I have been paying close attention to fintech for a while now. What happened in 2025 was not a regular market cycle. It was the industry finally deciding what it wants to be. The hype era is behind us. What replaced it is more interesting: a handful of companies pulling far ahead, a technology shift that changes who does the transacting, and a long-running debate about what money looks like getting resolved in real time.
Here is where things stand as of early 2026.
01The funding picture changed shape
After years of decline, fintech investment came back hard in 2025.
| Metric | 2025 | Change vs 2024 |
|---|---|---|
| Total global fintech funding | $52.7B | +35% |
| Deal count | 3,631 | -19% |
| Average deal size | ~$20M | +55% |
| Fintech share of global VC | 11% | 2nd only to AI |
Investors wrote fewer cheques and made them bigger. That tells you something. This is not a broad recovery where everyone gets funded. Capital is concentrating at the top. The middle is getting squeezed out.
B2B fintech dominated. In Q2 2025, B2B companies took 60% of the top payments rounds and 50% of the top banking rounds. Consumer plays, undifferentiated neobanks, and standalone BNPL products struggled to raise.
The IPO window also opened for the first time since 2021. Three flagships listed in 2025:
- Circle (June 5) — shares jumped roughly 168% on day one
- Chime (June 11) — raised $864M at an $11.6B valuation, first profitable quarter before listing
- Klarna (September) — raised $1.37B at a $17.4B valuation, swung from a $244M loss to $21M profit
The bar was the same for all three: show profitability or at least a credible path to it. The market rewarded the companies that made the shift.
02Digital currency is no longer theoretical
The stablecoin market grew 49% in 2025, climbing from $205 billion to $306 billion in total market cap. Stripe reported that stablecoin payment volume doubled to roughly $400 billion, with about 60% of that coming from B2B transactions. These are not speculative numbers. This is operational volume moving through live payment rails.
"Stablecoin payment volume effectively doubled to $400 billion in 2025, with around 60% of that volume coming from B2B payments, not retail speculation." — Stripe 2025 Annual Letter
The policy shift that enabled this was the GENIUS Act, signed in July 2025. It was the first proper federal framework for stablecoins in the U.S. Key requirements:
- 1:1 backing in cash or short-dated Treasuries
- Monthly reserve disclosures
- Audited financials for issuers above $50B outstanding
- Super-priority claims for stablecoin holders in insolvency
- Dual federal and state regulatory track
Following the law, Circle, Ripple, and Paxos all received provisional OCC banking charters.
At the same time, the U.S. formally walked away from a central bank digital currency. An executive order signed in January 2025 prohibited federal agencies from developing or promoting a CBDC and explicitly backed private stablecoins as the American alternative. The framing was about privacy. The actual outcome is that the dollar's global footprint is now being extended through Tether and USDC rather than through the Federal Reserve.
Other countries are taking different paths:
| Country | Approach | Status as of 2026 |
|---|---|---|
| USA | Private stablecoins via GENIUS Act | Live, $306B market cap |
| China | e-CNY CBDC | Restructured Jan 2026 to interest-bearing model |
| EU | Digital euro | Preparation phase, targeting 2029 |
| India | e-Rupee | Pilot, ₹10.16B in circulation |
China's e-CNY has processed roughly 16.7 trillion yuan in cumulative transactions but represents less than 0.2% of domestic payments. Alipay and WeChat Pay still dominate. Beijing restructured the e-CNY effective January 2026 to make it interest-bearing — a clear acknowledgment that the original model was not getting traction.
The honest read: stablecoins won the digital currency race before governments could finish drafting their rulebooks.
03Who is actually winning
Stripe is operating in a different tier entirely. It processed $1.4 trillion in payment volume in 2024, up 38% year over year. In 2025 it acquired Bridge for $1.1 billion, launched stablecoin financial accounts available in 101 countries, and introduced Open Issuance, which lets any business create and manage its own stablecoin with reserves managed by BlackRock and Fidelity. Stripe is not a payments processor anymore. It is building the infrastructure layer for how money moves globally, including programmable money.
In neobanking, the leaders are not where most people expected.
Nubank is the operational winner by a wide margin:
- 120 million customers
- ~$13B in annualized revenue
- $557M net profit in Q1 2025 alone
- ~$72B market cap
Revolut hit $75 billion in a late-2025 secondary share sale, on the back of $1.4 billion in pre-tax profits and 52.5 million customers across 38 countries. It is preparing to launch in Mexico, Colombia, Argentina, India, and Australia.
Chime has 22 million U.S. customers but still generates roughly 80% of its revenue from interchange. The IPO made the gap clear: Chime at $11.6 billion, Nubank at $72 billion. Pure interchange models cannot sustain the same multiples as diversified financial platforms.
Worth noting: only about 11 to 18% of the world's roughly 400 neobanks were profitable in 2025. The ones doing it at scale — Nubank, Revolut, WeBank, Starling — all share the same pattern. They moved beyond interchange into lending, subscriptions, FX, and wealth.
In B2B, Ramp is the breakout story. It closed 2025 valued at $32 billion, with 50,000 customers and over $100 billion in annualized purchase volume. It raised over $1 billion across three rounds in five months. The gap between Ramp and every other spend management company is widening.
04AI is not a feature anymore
This is the part that matters most right now, and it is where 2025 made a real shift.
Ramp launched its first set of AI agents for financial controllers in July 2025. These are not assistants. They act.
- Policy Agent — reviews every expense, auto-approves low-risk spend, escalates only the 10 to 15% that need human review. In a single month it prevented 511,000 out-of-policy transactions, saving customers $290 million.
- Treasury Agent — moved $5.5 billion from idle cash into 4% yield instruments automatically.
- Fraud Agent — caught a $49,000 AI-generated fake invoice before it was paid.
- AP Agents — handles invoice coding, fraud detection, approvals, and payments. Ramp claims 7x fewer clicks per invoice compared to Bill.com.
Zip went further. In June 2025 it launched 50 purpose-built AI agents for procurement. Not general chatbots. Specialized agents that each handle one job:
- Trade Policy Analyzer — monitors global tariff changes and recalculates supplier costs in real time
- Intake Validation Agent — at one customer, saved 1,400 hours annually by eliminating duplicate vendor reviews
- Price Negotiation Agent — surfaces overpayments and negotiates better deals across software, IT, services, and advertising
Zip's goal: within five years, 90% of its reviews handled by agents as it scales toward a billion reviews annually. Its current customers include OpenAI, Coinbase, Anthropic, HP, Lyft, and Snowflake.
The pattern is consistent across companies. Finance teams spend most of their time on high-volume, low-judgment tasks: expense review, invoice processing, policy enforcement, compliance audits. Agents handle the volume. Humans handle the edge cases. The ratio shifts over time.
The next step, already being built, is agents that transact autonomously. Stripe and OpenAI co-released an Agentic Commerce Protocol that powers ChatGPT's checkout. JPMorgan announced a partnership to let AI agents transact against merchant catalogs at bank-grade infrastructure. Visa launched Intelligent Commerce for the same purpose.
The unresolved question is not whether agents will buy things. It is who is liable when they buy the wrong thing.
05What this adds up to
Fintech in 2025 was not a recovery. It was a consolidation into a new shape. Looking at it from early 2026, the picture is clear.
A small group of companies built significant advantages in infrastructure, data, and AI. The gap between them and everyone else is widening. Mid-market neobanks, standalone BNPL products, and consumer wallets without differentiated economics are running out of time.
The two technologies that defined 2025 are stablecoins and AI agents, and they are converging. Stablecoins give agents programmable, globally transferable money. Agents give that money something to do autonomously. When those two things are fully integrated, software does not just manage money. It earns it, moves it, and spends it without waiting for a human to log in.
That is the shift. Everything else is noise.