Private credit and AI: what is the risk for investors?

Artificial intelligence has become one of the most closely watched risks in private credit: a large share of the loans made by major US managers over the past decade went to software companies, whose business models AI may disrupt. For a private investor in Spain, the lesson is not about technology but about basics: it matters a great deal what backs a loan and how capital is recovered if things go wrong.

In this article we explain what is happening, why it concerns supervisors and analysts, how lending against cash flows differs from lending against real estate, and which questions are worth asking before investing in private debt.

What is happening between private credit and artificial intelligence

For years, software was the favourite borrower of US private debt managers: recurring subscription revenue, high margins and low capital needs. According to a Bank for International Settlements bulletin (BIS Bulletin No 128, 14 July 2026), BDCs (business development companies, US private credit vehicles that report their portfolios loan by loan) have lent around USD 115 billion to software firms. That is about a fifth of all their lending and more than 80% of their technology portfolios.

The problem is that generative AI can replace software products or make them cheaper to build, increasing competition and casting doubt on the stability of that revenue. The BIS notes that, so far, there are hardly any arrears on these loans, but warns that spreads have narrowed and the cushion to absorb losses is thinner.

Third-quarter 2026 results show a more cautious sector: according to Bloomberg (12 November 2026), the largest listed BDCs made fewer loans than a year earlier, with rates rising and investors still wary of their software exposure.

Why software is a different risk: cash flow lending

Most of these loans are cash flow loans: they are granted on the basis of the company’s earnings or recurring revenue, not a physical asset. The security is usually the business itself (its shares, brands and contracts). If the business loses customers, its ability to pay and the value of what backs the debt fall at the same time.

Real estate secured lending works differently. The loan is backed by a property registered in Spain’s Land Registry, with a valuation and an LTV (loan-to-value, the ratio between the loan and the value of the asset). The property can also lose value, and enforcing the security takes time and costs money, but its value does not depend on an app remaining competitive. We explain this in more detail in our guide to asset-based finance.

Illustrative, hypothetical example: how much the security covers in an adverse scenario

Take two loans of EUR 1,000,000 each. The first is to a software company with recurring revenue of EUR 500,000 a year, valued at 5 times that revenue (EUR 2,500,000). The second is secured by a mortgage on a property valued at EUR 1,700,000 (an LTV of just under 59%), after deducting 10% for enforcement and sale costs. We compare what the security would be worth in three scenarios:

Illustrative example prepared by Ondara Group, November 2026
Scenario Software: company value Loan coverage Property: value net of costs Loan coverage
Base EUR 2,500,000 250% EUR 1,530,000 153%
Moderate (revenue −30%, multiple 3.5x / property −20%) EUR 1,225,000 122.5% EUR 1,224,000 122.4%
Severe (revenue −50%, multiple 2.5x / property −35%) EUR 625,000 62.5% EUR 994,500 99.5%

The example is hypothetical and simplified (it ignores default interest, enforcement timelines and company restructuring costs). It illustrates one idea: in a software loan, a fall in the business usually comes with a fall in the valuation multiple, and the two losses compound. In a mortgage loan, the security can also fall short in a very adverse scenario, but the starting point (a prudent LTV) makes the difference.

Data centre servers: artificial intelligence as a risk for private credit

Trends: what is happening worldwide and in Spain

What it means for a private investor in Spain

Very different products sit under the “private debt” label. A global private credit vehicle may be heavily exposed to US software companies; a bridge loan secured on a building in Barcelona carries different risks. Before investing, it is worth distinguishing:

  • What backs the loan: a specific, valued asset or the value of a company.
  • Sector concentration: how much a single sector weighs in the portfolio. See concentration risk in private debt.
  • How the portfolio is valued: how often and by whom (see private debt valuation).
  • Real liquidity: whether there is a fixed term or periodic redemptions that can be limited.

Risks to consider

  • Credit risk: the borrower may not pay, including in real estate secured deals.
  • Collateral value risk: security mitigates but does not eliminate risk; a property may be worth less than its valuation and enforcement takes time and costs money.
  • Illiquidity and term risk: capital is committed until maturity and the exit may be delayed.
  • Interest rate risk: with higher rates, refinancing at maturity is harder for the borrower.
  • Risk of loss: all investments carry risk, including partial or total loss of capital.

How Ondara Group works

Since 2018, Ondara Group has structured private debt transactions in Barcelona with real estate assets as collateral. We analyse each deal individually (the asset, the borrower and the exit plan), aim to diversify across transactions, and investors deal directly with the management team. We target double-digit returns, always as a target and never guaranteed. You can meet the team on our who we are page. The minimum investment is EUR 25,000.

Frequently asked questions

Why does artificial intelligence affect private credit?

Because a large part of US private credit has been lent to software companies, whose revenue may fall if AI replaces their products or lowers barriers to competition. If revenue falls, so does the ability to service the debt.

Are there already many defaults caused by AI?

According to the BIS (July 2026), software loans show hardly any arrears so far. The concern is forward-looking: a thinner cushion due to tight spreads and exposure concentrated in a few borrowers and a few managers.

Does this affect real estate private debt in Spain?

Indirectly. Spain’s alternative lending market relies mainly on property, not software companies. But a general tightening of credit or lower fundraising by global managers can influence conditions across the market.

Is a mortgage-backed loan immune to these risks?

No. Security reduces the expected loss if the borrower does not pay, but it does not eliminate it: the property can lose value and enforcement has timelines and costs. That is why the LTV and the analysis of the asset are key.

What should I ask before investing in private debt?

What backs each loan, its LTV, how the portfolio is valued, how much each sector or borrower weighs, and the term and real liquidity of the investment.

In summary

  • US BDCs have lent around USD 115 billion to software firms, a fifth of their lending (BIS, July 2026).
  • AI casts doubt on that revenue; defaults are low for now, but the cushion is thinner and European supervisors are watching the sector.
  • Lending against cash flows is not the same as lending against property: in the illustrative example, the severe scenario leaves the software loan more exposed.
  • In Spain, alternative lending relies mainly on property collateral, with an average LTV of 55% according to the Observatory.
  • All investments carry risk, including loss of capital: check what backs each loan.

If you would like to know how we invest, register as an investor and we will explain how it works. The minimum investment is EUR 25,000.

This article is for information purposes only and does not constitute an investment recommendation or financial advice. All investments carry risks, including the possible loss of the capital invested, and past or target returns do not guarantee future returns. Analyse each opportunity with a professional before investing.

Otros artículos