Global banks earned $1.3 trillion in net income in 2025 — the most profitable industry on earth, for the third consecutive year. And yet: fintechs have claimed 17% of industry revenues. Revolut has 69 million customers. Nubank 131 million. WeBank 420 million. Gen AI reached 45% of US working-age adults in two years — digital banking took 15 years for the same penetration. McKinsey's central scenario: if banks don't adapt, $170 billion erosion of profit pools — enough to push average returns below cost of capital. The most profitable industry in the world is being attacked from every direction simultaneously.
Not investment advice. Data sourced from McKinsey Global Banking Annual Review May 2026, Deloitte 2026 Banking Outlook, S&P Global Banking Predictions 2026, Backbase Banking Predictions 2026, Fortune Business Insights Neobanking Market 2025, BCG AI Reckoning May 2025. All figures current as of June 2026.
The paradox of global banking in 2026 is difficult to hold in the mind simultaneously. On one hand: record profits. Global bank net income rose 7% in 2025 to $1.3 trillion — more than any industry anywhere on earth, for the third consecutive year. Balances (deposits, loans, and assets under management) grew 6.5%. Investment banks had an exceptional run from geopolitical volatility. US banks' net interest margins actually improved by 9 basis points as rates remained higher than expected. By any conventional measure, global banking is thriving.
On the other hand: structural erosion. Fintechs have claimed 17% of industry revenues — up from essentially zero a decade ago. Transaction banking and distribution now account for 57% of banking profits — areas with far more competition and faster disruption than traditional balance sheet lending. Customer trust, banking's most durable competitive advantage, is shifting: fintechs and neobanks now score higher on trust metrics than incumbents in multiple markets surveyed by McKinsey. Gen AI is reaching consumers at a speed — 45% of US working-age adults in two years — that makes the previous digital disruption wave look slow. Stablecoins, private credit, embedded finance, and CBDC pilots (Letter 100) are each attacking a different part of the banking value chain simultaneously.
McKinsey's central warning is the most important number in banking strategy: if banks don't adapt, the central scenario predicts a $170 billion (9%) erosion of global banking profit pools — enough to push average returns below cost of capital for the first time since the 2008 financial crisis. Early AI adopters, conversely, could see a 4 percentage point improvement in return on tangible equity. The gap between leaders and laggards is widening faster than at any point in the industry's modern history.
Until 2025, neobanks were a growth story — impressive customer numbers, uncertain profitability. The 2025 McKinsey review identified the crossing of a critical threshold: neobanks are no longer trading profitability for growth. Nubank's ROE is approximately 30%. Revolut and Wise clock in at approximately 35%. Robinhood is over 20%. These returns exceed most incumbents. The story has changed from "they have customers but not profits" to "they have more customers, better products, higher returns, and they're just getting started." 21 US banking licence applications were filed by fintechs in 2025 — compared to one in 2024 and five in 2023. The neobanks are not attacking banking. They are becoming banks.
The most structurally important competitive advantage in retail banking has always been customer inertia — the hassle of switching accounts, cards, and loans, which allows banks to earn more from existing customers than market competition would dictate. AI is systematically destroying this advantage. AI agents can automatically move deposits to the highest-yielding accounts. Optimise credit card debt by finding lower-rate loans. Shop for financial products purely on value. McKinsey identified that banking profitability in deposits and credit cards has long relied on customer inertia. If AI agents eliminate that inertia — and the technology to do so already exists — the repricing could remove $170 billion from industry profit pools.
Letter 93 covered this in detail — private credit grew from $250 billion in 2008 to $1.75 trillion in 2025 at 22% CAGR, filling the vacuum left when Basel III forced banks to hold more capital against loans. The consequence for banks is structural: the most profitable lending business — leveraged finance for private equity-backed companies — is now largely owned by Apollo, Ares, Blackstone, and their peers. Banks are left with the regulated, capital-intensive, lower-margin lending that the private credit funds won't touch. The McKinsey review noted: "In lending, private credit firms have built substantial businesses. These firms operate under different regulation, but their critical difference is leaner operations."
Stablecoins threaten the payment infrastructure of banking at the moment when payments generate $2.5 trillion in annual revenue from $2 quadrillion in value flows. Citi's September 2025 report "Stablecoins 2030: Web3 to Wall Street" projects stablecoin market cap reaching $3.7 trillion by 2030. If even a fraction of dollar-denominated payments migrate to stablecoin rails, the interchange fee revenue that funds much of retail banking disappears. Embedded finance — financial services built into non-financial platforms (ride-hailing apps, e-commerce, accounting software) — further disintermediates the bank from the customer at the transaction moment.
JPMorgan is not a bank that is being disrupted — it is a bank that is doing the disrupting. It generated $58 billion in net income in 2025, the most of any bank in the world. It spent $17 billion on technology in 2025 — more than most fintech companies are worth. It has 180 AI use cases in production. Its retail banking franchise (Chase) has 80 million customers; its investment bank dominates every league table. JPMorgan is building the platform that other banks wish they were building — a technology company with a banking licence and the most valuable financial brand in the world.
Jamie Dimon's 2025 shareholder letter was the most important document in banking strategy of the year. It said: AI is not a feature. It is a fundamental transformation. JPMorgan is deploying AI across every function — trading, underwriting, compliance, customer service, fraud detection — not as individual pilots but as a comprehensive operational transformation. The implication: JPMorgan is trying to become the technological incumbent that is so far ahead that the neobanks can challenge it on customer experience but not on capability.
HSBC's strategic transformation under CEO Georges Elhedery — announced in late 2024 and executed through 2025 — is the most significant restructuring of a global bank since the financial crisis. The bank reorganised into four divisions: UK, Hong Kong, Corporate and Institutional Banking, and International Wealth and Premier Banking. Two regional banks (Americas, Middle East) were folded in. The headcount has been reduced; the cost base is being simplified. The strategic logic: HSBC is focusing on its irreplaceable advantage — the cross-border banking network connecting the world's two largest economies (China and the US) at a moment of geopolitical tension when that connectivity is extremely valuable to corporates on both sides.
HSBC's transaction banking franchise — trade finance, cash management, foreign exchange — serves the global supply chains that are being reorganised by China+1 (Letter 96), reindustrialisation (Letter 97), and food security (Letter 95). Every major corporate that is building a new supply chain link needs banking infrastructure in the countries at both ends. HSBC sits at that intersection by design — and that positioning is more valuable in 2026 than at any point in the past decade.
China will generate approximately $5 trillion in banking net interest income in 2026 — the largest single country banking market on earth, larger than the entire US banking sector. The Industrial and Commercial Bank of China (ICBC), Bank of China, China Construction Bank, and Agricultural Bank of China collectively manage over $25 trillion in assets. The scale is incomprehensible by Western standards. ICBC alone has approximately 680 million retail customers — more than the population of every country in Europe combined.
WeBank — Tencent's digital bank and the world's largest neobank by customer count at 420 million — has already reached 11th position by market value among China's banks. The coexistence of the biggest traditional banking system and the biggest neobank ecosystem in the world, in the same country, is producing the most intense competitive dynamics in global banking. The Chinese banking market is where the future of banking is being written fastest — but it is also the most opaque to Western investors due to state ownership, regulatory complexity, and geopolitical constraints on capital flows.
European banks have spent the period since 2008 in a state of chronic underperformance relative to their US peers — lower returns, heavier regulation, more fragmented markets, and the structural challenge of negative interest rates that persisted until 2022. The rate normalisation of 2022–2024 restored profitability. UK banks' NIM improved 6 basis points in 2025. UniCredit's aggressive cross-border M&A push — bids for Commerzbank in Germany and Banco BPM in Italy — reflects the consolidation logic that European banking has needed for a decade. UniCredit CEO Andrea Orcel is attempting to build the pan-European bank that the single market was theoretically designed to enable but that regulatory fragmentation has prevented.
The European defence spending boom (post-Ukraine, NATO 2%+ targets moving to 2.5–3%) is creating a major new lending opportunity for European banks — infrastructure financing, defence contractor lending, sovereign bond underwriting. BNP Paribas, Deutsche Bank, and UniCredit are all expanding their defence-adjacent lending books. The geopolitical moment is, paradoxically, good for European bank revenue — high volatility, rising defence spending, and the need to finance reindustrialisation (Letter 97) all generate fee income and lending growth.
Founded London 2015. Now the world's most valuable private fintech at approximately $45 billion. UK banking licence approved 2024 after 3-year wait. 21 US banking licence applications filed in 2025 industry-wide signals the direction. ROE of ~35% — higher than most G7 banks. Product: current accounts, savings, crypto, stock trading, insurance, business accounts. The super-app of European banking.
Founded São Paulo 2013. Now surpassed Petrobras as Brazil's most valuable company (Letter 86). Cost to serve $0.80/month vs R$30–35 for traditional banks. Expanding to Mexico, Colombia. The proof that a neobank can achieve institutional scale with returns that match or exceed incumbents. The template that every other neobank ecosystem is trying to replicate.
The world's largest digital bank by customer count. Backed by Tencent. Operates entirely without branches. Lends to individuals and small businesses using AI credit scoring on WeChat social and transaction data — the model that makes conventional credit bureaux obsolete. China's proof that a digital bank can serve half a billion people profitably without a single branch or ATM.
The US neobank pair. SoFi obtained its banking charter in 2022 — the strategic move that allows it to fund loans with deposits rather than expensive wholesale funding. Robinhood's evolution from a trading app to a full-service financial platform (banking, credit cards, retirement accounts) at ROE above 20% demonstrates the platform strategy working at US scale.
Wise (formerly TransferWise) has built the most efficient cross-border payments infrastructure in Europe at ROE ~35% — outperforming banks at the most difficult banking function. Trade Republic has become Germany's largest online broker and is expanding into banking — offering 4%+ interest on deposits at a moment when German banks offer near-zero. The cost of complacency, priced by customers in real time.
The UK neobank pair that created the playbook. Starling Bank became profitable in 2021 — the first UK neobank to do so — and has been building its business banking franchise. Monzo, valued at $5.9 billion in its 2024 funding round, has 10 million customers and files for a US banking licence in 2026 — taking the template that worked in the UK into the world's largest financial services market.
The speed asymmetry is the most important single fact in banking strategy in 2026. Every previous wave of technology disruption gave incumbent banks a decade or more to adapt — time to observe, pilot, learn, and deploy. The mobile banking wave took 10–15 years to reach mass adoption; during that time, incumbents watched neobanks emerge, assessed the threat, and eventually built their own mobile offerings. The AI wave is moving 5–7 times faster. Gen AI reached 45% of US working-age adults in two years. Banks that are still in the "observe and assess" phase are already losing.
BCG's May 2025 report stated plainly: "For banks, the AI reckoning is here." Only one in four banks worldwide is actively using AI to gain a competitive advantage. Three in four remain stuck in fragmented pilots and proofs of concept. The gap between AI leaders and laggards in banking is now measurable: McKinsey estimates a 4 percentage point ROTE differential between early AI adopters and slow movers — on a global average ROTE of 11.8%, a 4 point difference is existential.
The specific AI applications that matter most in banking are clear: AI-driven underwriting that can process loan applications in minutes rather than days; AI compliance monitoring that can review every transaction for regulatory risk in real time; AI-powered customer service that resolves most queries without human intervention; AI trading and market making that react to market conditions in milliseconds. These are not futuristic capabilities — JPMorgan, Goldman Sachs, and the leading European banks are deploying all four at scale right now. The banks that master AI deployment in 2026–2027 will have cost structures that make them nearly impossible to compete with. The banks that don't will face a choice between irrelevance and acquisition.
Regulatory moats are more durable than the disruption narrative suggests. The reason 21 fintech companies applied for US banking licences in 2025 — compared to one in 2024 — is that banking regulation is genuinely hard. A banking licence requires capital, compliance infrastructure, stress testing, consumer protection, anti-money laundering systems, and ongoing regulatory examination. The fintechs that want to be banks are discovering this. Revolut waited three years for its UK banking licence. The regulatory burden that makes banking frustrating for incumbents is the same burden that makes it difficult for challengers to fully replicate. The incumbents that use the regulatory window to close the AI gap are not as vulnerable as the disruption narrative suggests.
Scale and trust in wholesale banking are extremely sticky competitive advantages. The neobank disruption is primarily a retail banking story — consumer current accounts, savings, payments. Wholesale banking — corporate lending, capital markets, trade finance, derivatives — is a different competitive landscape. The relationships between JPMorgan, Goldman Sachs, HSBC, and the world's largest corporations have been built over decades and involve capabilities (credit capacity, global network, regulatory relationship management) that neobanks do not yet possess. The commercial and investment banking businesses of the major global banks are more defensible than their retail businesses. The disruption is real but its geography is more specific than the headlines suggest.
The acquisition strategy may be the incumbents' most underutilised weapon. US banks have $200+ billion in excess capital. The valuation gap between large incumbents (trading at 1.5–2.0× tangible book) and neobanks (some trading at 3–5× revenue but without the profitability that justifies the premium) creates acquisition opportunities. Every neobank that needs a banking licence, every payments company that wants bank deposits as a funding source, every wealth management platform that needs balance sheet — these are acquisition targets for incumbents who can buy the digital capability and customer base they need rather than building it from scratch. The banks that move from "building slowly" to "acquiring fast" in 2026–2027 may compress the disruption timeline in their favour.
Long-horizon thinking on capital, technology, and the forces shaping the next decade of wealth creation. Written from first principles. Not consensus. Not noise.