PM RESEARCH MEMO: Coinbase Global, Inc. (NASDAQ: COIN)

Date: 2026-10-01 | Classification: Research. Not financial advice, not a trade recommendation. | Frame: Constituent of Jordi Visser / 22V Research "Mag 1 Billion" AI-crypto index (46 names, launched 2026-09-23, bucket "Exchange, Brokerage & Market Structure"). The index composition and launch date come from the requester's screenshot. I could not find them in a live web search [External check needed].

Attribution legend: [S#] = sourced claim, with the number pointing to the source list at the end. (Inference) = analyst reasoning. [External check needed] = from background knowledge, no live source found. (unverified, no live source) = a figure I could not source.

1. EXECUTIVE SUMMARY

• Price and setup. The last close was ~$186.41, a market cap of ~$50.1B. The stock is down ~43.6% over 52 weeks and ~58% below its $444.65 ATH (2025-07-18). Trailing EPS is negative (−$3.69) [1]. BTC is ~$83.5K, ~34% below its $126K ATH [18]. COIN is trading as a crypto-beta drawdown name, not as an agent-rails name.
• Q2-26 was a miss, but the mix improved. Revenue was $1.22B against $1.29B consensus. Net loss was $359.5M, adjusted EBITDA +$207.8M (14th straight positive quarter). There was $52.4M of restructuring, tied to ~700 layoffs. The stock fell 5.4% after hours, to $154.68 [2][3].
• Subscription & Services is now ~48% of revenue. It came in at $555.1M, against transaction revenue of $599.2M [2][4]. Stablecoin revenue was $292M in the quarter (~$1.17B annualized), on ~$20B average USDC held in Coinbase products [3][4]. This is the line that maps most directly to Visser's thesis.
• Coinbase owns the economics of USDC distribution, and the deal is renewed. Circle renewed the revenue-share through 2029 on existing terms (confirmed on Circle's 2026-08-05 call). Coinbase keeps 100% of the reserve income on USDC held on its platform and 50% of the residual elsewhere [9][10]. USDC supply was ~$74.6B in early September 2026 [11].
• COIN owns the agent-payment standard. x402 launched in May 2025 and passed 100M machine-to-machine transactions in ~9 months [7]. Coinbase Business began accepting agent payments on 2026-07-23 and calls agentic payments one of its "most high-conviction bets" [5][6]. "Coinbase for Agents" launched 2026-06-11 and lets ChatGPT and Claude connect to user accounts [8]. 53% of Base documentation traffic came from AI agents starting June 2026 [5].
• What headlines miss: agent volume is real in count but trivial in dollars. x402 logged 3.1M operations worth ~$1.2M in May 2026, about $0.39 per operation [7]. Today's fees on that flow round to zero against a $50B market cap. (Inference) COIN captures agent value through the float, not the fee. USDC balances that agents and merchants hold on Coinbase rails earn reserve income. So the thing to track is "agent-held USDC balances," not "x402 volume."
• Second non-obvious angle: regulation cuts both ways on the key line. The CLARITY Act is stalled over stablecoin yield/rewards, with banks opposed [14][15][17]. Coinbase reportedly told Senate offices it cannot support the latest language [14]. Galaxy cut its 2026 passage odds to 60% [16]. A bill that bans rewards could raise the economics Coinbase keeps (it pays out less) or lower them (USDC sitting on Coinbase becomes less attractive). The direction is unresolved [External check needed].
• Derivatives are a rail-adjacent bet. On 2026-05-28/29 the CFTC gave Coinbase Financial Markets a no-action letter, making it the first US FCM allowed to offer global crypto perpetuals, routed to Deribit liquidity [12][13]. Perps are the 24/7 instrument agents are most likely to trade programmatically. Prediction markets grew 106% q/q to >$100M annualized revenue [2].
• Operating-leverage problem. The FY26 opex guide is $4.34-4.60B [2], against a Q2 revenue run-rate of ~$4.9B. (Inference) Margin depends almost entirely on crypto prices and volume recovering, not on agent adoption.
• Time horizon. Weeks: CLARITY floor action (status unclear as of 2026-10-01; a possible US shutdown may squeeze the calendar [18]). Quarters: Q3 print ~2026-10-29 [19]. Years: agent float on USDC/Base becomes material only if agent-held balances reach the tens of billions (Inference).
• Visser fit: 7/10. COIN is the most complete rail-owner in public equities: USDC economics, Base, x402, custody, and perps. Points come off because COIN is an exchange whose P&L is still dominated by retail transaction cyclicality, and because Visser explicitly prefers rails over individual winners.
• Stance: Neutral, conviction 3/5. The franchise fits the thesis structurally, but the near-term P&L does not yet reflect it. The research hypothesis is that COIN re-rates on the thesis only once agent/stablecoin KPIs become a disclosed, growing segment. Until then it trades like BTC beta.

2. SOURCE-ACCURATE SUMMARY (by topic, with attribution)

Market data (fetched 2026-10-01) - COIN last $186.41, −0.92% d/d. Market cap $50.13B. EPS −$3.69. Down 43.60% over 1 year. ATH $444.65 (2025-07-18) [1]. The page said "Market closed" and did not stamp the session date. Treat this as the most recent close, probably 2026-09-30. Other aggregators in search returned stale values ($242.30, $204.23); I discarded them. - BTC $83,485, ATH $126,080, market cap $1.678T, dominance 56.6% [18]. - Street: one search summary cites an average 12-month target of ~$261.69 and says COIN was "down 34% YTD in 2026." The dates of individual actions are not confirmed: BofA $216 Neutral, Compass Point Sell, Barclays to Equal-Weight at $169 [20][21]. Treat these as directional, with timing unverified.

Q2-26 results (reported ~2026-07-30, after the close) [2][3][4] - Revenue $1.220B (consensus $1.290B). Transaction revenue $599.2M. S&S $555.1M (48% of net revenue, up from 29% less than two years earlier). - Net loss $359.5M ($1.36/diluted share). The driver was not itemized in my sources. (Inference) It likely includes marks on crypto investments plus restructuring (unverified, no live source). - Adjusted EBITDA $207.8M. Restructuring $52.4M (~700 layoffs). - Record 10.3% share of crypto trading volume (vs 9.1% in Q1). Spot volume fell >20% q/q as volatility hit multi-year lows. - Stablecoin revenue $292M. Average USDC in Coinbase products ~$20B, said to be ">30% of total USDC" at quarter-end. Stablecoin transaction volume on Base grew 7x y/y. Market-wide stablecoin volume was >$37T YTD, 79% of it USDC and Coinbase partner stablecoins. - Prediction markets: +106% q/q, >$100M annualized revenue. - FY26 opex guide $4.34-4.60B.

Agentic stack [5][6][7][8][23] - x402 is an open HTTP-402 stablecoin payment standard (May 2025). It passed 100M machine-to-machine transactions in ~9 months. May 2026: 3.1M operations, ~$1.2M value. - 2026-06-11: "Coinbase for Agents" launched, connecting AI assistants to Coinbase accounts for trading and data. - 2026-07-23: Coinbase Business merchants can accept agent payments in USDC via x402 with no extra setup. Coinbase Business had 5,000 customers, $1B in combined payments and trading volume, and a 12-person team. Early use is bots buying cloud credits, domains, and API access. The Block's article did not disclose the monetization model.

Stablecoin economics [9][10][11] - The Circle revenue-share was renewed through 2029 on existing terms: 100% of reserve income on USDC held on Coinbase, 50% of the residual off-platform. - USDC supply was ~$74.6B (early September 2026), +$0.6B in one week.

Regulation [12][13][14][15][16][17] - CFTC no-action letter (2026-05-28/29): Coinbase is the first FCM for global perps for US clients, routed to Deribit. - CLARITY (H.R. 3633): passed the House in July 2025 and Senate Banking 15-9 in May 2026. The August floor vote was postponed to after the recess (Senate returned ~2026-09-14). Republicans are split on stablecoin yield, and banks oppose rewards. Coinbase reportedly cannot support the latest language. More than 200 firms, Coinbase included, urged a floor vote. Galaxy cut 2026 passage odds to 60%. I found no live confirmation of a September floor vote or its outcome.

Other - CFO Alesia Haas spoke at Citi's 2026 Global TMT conference on 2026-09-10 [22]. I did not retrieve the content. - A Base native token has been "under consideration, no timeline." The date of that statement is unclear and may be 2025 [External check needed] [23].

3. SYSTEMS MAP / VALUE CHAIN

           AI AGENTS / LLM PLATFORMS (OpenAI, Anthropic, agent frameworks)
                │  (Coinbase for Agents, AgentKit/CDP wallets [ext. check], x402 client libs)
                ▼
  ┌──────────── DEMAND / ACCESS LAYER ─────────────┐
  │ Coinbase retail app · Coinbase Business (x402  │  ← fees: spread/commission (cyclical)
  │ merchant checkout) · Prime/custody (institutions,│  ← custody fees [ext. check]
  │ ETF custody [ext. check])                       │
  └────────────────────────────────────────────────┘
                │
  ┌──────────── SETTLEMENT RAIL ───────────────────┐
  │ Base L2 (Ethereum rollup) — sequencer revenue   │  ← sequencer margin (in S&S/other) [ext. check]
  │ x402 open standard (no direct toll observed)    │  ← ~$0 direct fee today [7]
  └────────────────────────────────────────────────┘
                │
  ┌──────────── MONEY LAYER ───────────────────────┐
  │ USDC (Circle issuer; reserves in T-bills/repo)  │  ← 100% on-platform reserve income +
  │ ~$74.6B supply [11]; ~$20B on Coinbase [3]      │    50% residual off-platform [10]
  └────────────────────────────────────────────────┘
                │
  ┌──────────── RISK-TRANSFER LAYER ───────────────┐
  │ Deribit (options/perps), CFM US perps [12][13]  │  ← derivatives fees, 24/7
  │ Prediction markets (>$100M ARR) [2]             │
  └────────────────────────────────────────────────┘
       Upstream dependencies: Ethereum L1 (DA/security), Circle (issuance, reserve mgmt),
       US Treasury bill rates (the stablecoin P&L is a rate trade), banking partners, CFTC/SEC.

Bottlenecks and leverage points (Inference unless cited)

Where constraints create opportunity (Inference). The binding constraint for agent commerce is not throughput, since Base already handles millions of operations. It is compliant custody plus KYC-linked agent identity plus fiat on/off-ramps. Coinbase is one of very few entities that holds all three under US regulation. That is the defensible part of the thesis. The protocol (x402) is deliberately open and is not the moat.

4. SECOND- AND THIRD-ORDER EFFECTS

1. Agents transact with stablecoins → agents must hold working-capital balances (float) → aggregate agent-held USDC grows faster than payment value → COIN's reserve-income line grows with a lag, independent of exchange volume. Investment relevance: the one channel through which the Visser thesis reaches COIN's P&L at scale. KPI to watch: on-platform USDC (~$20B now [3]). (Inference)
2. CLARITY restricts stablecoin rewards → exchanges can no longer pay holders USDC rewards → (a) Coinbase's rewards expense falls, so net take rises, but (b) retail/agent incentive to park USDC on Coinbase falls, so on-platform balances drift toward the 50% residual tier → net effect on the $292M line is ambiguous. Relevance: binary legislative risk sitting on ~24% of revenue. Requires modeling Coinbase's rewards payout, which I did not obtain [External check needed].
3. Fed easing lowers T-bill yields → reserve income per dollar of USDC falls → COIN needs USDC balance growth just to hold stablecoin revenue flat → at the same time, lower rates historically help crypto prices and transaction revenue. Relevance: COIN carries an internal rate hedge. Stablecoin and transaction revenue respond to rates in opposite directions, which dampens rate sensitivity but blurs the agent signal. (Inference)
4. US perps via Deribit + 24/7 agent trading → machine-driven derivatives volume concentrates on regulated venues with API-first access → Coinbase competes with CME (24/7 crypto futures [External check needed]), Kalshi [12], and offshore venues → third order: if agents arbitrage across venues, fee compression accelerates. Relevance: volume upside, but fee-per-trade pressure. Barclays reportedly cited "fee compression and structural shifts" [20]. (Inference)
5. The x402 open standard spreads (merchants, API providers) → rival wallets and chains adopt x402 (it is chain-agnostic [5]) → Coinbase seeds a market that competitors can settle on Solana or other L2s with non-USDC stablecoins → third order: COIN's share of agent flows decays even as the category grows. Relevance: the classic standards trap. Track the share of x402 settlement on Base vs other chains and in USDC vs other stablecoins. (Inference)

5. SCENARIO FRAMEWORK (12-24 months; probabilities are subjective analyst estimates)

6. COMPANY / ASSET WATCHLIST (research candidates, not recommendations)

COIN — Coinbase Global (primary) - Thesis: The only scaled, US-regulated equity that combines USDC economics, an L2 (Base), the agent-payment standard (x402), custody, and regulated perps. The agent thesis enters the P&L through float, not fees. - Key metrics: stablecoin revenue and average on-platform USDC; S&S share of revenue; trading market share (10.3% [2]); Base stablecoin volume; any disclosed x402 or agent-wallet KPIs; adjusted EBITDA vs the opex guide. - Catalysts: Q3 print ~2026-10-29 (forecast) [19]; CLARITY floor action (TBD); Circle Q3 print (date unverified); any Base token decision (no timeline) [23]. - Risks: crypto beta, rate compression, stablecoin-yield legislation, fee compression, ongoing GAAP losses.

CRCL — Circle Internet Group [External check needed for listing details] - Thesis: The other half of the USDC economics, with more pure exposure to the issuer and to "stablecoin as agent money." It shares the 50% residual with COIN [10]. - Metrics: USDC supply, distribution cost as a % of reserve income (Coinbase's take), share of USDC off Coinbase. Catalyst: Q3 print. Risk: heavy dependence on Coinbase distribution, and rates.

HOOD / CME / Kalshi (private) — competitive read-across on perps, prediction markets, and 24/7 trading [12]. Research only.

7. DILIGENCE QUESTIONS & RESEARCH AGENDA

8. RISK ANALYSIS

• Thesis risk. Agent payments may never become material in dollar terms. May 2026 x402 value was ~$1.2M [7]. If agent flows stay micropayments, the float stays small, and COIN remains an exchange story. The open standard also lets rivals capture the settlement leg.
• Timing risk. The legislative calendar is shrinking. Galaxy puts 2026 passage at 60% [16], and a possible US shutdown around 2026-10-01 is flagged in market news [18]. The Q3 print lands into a weak-volume, low-volatility tape [2].
• Execution risk. 700 layoffs in Q2 [2] while building agent products with small teams (Coinbase Business had 12 people [5]). Opex discipline vs investment needs. The Base token decision could create regulatory or dilution complications [External check needed].
• External risk. Fed easing compresses reserve income. Bank lobbying against stablecoin rewards [17]. Fee compression and analyst skepticism (Compass Point Sell, Barclays Underweight→EW, timing unverified) [20]. BTC drawdown: −34% from ATH [18].
• Concentration risk. The USDC economics depend on a single counterparty contract with Circle, renewed through 2029 [9]. The terms are stable, but the renewal concentrates the exposure.

9. FIT WITH VISSER "GHOST RAILS" THESIS — 7 / 10

SCOPE DISCLOSURE

• CHECKED: live web search and fetch on 2026-10-01: TradingView quote, CoinGecko BTC, Q2-26 coverage (Merca2, Odaily, Pulse2), The Block (x402), CLARITY Act coverage (PrimeXBT, American Banker, The Defiant, CryptoBriefing), CFTC perps (CoinDesk, Coinpaprika), Circle revenue-share (KuCoin, Decrypt), analyst-rating aggregators.
• NOT CHECKED: Coinbase 10-Q/shareholder letter primary documents. Exact quote timestamp. Balance sheet, cash, and share count. Any confirmed CLARITY outcome after mid-September. Base sequencer financials. The 22V index document itself (it did not surface in search). Dates on individual analyst actions. Circle's listed-equity details.
• WOULD CHANGE THE CONCLUSION: (1) Q3 disclosure of on-platform USDC well above $25B, or of agent-wallet balances, would move the Visser fit to 8-9 and the stance toward constructive. (2) A CLARITY outcome that pushes USDC off-platform would move the stance to avoid. (3) Evidence that x402 or Coinbase Payments takes a meaningful fee would add a second capture channel.

SOURCES

1. TradingView — COIN quote: https://www.tradingview.com/symbols/NASDAQ-COIN/
2. Merca2 — Q2-26 results (2026-07-31): https://www.merca2.es/2026/07/31/resultados-coinbase-q2-decepcionan-2427895/
3. Odaily — Q2 revenue $1.22B, net loss $359M: https://www.odaily.news/en/newsflash/505321
4. Pulse2 — Q2 record share / S&S mix: https://pulse2.com/coinbase-everything-exchange-drives-3rd-consecutive-quarter-of-record-crypto-trading-volume/amp/
5. The Block — agentic payments "most high-conviction bet" (2026-07-23): https://theblock.co/post/409539/coinbase-ai-payments-most-high-conviction-bet-businesses-accepting-agent-payments
6. Stablecoin Insider — Coinbase USDC payments from AI agents via x402: https://stablecoininsider.org/coinbase-usdc-payments-ai-agents-x402/
7. Pasquale Pillitteri — AiFi / x402 metrics: https://pasqualepillitteri.it/en/news/11787/aifi-coinbase-programmable-money-ai-agents
8. Let's Data Science — Coinbase enables AI agents to trade and pay: https://letsdatascience.com/news/coinbase-enables-ai-agents-to-trade-and-pay-e12bdc0d
9. KuCoin — Circle renews USDC revenue-sharing through 2029: https://www.kucoin.com/blog/circle-renews-usdc-revenue-sharing-agreement-with-coinbase-through-2029
10. Decrypt — Coinbase takes 50% of residual USDC reserve revenue: https://decrypt.co/312757/coinbase-circles-residual-usdc-reserve-revenue-filing
11. KuCoin — USDC circulation +$600M in 7 days: https://www.kucoin.com/news/flash/usdc-circulation-increases-by-600m-in-7-days
12. CoinDesk — CFTC opens crypto perp door (2026-05-28): https://www.coindesk.com/policy/2026/05/28/u-s-cftc-opens-crypto-perp-door-with-approval-of-first-regulated-firm
13. Coinpaprika — Coinbase wins CFTC approval for global perps: https://coinpaprika.com/news/coinbase-cftc-approval-global-crypto-perps/
14. PrimeXBT — Senate delays CLARITY vote to September: https://primexbt.com/news/senate-delays-clarity-act-vote-to-september-as-stablecoin-yield-fight-persists/
15. American Banker — Senate punts CLARITY Act to September: https://www.americanbanker.com/news/as-divisions-linger-senate-punts-clarity-act-to-september
16. The Defiant — 200 firms urge vote; Galaxy cuts odds to 60%: https://thedefiant.io/news/regulation/200-crypto-firms-urge-senate-vote-clarity-act-galaxy-cuts-passage-odds-60-percent
17. CryptoBriefing — banks oppose stablecoin rewards: https://cryptobriefing.com/clarity-act-faces-senate-vote-as-banks-oppose-stablecoin-rewards/
18. CoinGecko — Bitcoin price (fetched 2026-10-01): https://www.coingecko.com/en/coins/bitcoin
19. Wall Street Horizon — Coinbase earnings calendar: https://www.wallstreethorizon.com/coinbase-earnings-calendar
20. StreetInsider — COIN rating history: https://www.streetinsider.com/rating_history.php?q=COIN
21. TipRanks/The Fly — BofA PT $216: https://www.tipranks.com/news/the-fly/coinbase-price-target-lowered-to-216-from-217-at-bofa
22. 01net — Coinbase at Citi 2026 Global TMT Conference: https://www.01net.it/coinbase-to-participate-in-citis-2026-global-tmt-conference/
23. Let's Data Science — Coinbase pushes Base as AI-agent hub: https://letsdatascience.com/news/coinbase-pushes-base-as-ai-agent-hub-bc32e1ef

Research memo for internal and LP discussion. Not investment advice. No position recommendation is made.