How to Start a Neobank in 2026: Complete Guide [With Costs & Licensing]

The neobank space had a reckoning between 2020 and 2024. Xinja returned AU$100 million in customer deposits. Volt Bank surrendered its banking licence. Simple closed. Bank North collapsed mid-Series B. And when Synapse filed for bankruptcy in 2024, it left between $85 million and $96 million in customer funds unaccounted for — an event the CFPB is still working through.

None of that means the opportunity closed. The global neobanking market was valued at $210–$230 billion in 2025, growing at 36–62% annually. SME accounts now drive 65–68% of that market — a segment most legacy banks still serve with 1990s-era digital infrastructure.

What changed is the cost of getting things wrong. Regulators are sharper, BaaS dependencies carry real bankruptcy risk, and “we’ll figure out monetization later” no longer survives a Series A. This guide covers what it actually takes: the licencing math, the tech stack decisions, where development money goes, and what specific companies failed to get right.

Key Takeaways

• The global neobanking market reached $210–$230 billion in 2025, projected to grow to $4.99–$9.38 trillion by 2033–2035
• Three licencing paths: BaaS partnership (3–6 months), EMI licence (6–12 months), full banking charter (15+ months, $5M+ capital)
• MVP development costs range from $50,000 for a basic build to $350,000+ for a full-scale platform — before licencing
• US neobanks face $250,000–$350,000 in direct year-one licensing costs across 53 state money transmitter licences
• SME banking accounts for 65–68% of neobank market revenue — the highest-value customer segment by ARPU

Quick answer: Starting a neobank in 2026 requires five decisions: a target niche and revenue model, a licencing path (BaaS, EMI, or charter), a core banking platform, a KYC/AML stack, and an MVP feature set. Total costs range from $300,000 to $2 million+ depending on jurisdiction and build approach.

What Is a Neobank in 2026 — and Why the Window Is Still Open

A neobank is a digital-only financial institution with no physical branches. The practical difference from a fintech app is regulatory: neobanks offer core banking products — accounts, deposits, card programmes — through their own licence or a licenced banking partner.

The market numbers are real. The 2026 neobanking market sits at $210–$230 billion. Long-term projections range from $4.99 trillion to $9.38 trillion by 2033–2035, driven by smartphone penetration in emerging markets and open banking API adoption. The growth leader isn’t retail consumers switching their primary accounts — that’s been slower than early predictions. The actual growth driver is B2B: business accounts and SME solutions hold a 65–68% share of neobank market revenue in 2026, driven by demand for expense tracking, invoicing integration, and real-time business payment tools.

Where the window stays open: most legacy banks still operate SME banking on infrastructure that hasn’t changed since the early 2000s. A neobank doesn’t need to compete with Revolut’s 45 million users. It needs to be meaningfully better than the local commercial bank for one specific use case — gig economy payments, cross-border SME transfers, teen banking in an underserved market — and build from there.

Who Is Building Neobanks in 2026? Five Operator Profiles

The people searching “how to start a neobank” are not a uniform group. Five distinct operator profiles drive most new neobank projects, each with different constraints and failure modes:

The profile you fit determines which licencing path makes sense, which tech architecture is appropriate, and where your real capital risk lies.

Step 1 — Define Your Niche and Build a Revenue Model That Survives

Why Pure Debit Interchange Isn’t a Business Model

The failure pattern here is well-documented. Simple and Cogni, both US neobanks, built large user bases on debit card programmes. Neither became profitable. Debit card interchange fees cannot cover customer acquisition costs at any reasonable scale — the maths doesn’t close without a second revenue stream.

Xinja is the cleaner case study. They launched a high-yield savings account with 1.5–1.8% interest rates before building a single lending product. Every customer cost them money from day one. When pandemic conditions dried up their funding pipeline, there was no path to viability. AU$100 million returned to depositors, licence surrendered.

The revenue model must be solved before architecture decisions are made. The structures that work combine:

• Interchange fees — baseline revenue, not sufficient alone
• FX markup on multi-currency and international transfers
• Premium subscription tiers — travel insurance, higher limits, cashback
• Embedded lending — BNPL, overdraft, micro-loans; this is where the actual margin lives
• SME account fees — invoicing tools, expense management, payroll integration; highest ARPU of any neobank segment
• Stablecoin and crypto services — where authorised under MiCA (EU) or relevant regional frameworks

Choosing a Niche With Real Demand in 2026

The verticals with demonstrated traction:

• SME business banking — underserved by incumbents, 65–68% of neobank market value
• Gig and freelance economy — contractors, marketplace sellers, independent service providers who need faster payouts and simpler expense management
• Remittance corridors — MENA to South Asia, Latin America to US; cross-border payments with transparent FX rates
• Stablecoin-first global accounts — reaching the estimated 1.4 billion unbanked adults in markets with weak local banking infrastructure
• Emerging markets — India (UPI-integrated bank partnerships), Nigeria (post-2026 AML reform framework), Brazil (Nubank’s model of eliminating hidden fees for underserved populations)

The common thread: pick a segment where the incumbent’s product is measurably bad, and be better at the specific thing that segment needs most.

Step 2 — Choose Your Licencing Path

Three paths. Each has real costs, real timelines, and real constraints.

Path 1 — BaaS / Sponsor Bank Partnership (3–6 Months)

The BaaS model lets you launch in 3–6 months with minimal capital requirements. A licenced banking partner holds regulatory responsibility; you build the product and customer relationships on top. The tradeoff: revenue share runs 20–40% of transaction fees and interchange, which compresses margins significantly at scale.

The structural risk became concrete in 2024. Synapse Financial Technologies filed for bankruptcy, leaving $85–96 million in customer funds unaccounted for across dependent platforms including Yotta, Juno, and Copper. The CFPB issued a judgment against Synapse in 2025. Every BaaS-dependent neobank now needs direct contractual access to core account records and a documented contingency if the middleware provider becomes insolvent — not a nice-to-have, a survival requirement.

Path 2 — Electronic Money Institution Licence (6–12 Months)

An EMI licence gives direct control over customer ledgers and e-money issuance without the credit liabilities of a full banking licence. No fractional reserve requirements, no deposit guarantee scheme obligations.

Capital requirement: €350,000 initial paid-up capital for unrestricted EMIs.

Costs by jurisdiction:

Lithuania remains the most used EU entry point for early-stage neobanks: direct SEPA access via the Bank of Lithuania’s CENTROLINK, a predictable licencing timeline, and lower local staffing costs compared to Western European jurisdictions. The minimum of four local employees is a real requirement, not optional.

One planning note: the EU reached a provisional political agreement on PSD3 in late 2025. Full transposition is expected by late 2027, with a 24-month grandfathering period for existing EMIs. If you’re applying for an EU licence now, compliance planning should account for PSD3, not just PSD2.

Path 3 — Full Banking Charter (15+ Months)

A full banking licence unlocks deposit-taking, lending, and the complete product suite. It also requires $5–$15 million or more in initial capital, a multi-year regulatory approval process, and ongoing capital adequacy compliance. For early-stage startups, the capital cost alone is prohibitive.

This path is realistic for corporate spinoffs with parent institution support, or for established neobanks expanding their regulatory footprint. Nubank filed for an OCC charter in late 2024 after achieving profitability across Latin America. N26 and Monzo both hold full banking licences in their home markets.

Step 3 — Select Your Core Banking Platform

The core banking platform runs your ledger — every account balance, transaction record, and daily reconciliation. Choosing this wrong costs months of rebuild time and carries real regulatory exposure if audit processes fail.

On Solaris: The BaFin situation has materially changed the risk profile. Multiple neobanks that built on Solaris have since migrated. If your build is EU-focused, evaluate Equals, Mambu, or one of the alternatives below.

Alternatives for EU/UK: Codego supports crypto-funded card programmes and completes BaaS migrations in 6–10 weeks. Gemba (FCA-regulated) assumes regulatory liability and risk management on behalf of the neobank — meaning you can offer GBP accounts and card programmes without carrying the compliance burden directly.

Step 4 — Build Your Neobank Tech Stack

Mobile and Web Frontend

For most neobank MVPs, Flutter or React Native is the right call — single codebase, faster iteration, lower maintenance costs. Native dual apps are justified when deep device integration is required: hardware security modules, specific biometric implementations, or platform-specific banking security frameworks. For the majority of products, that level of native integration isn’t needed at MVP stage.

Backend Architecture

Golang is the right choice for high-throughput ledger environments processing hundreds of thousands of transactions per hour. For an MVP, it’s over-engineered. The pattern that works: Node.js or Python for initial build, migrate transaction-critical ledger services to Golang at the point when throughput actually demands it.

API Integration Costs

KYC/KYB Provider Comparison

The stacked fees problem: A base KYC check listed at $1.85 often hides downstream costs: document classification ($0.40) + face matching ($0.30) + active liveness screening ($0.20) + sanctions screening ($0.15). Real cost per onboarded user: $2.90+. At 100,000 users, the difference between the listed price and the actual cost is $105,000. Model this in your unit economics before selecting a provider.

Payment Rails by Target Market

• United States: Integrate FedNow (Federal Reserve’s 24/7 instant settlement rail, operational since 2023) alongside RTP. FedNow is particularly relevant for payroll-focused or gig economy products — instant settlement at any hour is a real product differentiator. Use ACH for high-volume, non-urgent batch processing.
• Europe / UK: SEPA Instant for euro transfers across 36 countries; UK Faster Payments for GBP. PSD2 open banking API connectivity required for account aggregation features.
• Cross-border: SWIFT for high-value international transfers. ISO 20022 migration is underway — it provides richer transaction data and faster automated reconciliation, and it’s worth building to that standard from the start.

Card Issuance

Marqeta handles enterprise-grade card programmes with JIT Funding 2.0 — real-time transaction controls at the authorization level. Setup runs $50,000–$150,000 with $5,000–$10,000 monthly minimums. Galileo offers lower entry costs for US-focused builds. Both require substantial API integration work.

AI Features Already Running in Production

Not theoretical roadmaps. These are live:

• bunq “Finn”: In-app spending AI for budget tracking and savings nudges. Bunq’s own data shows users trust bank-built AI nearly twice as much as generic chatbots — relevant when designing your AI interface strategy.
• Revolut APP fraud detection: ML-based system that identifies transaction patterns suggesting a user is being coached by a fraudster, then declines the payment and guides them through an in-app intervention. Revolut reports a 30% reduction in fraud losses from card scams since launch.
• Credit decisioning: Transaction history analysis for risk-adjusted lending decisions — running in production at multiple mid-sized neobanks.

For MVP, the AI layer isn’t required on day one. But designing your data architecture to support it from the beginning is worth the small upfront investment in schema planning.

For full-cycle neobank engineering — BaaS integration, KYC/AML stack, card issuance, and payment rails — see our neobank app development services. If your product includes a standalone digital wallet component, see our digital wallet app development page.

Step 5 — Navigate Compliance by Region

Compliance is the operational area where first-time neobank founders most consistently underestimate both cost and complexity.

United Kingdom: The FCA’s Electronic Money Institution regime is active and well-understood. Neobanks also need to comply with the Payment Services Regulations (PSRs), PSD2 open banking API obligations, and GDPR data handling requirements. The FCA runs an Innovation Sandbox (DISC) for early-stage applicants who want to test products before full authorisation.

European Union: PSD2 requires that licensed banks open customer account data to authorised third parties via APIs. The transition to PSD3 — with a provisional political agreement reached in late 2025 — will standardise API performance requirements, tighten fraud liability frameworks, and unify PI/EMI licencing criteria across member states. Transposition expected by late 2027, with a 24-month grandfathering period for existing licences. Build for PSD3 now.

Under the EU’s MiCA framework, any neobank issuing asset-referenced tokens or e-money tokens must obtain authorisation as an EMI or credit institution. This matters for any product with stablecoin or tokenized rewards components.

United States: Neobanks must either obtain Money Transmitter Licences (MTLs) across all 53 US jurisdictions or operate via FBO (For Benefit Of) partner bank arrangements. Direct year-one licensing costs run $250,000–$350,000, covering state application fees ($100,000–$130,000), surety bond premiums ($120,000–$160,000), and registered agent filings ($15,000–$30,000). Annual renewal costs run $225,000–$280,000.

Beyond licencing, FinCEN and the Bank Secrecy Act require: a written, board-approved AML programme; a designated compliance officer; and automated transaction monitoring systems capable of generating Suspicious Activity Reports (SARs) and Currency Transaction Reports (CTRs).

United Arab Emirates: The DIFC and ADGM financial free zones offer common-law regulatory frameworks with structured pathways. DIFC Category 3D (Money Services Business): $200,000 minimum base capital. ADGM Category 3C: $250,000. Statutory setup costs in either zone run approximately $150,000 in year one, excluding salaries and office rent. Both zones offer Innovation Testing Licences (ITLs) for early validation.

FATF Travel Rule: Applies globally. Financial institutions and VASPs must collect and share originator and beneficiary information for transactions exceeding specified thresholds. Any neobank handling cross-border transfers needs this compliance layer built into the transaction processing architecture.

Step 6 — Assemble Your Launch Team

The technical team structure that works for a neobank launch:

• CTO / fintech architect — responsible for core ledger design, API layer, and security architecture
• Compliance officer — required for EMI applications; needs prior experience with the target regulatory body
• Fraud and risk analyst — transaction monitoring rules, KYC escalation handling
• Product manager with fintech background — ideally someone who has shipped a regulated financial product before
• UX designer — mobile-first, with experience designing trust-sensitive interfaces (banking UX is not the same as consumer app UX)
• BaaS integration engineer — dedicated to the middleware layer; this is where delays usually happen
• DevSecOps engineer — security-first infrastructure from day one; financial-grade applications have different security surface areas than typical SaaS

On the build vs. outsource question: Eastern Europe development teams run $35–$85/hour against $95–$200/hour for North American teams. For a 1,000-hour project, that difference is $60,000–$115,000. The relevant question isn’t cost alone — it’s whether the outsourced team has prior experience with regulated financial product architecture. For an overview of how OmiSoft structures fintech software development engagements, including compliance integration and BaaS setup, see the service page.

Step 7 — Build, Launch, and Scale Your MVP

Phase 1: Strategic Foundation (Days 1–30)

Define the target niche and map unit economics in detail — not at the level of “we’ll earn interchange fees” but at the level of: what is our projected CAC, what is the 12-month LTV for each product tier, and at what user volume does the model break even? Model the regulatory path based on your geography and product scope. Choose BaaS or EMI based on both capital availability and timeline requirements.

Phase 2: Vendor Selection and Architecture Design (Days 31–60)

Procure the core ledger. Establish contractual direct access to core account records — not optional if you’re on a BaaS model, non-negotiable after Synapse. Configure primary and backup KYC/KYB providers: running a single KYC provider without a fallback is an operational risk. Implement automated transaction monitoring from day one.

Phase 3: Integration and Testing (Days 61–90)

Deploy payment rail integrations for your target market. Configure the card issuance programme. Run stress tests on ledger performance and transaction reconciliation loops. Verify AI fraud detection systems before launch — not after. Test sub-ledger reconciliation against real bank statement data.

Phase 4: Soft Launch and Scale (Days 91+)

Release to a limited beta group before public launch. The metrics that matter:

• Activation rate: Industry benchmark is 32%. Below this, your onboarding flow has friction that will compound at scale.
• Day 7 retention: Target 40%+. If users aren’t returning within a week, the core product isn’t solving the problem clearly enough.
• Support ticket volume: High ticket volume in the first two weeks usually means the core user journey has an unresolved UX problem, not a support staffing problem.

As transaction volumes grow, migrate high-throughput card and payment flows to optimized rails to improve interchange margin capture.

How Much Does It Cost to Start a Neobank in 2026?

Development Costs by Project Scope

Using white-label platforms or configurable SaaS core banking engines reduces infrastructure costs by up to 70% compared to custom builds — but adds ongoing SaaS fees and limits long-term product differentiation.

Licencing and Regulatory Costs

Ongoing Annual Operating Costs

Once live, the annual operating budget typically covers:

• BaaS platform fees: 20–40% of transaction and interchange revenues (on BaaS model)
• Cloud infrastructure and ledger licensing: Enterprise ledger systems and secure hosting can exceed $50,000/month at scale
• Compliance monitoring and audits: $500–$2,000/month
• 24/7 customer support operations: $5,000–$20,000/month
• Customer acquisition and marketing (initial): $50,000–$200,000

Developer Rates by Region

Eastern Europe offers the strongest balance of cost, fintech domain expertise, and timezone overlap with EU markets — which is relevant for compliance reviews and live integration work during EU banking hours.

Why Neobanks Fail — Lessons from Real Closures

Five documented failures worth understanding before you build.

Xinja (Australia, 2020)

Xinja offered a savings account (Stash) at 1.5–1.8% interest before launching any lending products. The structural result: every user cost them money from day one. When pandemic conditions ended the funding environment, Xinja surrendered its banking licence and returned AU$100 million in customer deposits.

The principle: A savings product with no lending revenue creates a negative carry that only venture capital can absorb. Once funding stops, the model has no path forward.

Volt Bank (Australia, 2022)

Volt invested heavily in building a proprietary custom ledger before launching public products — high burn rate, long development cycles, and mortgage origination timelines that didn’t satisfy investors looking for early traction. The result: licence surrender and AU$100 million returned to depositors in June 2022.

The principle: Custom core infrastructure before product-market fit is a capital trap. White-label platforms or configurable cores exist precisely to avoid this. Build for differentiation; buy for infrastructure.

Bank North (UK, 2022)

Bank North operated under a restricted banking licence — a structure that requires continuous equity capital raises to meet regulatory capital thresholds before taking full retail deposits. They were unable to close a £30 million Series B within their regulatory timeline and initiated a solvent wind-down in October 2022.

The principle: Restricted banking licences create a specific capital timing risk that doesn’t exist in BaaS or EMI structures. If you go this route, the fundraising pipeline needs to be de-risked before the regulatory clock starts.

Simple and Cogni (United States)

Two US neobanks with solid product design that couldn’t reach profitability. The business model — debit card interchange only — couldn’t offset customer acquisition costs at any viable scale. No lending products, no subscription tiers, no high-margin services.

The principle: Debit interchange is a cost of doing business, not a business model. Any neobank built on interchange alone will eventually require either a pivot or an exit.

Synapse Financial Technologies (United States, 2024)

Not a neobank itself, but a BaaS middleware provider that multiple neobanks depended on. Synapse’s bankruptcy in 2024 resulted in $85–96 million in customer funds going missing across dependent platforms. The CFPB issued a formal judgment in 2025. The platforms built on Synapse — Yotta, Juno, Copper — had no direct access to underlying account records and no independent path to customer fund recovery.

The principle: BaaS dependency without direct account record access is an unacceptable operational risk. Any contract with a BaaS provider should include: direct API access to core ledger records, fund reconciliation rights, and contractual fallback procedures in the event of provider insolvency.

By Dmytro Romaniuk

CEO & Founder

15 mins read | Jun 11, 2026

• 
• 
•