For the past decade, founders and investors were told the same thing: avoid credit risk, stay off the balance sheet, and keep fintech “asset-light.” That advice worked when data was patchy, capital was scarce, and operations were mostly manual.

But the ground has shifted. In 2025, infrastructure is cheaper, data’s available in real time, and capital is flowing again. Clinging to an asset-light model now doesn’t de-risk your business, it caps your upside.

Here’s the contrarian take: the next generation of breakout fintechs will own (at least part of) the infrastructure. Not for bragging rights, but because the economics finally make sense. And ignoring them means leaving margin on the table.

What went wrong last time

The last wave of balance sheet fintechs ran into the same wall: their unit economics never became a burden. On paper, lending margins looked attractive. In practice, costs killed the model.

Costs stacked the wrong way:

• Cost of capital (CoF): unless you were a bank, senior funding was chunky, slow, and pricey. Advance rates were conservative, covenants unforgiving.
• Cost of operations: onboarding, KYB/KYC, invoice checks, reconciliations, collections — armies of people, brittle workflows, error-prone handoffs.
• Cost of risk: underwriting on stale data; little telemetry post-disbursement; fraud and first-payment-defaults hiding in the noise.
• Cost of acquisition: distribution wasn’t embedded. CAC amortized over short tenors killed contribution margins.

The math in one line:

net excess spread = yield - (CoF + losses + opex + CAC amortized)

For most “lenders,” that equation went negative as soon as rates moved above zero. Which is why the last cycle left everyone repeating the same mantra: stay asset-light.

What changed (and why the equation now clears)

The constraints that made balance sheet fintech a losing game last decade are gone. Technology, regulation, and capital markets have all moved, and the economics now look very different.

1) AI collapsed opex and improved losses

Tasks that once required armies of analysts are now automated with AI: KYB/KYC, document review, invoice validation, account categorization. Fraud detection and early-warning signals kick in before problems surface. Collections are agent-assisted, not fully manual.

The result: fewer human touches, faster cycles, and sharper credit segmentation. Losses move from blunt cutoffs to policy-driven micro-decisions.

2) e-invoicing created clean, real-time rails

Mandated formats and structured invoice data mean transactions are standardized. Counterparties are verifiable, payment states are clear, and fraud checks move upstream. Working capital can now be monitored continuously instead of waiting for quarterly statements.

3) Private debt scaled and standardized

Institutional investors have an appetite for this asset class, and securitization programs are now repeatable. Advance rates are more predictable, execution frictions slimmer. With institutional-grade data and servicing, you can fund like a bank without being one.

4) Distribution is embedded

Marketplaces, SaaS platforms, banks, and payment providers have become the new distribution channels. Customer acquisition is built into the product flow instead of showing up as a standalone P&L line item.

Net effect: net excess spread turns positive — and scalable.

How to build he new balance sheet stack (so it doesn’t blow up)

Owning assets doesn’t have to mean owning operational headaches. Today’s balance-sheet fintechs run on infrastructure designed to keep risk transparent, processes automated, and costs under control.

Here’s what the modern stack looks like:

Origination rails

• Direct APIs into accounting, banking, invoicing, and commerce platforms.
• Real-time eligibility checks (and no more “apply and pray”).

Risk & pricing

• A policy engine with hard guardrails on eligibility, limits, and tenors.
• Dynamic daily pricing tied to risk telemetry and cost of funds.
• Static-pool and cohort monitoring embedded from day one.

Funding

• Bankruptcy-remote SPVs with senior, mezz, and equity layers.
• Programmatic issuance with clear triggers, cash waterfalls, and lockbox structures.
• Advance rates flexing based on data quality and servicer performance.

Servicing

• Automated cash application and reconciliations.
• Invoice state machines (issued, approved, disputed, paid) to track every stage.
• Agent-assisted collections supported by reason codes, playbooks, and promises-to-pay.

Monitoring

• Node-based risk models that capture counterparty, sector, and network effects.
• Bank-account telemetry on balance volatility, inflow sufficiency, and concentration.
• Early-warning rules for payment slippage, broken debit cadences, and credit line stacking.

With this stack, humans move from “doing” to “deciding.” That’s where the margin comes from.

A quick unit-economics teardown (illustrative)

The old model collapsed because the spread went negative. In today’s environment, the math looks very different. Here’s a simplified walk-through for short-tenor working capital, with potential numbers as an illustration:

• Gross yield (Y): 12%
• Cost of funds (S): 5.0% (programmatic issuance)
• Expected losses (L): 2.0% (based on 12-month static pools)
• Opex (O): 1.5% per € financed (post-automation)
• CAC (C): 0.5% per € (embedded distribution)

Now the math: Y – (S + L + O + C) → net excess profitability ≈ 3%.

3% may look modest, but remember: these assets turn over quickly. Scale that over high-velocity turnover and you get real ROE above 30%, without relying on fantasy multiples.

The point isn’t the exact numbers. It’s that data + automation + programmatic capital move you solidly to the right side of zero.

“But isn’t asset-light safer?”

Until recently, this wasn't even a debate. Why would anyone choose the grinding complexity of credit risk and regulatory capital when you could build elegant SaaS?

Recurring revenue, seat expansion, tidy multiples: the choice was obvious.

Then AI arrived and broke the model.

Consider what's happening to traditional SaaS economics: when one person with Claude or ChatGPT can do what used to require a ten-person team, user-based pricing doesn't just wobble, it collapses. That lovely ARR growth suddenly looks fragile when your enterprise client realises they need 80% fewer seats.

Meanwhile, AI transforms the parts of financial services that used to be prohibitively expensive. All those manual KYC checks, document validation, payment reconciliations, and early collections are now automated. The operational complexity that killed margins becomes manageable.

Here's the crucial bit: ChatGPT can build you a brilliant cash flow model, but it cannot lend you fifty thousand euros.

Only institutions with balance sheets can do that. Regulatory permissions matter. Capital allocation remains stubbornly analogue, and that's precisely the point.

When you operate asset-light, you're taking credit risk without credit margins. You depend entirely on someone else's underwriting (and their generosity with economics). But when you hold the asset, you control the three things that matter: how you price risk, how you distribute product, and how you collect payment.

Software gets commoditized, but capital remains scarce. In an AI world, that's where the defensible business lives.

Where this already works

We can apply different wrappers with the same underlying truth: own the risk policy, instrument the data, and standardize the funding.

• B2C BNPL (like Klarna): Short duration, embedded distribution, and rich consumer data lead to predictable losses and real scale.
• Embedded merchant capital (Stripe/Block/Shopify/Adyen): In-product offers, repayment via daily take-rate, and platform telemetry provide low CAC and tight risk loops.
• Invoice financing (Iwoca + embedded in ERPs/AR tools/marketplaces): E-invoicing and lockbox/virtual IBANs offer clean collateral, short tenor, and securitization.

Why this matters for Europe (and for banks)

Europe’s e-invoicing wave and SMB digitization create rare alignment. Regulators want transparency, SMBs want liquidity, banks want capital-efficient growth, and investors want yield with real collateral.

A modern balance-sheet platform can slot into that gap, either lending directly to SMBs or powering banks with modular servicing, data, and risk.

What we’re building at Defacto

We’ve already financed €1B+ for 17,000 SMBs and run a €300m fund with institutional partners.

Our latest €16M round lets us push the stack further, and lets banks and platforms use it without rebuilding it themselves.

We’re building:

• Better data pipes‍
  Defacto will plug into more ERPs, invoicing tools, and banks. Customers get cleaner invoice statuses, and clear maps of “who owes whom.”

• Servicing that banks can trust
  ‍We handle KYB/KYC, invoice checks, cash posting, and collections as a product. So a bank doesn’t need a big team to scale.

• Risk rules that run themselves
  ‍Limits update automatically based on live data. Covenants are checked by the system. Alerts trigger actions (pause draws, adjust pricing) without waiting for a meeting.

Back to the balance sheet: Infrastructure meets opportunity

Owning a balance sheet used to be baggage. But today, it’s leverage—the good kind.

If you can see the cashflows, price the risk, automate the work, and fund programmatically, balance-sheet fintech beats asset-light on real margins.

This fundraising helps us provide the critical infrastructure that lets balance-sheet businesses, and the banks that lend to them, scale efficiently.