← back to ranking

COIN · Coinbase Global, Inc.

Financial - Data & Stock Exchanges · mkt cap $45.8B · calls: Q1 FY2026 vs Q4 FY2025
67.0 conviction · conf-adj 64

conf 3/10 partial

enthusiasm:27.0 · trend:8 · quantifies:5 · impact:0 · under_radar:5 · credibility:5 · business_impact:8 · disruption:0 · commitment:6 · confirmation:3

Enthusiasm latest 9 / prev 7 (rising)

Coinbase’s AI thesis intensified from Base/stablecoin wallets for AI agents in Q4 FY2025 to a broader Q1 FY2026 claim that Coinbase is becoming both AI-native internally and the leading stack for agentic commerce. The credible parts are quantified: agentic transaction share on Base, USDC settlement share, pull-request productivity, test coverage, and restructuring cost removal. The main caveat is that direct AI revenue was not quantified; the topline case is still mostly usage-share and ecosystem positioning rather than disclosed dollars.

GROUNDED NEXT-FY IMPACT vs CONSENSUS

Grounded on actual base — revenue $7.2B · net income $1.3B · net margin 17.6% · diluted EPS 4.45

These are next-fiscal-year annual uplift estimates, not next-quarter numbers.

Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: 31.34% · vs analysts: unclear · priced in: medium · confidence: 3/10

ClaimFigureArithmeticNext-FY Rev %Next-FY EPS %
Restructuring removed ~$500M of total costs (mkt headwinds + AI-native ops)
cost		$500M run-rate cost reduction$500M * (1-0.21 tax) = $395M after-tax; $395M / $1,260.327M NI = 31.34% EPS uplift (≈+$1.40 on $4.45 EPS). Bundled with non-AI market-headwind cuts, so AI-attributable share is a fraction (≈15.7% at 50% attribution). Topline ~0 (pure opex action).031.34
Pull requests per engineer up ~78-80% YoY
productivity · soft		+78-80% YoYEngineering-velocity proxy and mechanism behind the $500M cut, not an independent $ figure; no headcount cut or $ saving disclosed → null.
Integration test coverage up 3x in 6 months
productivity · soft		3xQuality/throughput metric, no revenue or cost dollars attached → null.
Anthropic Mythos finds vulns 99%+ of engineers would miss
productivity · soft		99%+Capability of a third-party model; no quantified security-cost or revenue impact to COIN → null.
>90% of onchain agentic transaction volume on Base
engagement · soft		>90% shareMarket-share of a nascent, UNSIZED agentic-volume base; no $ volume or Base fee economics disclosed → null. Strategic, not yet financial.
Agents use USDC 99%; 99% of x402 settled in USDC
engagement · soft		99%COIN shares USDC reserve-interest economics, but agent-payment USDC volume is not sized in $, so interest-revenue uplift is unquantifiable → null.
90% of agentic stablecoin volume settled on Base
engagement · soft		90%Duplicate share metric; unsized → null.
Soon billions of agents transacting on these rails
other · soft		billions of agentsUnanchored forward-looking TAM language; no figure or timeframe to size → null.

Assumptions: Current revenue $7.181325B, net income $1.260327B, net margin 17.55%, tax rate 21%, diluted shares ~287.2M. The only dollar-anchored claim is the $500M run-rate cost reduction, sized as pure bottom-line ($395M after-tax), topline ~0, treated as a full forward-year run-rate. CRITICAL: the $500M is explicitly bundled 'market headwinds AND AI-native operations', so it is NOT purely AI — true AI-attributable EPS uplift is materially lower (≈15.7% at 50% attribution). All Base/USDC/x402 figures are share-of-nascent-market metrics with no underlying $ volume disclosed, hence unsizeable; no agentic revenue dollars flow.

Top line: No quantifiable AI revenue uplift. Every topline AI claim (>90% of agentic volume on Base, 99% USDC settlement, 99% x402 in USDC) is a market-SHARE metric of a still-tiny, undisclosed agentic-payment base — strategically real (Base L2 + USDC rails positioned as the default agent payment layer) but not dollar-sized, so est_rev_uplift_pct is null. 'Billions of agents' is TAM rhetoric. None is supplier-side: COIN isn't selling AI compute/chips, it's positioning its own crypto rails as AI-agent infrastructure (adopter), so supplier uplift is 0.

Bottom line: The entire sizeable impact is cost-side: the $500M run-rate cut = $395M after-tax = 31.34% of $1.260B net income (≈+$1.40 on $4.45 EPS) IF fully AI-driven. It is not — management bundles market-headwind layoffs with the AI-native operating model, so the genuinely AI-attributable EPS uplift is materially lower (≈15.7% at 50% attribution). AI engineering productivity (PRs/engineer +78-80%, test coverage 3x) is the qualitative enabler but carries no independent dollar figure.

Consensus already models a brutal 2026: revenue $6.084B (−15.3% vs $7.181B) and EPS $0.90 vs $4.45 (−79.8%) — a crypto-cycle/trading-revenue collapse, not an AI story. The $500M cost action is almost certainly already embedded in post-call estimates (it cushions, not reverses, the decline), so the one hard AI-adjacent number is largely priced in. The topline agentic-rails optionality (Base/USDC dominance of agent payments) is NOT in consensus but is unquantifiable today, so it can't be scored as a beat. Net: the sized piece is priced in (medium), the upside is real but unmeasurable.

MODEL CONSENSUS (impact)

partial

Both agree: sole hard claim is the $500M AI-native cost cut (31.34% EPS, bundled so true AI share lower); all agentic-rails metrics unsizeable.

Conflicts reconciled
FieldOpus 4.8GPT-5.5
Rev uplift %–0
EPS uplift %31.340.31341072594652025
Priced inmediummedium
vs analystsunclearunclear
Confidence43
Top lineNo quantifiable AI revenue uplift. Every topline AI claim (>90% of agentic volume on Base, 99% USDC settlement, 99% x402 in USDC) is a market-SHARE metric of a still-tiny, undisclosed agentic-payment base — strategically real (COIN's Base L2 + USDC rails are positioned as the default agent payment layer) but not dollar-sized, so est_rev_uplift_pct is null. The 'billions of agents' line is pure TAM rhetoric. None is supplier-side: COIN isn't selling AI compute/chips, it's positioning its own crypto rails as AI-agent infrastructure (adopter).No hard AI revenue uplift is calculable: the 90%-99% Base/USDC/x402 figures are mix or share metrics on undisclosed volumes, so the quantified next-FY revenue uplift is $0 / $7.181325B = 0.00%.
Bottom lineThe entire sizeable impact is cost-side: the $500M run-rate cut = $395M after-tax = 31.34% of $1.260B net income (≈ +$1.40 on $4.45 EPS) IF the full amount were AI-driven. It is not — management explicitly bundles market-headwind layoffs with AI-native operating model, so the genuinely AI-attributable EPS uplift is materially lower (≈15.7% at 50% attribution). AI engineering productivity (PRs/engineer +78–80%, test coverage 3x) is the qualitative enabler of that cut but carries no independent dollar figure.The only hard bottom-line figure is the $500M cost reduction: after 21% tax, $395M / $1.260327B current net income = 31.34% EPS uplift. PR, test coverage, and security claims are operationally relevant but unanchored to dollars.
ReasoningConsensus already models a brutal 2026: revenue $6.084B (−15.3% vs $7.181B) and net income $152.5M (−87.9% vs $1.260B, EPS $0.90 vs $4.45) — a crypto-cycle/trading-revenue collapse, not an AI story. The $500M cost action is almost certainly already embedded in those post-call estimates (it cushions, not reverses, the decline), so the one hard AI-adjacent number is largely priced in. The topline agentic-rails optionality (Base/USDC dominance of agent payments) is NOT in consensus but is unquantifiable today, so it can't be scored as a beat. Net: the sized piece is priced in (medium), the upside is real but unmeasurable.Consensus 2026 revenue of $6.084138B is -15.28% vs the $7.181325B current base, and consensus 2026 EPS of $0.89916 is -79.79% vs current EPS of $4.45. The hard AI math adds no calculable revenue and $395M after-tax savings, equal to about $1.38/share on 287.209M diluted shares, but the claim is likely known to analysts and not purely AI-attributable. Consensus 2027 revenue of $7.596035B is only +5.77% vs current revenue and EPS of $5.18528 is +16.52% vs current EPS, so the disclosed AI claims do not clearly prove upside beyond consensus without monetization data.

Rows highlighted where the two models disagreed.

QUANTIFICATIONS
Future AI-agent transaction population: billions of agents (forward-looking, topline)
“There will soon be billions of agents transacting and they need rails that can keep up.”
Onchain agentic transaction volume on Base: over 90% (Q1 FY2026, topline)
“We're also leading on the next frontier with over 90% of onchain agentic transaction volume happening on Base.”
AI-agent crypto payment mix using USDC: 99% (Q1 FY2026, topline)
“And when agents pay with crypto onchain, they use USDC 99% of the time and over 90% of those transactions are happening on the Base chain in Q1.”
AI-agent crypto payment mix on Base: over 90% (Q1 FY2026, topline)
“And when agents pay with crypto onchain, they use USDC 99% of the time and over 90% of those transactions are happening on the Base chain in Q1.”
Pull requests per engineer: up almost 80% (year-over-year, bottomline)
“Our product velocity is already increasing rapidly. The number of pull requests per engineer is up almost 80% year-over-year.”
Integration test coverage across core services: up 3x (last 6 months, bottomline)
“Integration test coverage across core services is up 3x in the last 6 months.”
x402 transactions settled in USDC: 99% (Q1 FY2026, topline)
“One is that 99% of the x402 transactions right now are settled in USDC, that's from Q1.”
Agentic stablecoin transaction volumes settled on Base: 90% (Q1 FY2026, topline)
“90% of the agentic stablecoin transaction volumes were settled on Base in Q1.”
Pull requests by engineers: up 78% (year-over-year, bottomline)
“So as we shared, pull requests are up by engineers by 78% year-over-year.”
Cost reduction from restructuring actions involving market headwinds and AI-native operations: $500 million (compared to Q4 2025 run rate, bottomline)
“With regard to specific dollars, the actions removed about $500 million of total costs as compared to the Q4 2025 run rate.”
Security vulnerabilities detectable by Anthropic Mythos model: 99% plus of human engineers (recently, bottomline)
“And we saw a glimpse of this recently actually with the Mythos model that Anthropic put out, where it's actually able to find security vulnerabilities that 99% plus of human engineers would not have been able to find.”
PAST (realized)
CURRENT (now)
FORWARD (guidance)
TRACK RECORD — PROMISE vs DELIVERY

50/100 track record   too-early  6 calls reviewed

AI is essentially absent from Coinbase's earlier calls and emerges only in the most recent ones as an 'AI-native' efficiency push; the sole quantified, AI-attributed target — the 2026 ~$500M expense reduction — was just made and has not reached its judging window, so their AI delivery track record is currently untestable.

Transition to an 'AI-native company': 2026 adjusted expenses of $4.3B-$4.6B, ~$500M below the Q4 2025 annualized exit rate — promised Q1 FY2026
too-early Announced in the most recent call alongside a headcount cut ($50-60M Q2 restructuring charge); the 2026 timeframe has not elapsed, so the AI-attributed cost-out cannot yet be judged
AI coding productivity: pull requests per engineer up ~80% YoY and integration test coverage up 3x over 6 months, cited as proof of the AI-native shift — promised Q1 FY2026
too-early Presented as already-achieved metrics in the same final call, not a prior forward target verifiable against a later call
PRICED-IN (REFINED)
MEDIUM

Est. revisions falling  ·  Fwd P/E 23.2  ·  EV/Sales 7.5x

AI claim maps to Bank Servicing, Consumer, Net, Bank Servicing, Institutional, Subscription and Circulation, Blockchain Infrastructure Service

Estimate momentum looks falling: recent price-target averages are down sharply from the last-year average, ratings are broadly flat to slightly less bullish versus early 2026, and forward revenue/EPS estimates are volatile rather than steadily compounding. Valuation is still rich, with 23.2x forward P/E, 57.5x TTM P/E, and about 7.5x EV/sales, so the market is paying up despite weaker revision signals. AI upside would most plausibly show up in consumer/institutional exchange servicing and blockchain infrastructure services rather than all revenue. Because revisions are not rising but valuation is stretched, the AI upside looks only partially priced in, not clearly low or high.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
1Q4 FY20241Q1 FY20251Q2 FY20251Q3 FY20256Q4 FY20259Q1 FY2026

AI enthusiasm across 6 calls — trend ↗ rising

AI was absent until late 2025, then became a concrete Base, USDC, x402, and agentic commerce growth thesis.

RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY

8/10 qualitative impact   material  medium-term · mixed evidence

Where AI matters: Base/USDC agent payments and AI-native cost structure

Coinbase has concrete AI-adjacent traction in agentic payments, with over 90% of onchain agentic volume on Base and 99% USDC settlement, plus real internal productivity claims tied to a $500M cost-removal program. The revenue opportunity is still unsized and may be a small base today, so this is material optionality rather than proven transformation.

Caveats: Agentic-payment volumes are share metrics on an undisclosed and likely nascent base; The $500M cost reduction is bundled with market-headwind restructuring, not purely AI-driven; Open standards such as x402 may benefit the ecosystem without ensuring Coinbase captures economics; AI-generated code and AI-native operations raise security, compliance, and quality-control risk

AI DISRUPTION / CANNIBALIZATION RISK  two-sided · 3/10

AI does not directly automate away Coinbase's core regulated exchange, custody, stablecoin, or blockchain infrastructure roles, and agent activity may increase crypto transaction demand. The risk is that AI agents abstract away Coinbase's consumer interface or route through cheaper open rails, pressuring retail fees and making Base/x402 economics more commoditized.

OPTIONS / MARKET STRUCTURE

option liquidity: good

ATM IV
—
TYPICAL BID-ASK
—
OPEN INTEREST
—

proxy inputs — dollar-ADV $1.9B · beta 3.381 · px $173.99

source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.

CONFIRMATION — INSIDERS · 13F · LANGUAGE
Mixed — insiders selling, institutions adding, management language 8/10 committed.
INSIDERS selling 35 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) adding as of 2026-03-31: 143 new / 229 closed positions; 644 increased / 384 reduced; institutional ownership -2.97pp; -115 net 13F holders
MGMT LANGUAGE 8/10 committed Concrete current usage metrics make the stance committed, though framed around agent-economy infrastructure rather than internal AI automation.
commit “over 90% of onchain agentic transaction volume happening on Base.”
commit “USDC and Base are now powering the majority of onchain stablecoin transactions for AI agents.”
hedge “agentic commerce is the next frontier.”
VERBATIM AI QUOTES
“And now crypto has a new catalyst, AI. There will soon be billions of agents transacting and they need rails that can keep up. Crypto is the only option that checks all 3 boxes, fast, cheap and global.”
— Brian Armstrong, Q1 FY2026
“We're also leading on the next frontier with over 90% of onchain agentic transaction volume happening on Base.”
— Brian Armstrong, Q1 FY2026
“And we're also building stablecoin infrastructure for agents. So USDC and Base are now powering the majority of onchain stablecoin transactions for AI agents.”
— Brian Armstrong, Q1 FY2026
“And when agents pay with crypto onchain, they use USDC 99% of the time and over 90% of those transactions are happening on the Base chain in Q1.”
— Brian Armstrong, Q1 FY2026
“We're seeing agents also use the x402 protocol for a wide variety of use cases, including trading, AI inference, media generation, storage and more.”
— Brian Armstrong, Q1 FY2026
“In short, Coinbase is at the center of the agent economy.”
— Brian Armstrong, Q1 FY2026
“As we mentioned in Tuesday's announcement, we are transitioning to be an AI-native company.”
— Alesia Haas, Q1 FY2026
“Our product velocity is already increasing rapidly. The number of pull requests per engineer is up almost 80% year-over-year.”
— Alesia Haas, Q1 FY2026
“Integration test coverage across core services is up 3x in the last 6 months.”
— Alesia Haas, Q1 FY2026
“The ability to scale our team members and their ability to iterate and improve our products at these speeds is a game changer for our execution throughput and efficiency.”
— Alesia Haas, Q1 FY2026
“It allows agents to spend small or large amounts attached to any request, whether that's to e-commerce checkout or any other agent in the world.”
— Brian Armstrong, Q1 FY2026
“So your question asked about how this helps Coinbase. Well, there's a couple of ways. One is that 99% of the x402 transactions right now are settled in USDC, that's from Q1.”
— Brian Armstrong, Q1 FY2026
“90% of the agentic stablecoin transaction volumes were settled on Base in Q1.”
— Brian Armstrong, Q1 FY2026
“These are all products that we either co-created or helped create, and it's been a really great journey to see that all those pieces come together to become the leading stack for agentic commerce.”
— Brian Armstrong, Q1 FY2026
“We definitely saw market headwinds, and we have definitely also seen a transition to AI-native operations.”
— Alesia Haas, Q1 FY2026
“So as we shared, pull requests are up by engineers by 78% year-over-year.”
— Alesia Haas, Q1 FY2026
“I think that we're going to only see more and more of our work being done by AI in all of our functions over time.”
— Alesia Haas, Q1 FY2026
“Base set a new transaction all-time high with AI agents adopting stablecoin wallets.”
— Brian Armstrong, Q4 FY2025
“Base is quickly establishing itself as the onchain home for AI.”
— Brian Armstrong, Q4 FY2025
“We're even seeing these AI agents adopt stablecoins for payment, and I believe that stablecoins will be the default payment method for AI agents.”
— Brian Armstrong, Q4 FY2025
“we put out some really useful tools for developers to just get any AI agent a crypto wallet and begin to make stablecoin payments and begin to complete agentic commerce essentially.”
— Brian Armstrong, Q4 FY2025
“L2s should focus on value-added features, including AI, privacy, et cetera.”
— Brian Armstrong, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, @Architect9000): It was fairly alarming in your note earlier this week, Brian, to hear that nontechnical developers are pushing code -- AI code into production. Is that really true? And how is Coinbase going to marry AI's ability to move fast while preserving high quality and brand trust.
A: Brian Armstrong: "we encourage product managers, designers, other nontechnical employees to use AI agents to draft code, that's getting easier to do. But human engineers still review all code before it goes into production." Alesia Haas: "our investment in quality, our investment in integration testing is exceeding the pace of our growth in new pull requests."
Q (Q1 FY2026, Rayna Kumar at Oppenheimer): As we get closer to the commercialization of agentic payments at scale, can you talk about the particular opportunity you see for x402, specifically, how should we think about incremental USDC on platform growth from x402 adoption? And over time, how meaningful could transaction fees on Base and from the x402 facilitator really become?
A: Brian Armstrong: "One is that 99% of the x402 transactions right now are settled in USDC, that's from Q1." Brian Armstrong: "90% of the agentic stablecoin transaction volumes were settled on Base in Q1." Brian Armstrong: "it's been a really nice thing that's grown from out of Coinbase to become an open standard that has secondary effects on all of our various products."
Q (Q1 FY2026, Alex Markgraff at KBCM): Can you walk us through the RIF? I think many folks are curious to understand how much is a function of the current environment versus AI leverage? And what do you anticipate in terms of cost savings, both in quarter and run rate?
A: Alesia Haas: "We definitely saw market headwinds, and we have definitely also seen a transition to AI-native operations." Alesia Haas: "So as we shared, pull requests are up by engineers by 78% year-over-year." Alesia Haas: "With regard to specific dollars, the actions removed about $500 million of total costs as compared to the Q4 2025 run rate."
Q (Q4 FY2025, @InternetToken): with Base TVL and sequencer revenue growing strongly in late 2025, what percentage of overall subscription and services revenue do you expect Layer 2 activity from Base and partners to contribute in 2026? And are there plans to further incentivize builders there?
A: Brian Armstrong: "we put out some really useful tools for developers to just get any AI agent a crypto wallet and begin to make stablecoin payments and begin to complete agentic commerce essentially. And that started to get quite a good amount of traction."
Q (Q4 FY2025, Crypto Pete Christiansen at Citi): The debate further argues that L2s should focus on value-added features, including AI, privacy, et cetera. What's Coinbase's view on the Base L2 value prop going forward in this respect? And how might potential DeFi regulations shape Base's future?
A: Brian Armstrong: "We've seen that Base has rapidly become the #1 L2 on Ethereum." Brian Armstrong: "It's really great for payments. It's great for trading. It's great for DeFi."
Q (Q4 FY2025, Gus Gala at Monash Crespi Hart): How is this different from potential revenue S-curve? Contrast that with USDC on base for more consumer-centric volumes via x402.
A: Alesia Haas: "our goal is to drive a payments vertical, as Brian shared in our goals for 2026, where we create the best place for businesses to come transact in USDC on Base to enable their payments businesses."