The 'flight-to-quality' premium that COIN earned post-FTX fades as memory fades — by 2027 nobody will be paying Coinbase a trust premium. (e) Coinbase Trust Company status is a regulatory permission slip, not an economic moat; permission slips can be modified.

Why management is worse than it appears. Armstrong is a true believer, which is not the same as a great capital allocator. Three specific concerns: (1) chronic dilution of 31.6% over 10 years per the scorecard — most of it SBC at high prices, the worst form of capital allocation per Buffett [1]. (2) The 2021 hiring spree (5,000 employees in 18 months, then 25%+ layoffs in 2022–23) reflects pro-cyclical capital deployment — the opposite of 'be greedy when others are fearful.' (3) The Deribit-class derivatives M&A at peak-cycle pricing risks repeating 2021 mistakes. (4) Insider selling has been heavy at every cycle peak; Class B supervoting structure means founders can deploy capital however they like without check.

What bulls are extrapolating that won't hold. (a) USDC reserve income at current rates: this is a ~$700M-1B+ run-rate line that exists because Fed funds is ~5%. At 3% rates, this halves; at 1.5% rates, it goes to zero. The bull model assumes structurally high rates persist — a meaningful macro bet. (b) Spot ETF custody fees: Coinbase custodies most US BTC ETFs at ~10bps — competitive intensity from BNY, State Street, Fidelity will compress this to 2-3bps. (c) Trading volumes in 2024–25 are inflated by ETF arbitrage, retail FOMO from BTC at all-time highs, and meme-coin cycles. Through-cycle volumes are likely 40-60% lower. (d) International expansion margins assume regulatory acceptance comparable to US — naive in MENA, APAC, and post-MiCA EU. (e) Base L2 sequencer revenue: currently substantial, but L2s are commoditizing rapidly and Ethereum L1 fee structure changes (EIP-4844 already pressured L2 economics) can compress this further.

Valuation trap (multiple compression / regime change). Current $191 / $3.58B owner-earnings ≈ 13x owner-earnings cycle-peak. In a normal-cycle year, owner earnings are likely $1.5–2B (the trough of 2022 saw losses; the long-run mean is closer to $1.8B). At trough multiples for a cyclical business — 8-12x trough earnings — the stock is worth $60-90, not $310 (low IV). The IV range in the scorecard is anchored on a clamped-down 14% growth rate that itself was clamped from 71.7% — an admission that the actual data is too volatile to project. The reverse-DCF implied growth of 2.4% looks easy to clear, but the bear point is that owner earnings could be NEGATIVE in a meaningful share of years, making any constant-growth DCF misleading. The 'cheap on trailing P/E of 20.2x' is a classic value trap — cyclical lows in P/E often precede earnings collapses, not multi-baggers.

If I am right, the stock could be worth $80 within 3 years.

Lollapalooza Bias Check

The biases active in me as I write this:

Recency bias (HIGH). I am writing in 2026 after a 2024–25 crypto bull run. Spot ETFs launched, regulation softened, BTC made all-time highs, and Coinbase's earnings exploded. My recent-data window is unrepresentative. The 26.2% ROIC cited in the scorecard was earned across one cycle — recency makes it feel structural.

Anchoring (MEDIUM-HIGH). The scorecard hands me an IV range of $310-483 against a $191 price, and I find myself anchoring on the 57% discount as if it were a margin of safety. The scorer notes ('base CAGR clamped from 71.7% to 14.0%', 'insufficient history') explicitly warn me the anchor is soft. I should treat IV as a wide range with the bottom potentially much lower than $310, not as a reliable target.

Authority bias (MEDIUM). The scorecard composite of 78 looks high. The composite is a deterministic Python score; it does not know that crypto-exchange owner-earnings are not Coke owner-earnings. I should not let the score do the qualitative work for me.

Confirmation bias (LOW-MEDIUM). I started skeptical (crypto = too-hard) and am finding mostly confirming evidence. I corrected by genuinely steel-manning the bull case in the thesis section: real network effects, regulatory clarity, fortress balance sheet, owner-operator. The bull case is real; my conclusion is that it is not enough at this price.

Social proof (LOW). Most equity analysts publishing on COIN are bullish; this would normally pull me. I notice the pull and discount it — analyst coverage of cyclicals at peaks is structurally bullish.

Deprival super-reaction (LOW-MEDIUM). There is a fear of missing out — if crypto adoption goes parabolic, COIN could 3x from here. This is a call-option payoff structure, and Munger warns against confusing call-option payoffs with compounder ownership.

Incentive bias (relevant for management, not me). Armstrong's incentives via supervoting Class B + heavy SBC compensation tilt toward growth-at-any-price and dilution. I should weight stated guidance accordingly.

Net. Recency and anchoring are pulling me toward 'cheap.' The discipline of inversion and the circle-of-competence test pull me back toward 'too volatile / too policy-dependent.' The corrected position is closer to Avoid/Too Hard than to Buy.

10-Year Outlook

Same fundamental business model in 10 years? Probably no. Today Coinbase is primarily a retail spot-trading exchange + custodian + USDC partner + L2 operator. In 2036, if the business survives strongly, it is likely a regulated stablecoin/payments rail, custodian to institutional digital-asset products, and possibly an L2 / on-chain infrastructure provider — much less of a fee-per-trade exchange. The shape is changing.

Customer base larger? Plausibly yes in users, but per-user economics likely lower. Crypto adoption probably reaches 30–50% of US adults, but trading frequency and take-rate per user fall as the asset class normalizes. Net effect on revenue per user is ambiguous.

Profit per customer higher? Likely lower on a steady-state basis. Spot trading fees are commoditizing. Higher-margin products (staking, derivatives, USDC interest sharing) are policy-dependent — they could be much bigger or near-zero by 2036.

Moat wider? Regulatory moat narrows as regulation clarifies and competitors enter compliantly. Brand moat may persist if no major incident occurs. Network-effect moat narrows as DEXes and self-custody mature. Net: moat NARROWER in 10 years, not wider — the opposite of what a Buffett compounder needs.

Single biggest threat? Not a competitor or a hack. The single biggest threat is regulatory regime change combined with rate-cycle reversal — i.e., USDC interest income normalizes downward at the same time as a crypto bear market and re-tightening of crypto regulation. That triple combination compresses every revenue line simultaneously. Secondary threat: a security breach or bank-run-style event that breaks the trust-premium permanently.

Confidence. Buffett's standard is being able to look at the business in 10 years and recognize the same fundamental shape. Coke today and Coke in 1986 — same. AMEX today and AMEX in 1986 — same [1]. Coinbase today and Coinbase in 2036 — likely meaningfully different in business mix, regulatory regime, fee structure, and competitive position. That is the definition of a low-confidence forecast. Per the methodology's own rule, low confidence implies 'Too Hard.' I am one notch above that — Avoid — only because the qualitative bear case is concrete enough to act on, not because the long-term forecast is reliable.

CONFIDENCE: low

Position Guidance

• Recommendation: Avoid
• Conviction: medium
• Target buy price: $95 (≈ 30% margin of safety to the lower bound of likely through-cycle IV; ~50% below current price)
• Target trim price: $310 (the scorecard low IV; above this, even bear-case IV is exceeded relative to a reasonable through-cycle estimate)
• Position sizing: Zero for compounder-style portfolios. If held for trading/optionality, cap at 1-2% of portfolio with explicit cycle-trim discipline. Not eligible for 'core compound for decades' bucket.
• Why not 'Too Hard': the qualitative bear case is concrete enough to take a side. Confidence on the short side is medium; confidence on the long side is low. Net = Avoid at this price, watch for ≥50% drawdown to revisit.