Private Credit Deal Origination Insights: UK Direct Lending in Focus

Editor’s Note: Welcome to Octus’ new feature series on the U.K.’s M&A pipeline and its financing. In the second part of our coverage, we examine how AI disruption risk is bifurcating underwriting in the U.K. mid-market, with AI-resilient credits attracting favorable leverage and pricing while AI-exposed borrowers face tighter terms – and how that dynamic could be tested by the 2028 maturity wall.

Artificial intelligence is changing how private credit investors assess risk, with players in the U.K. mid market increasingly looking beyond traditional sector classifications to determine which businesses can withstand technological disruption.

For years, lenders could make relatively broad sector calls, avoiding areas such as retail, chemicals and construction while favoring software and other companies with a high portion of annual recurring revenue. The rapid adoption of AI is making those distinctions difficult to sustain as the technology creates both opportunities and disruption risks across sectors in the mid-market.

“AI has made that approach much less useful because disruption risk cuts across sectors,” David Hirschmann, co-head of Permira Credit and head of its private credit business, told Octus.

Permira Credit is a pan-European manager investing across Western and Northern Europe, including the U.K., France, Germany, Benelux and Scandinavia. Around one-third of its portfolio companies are headquartered in the U.K.

The U.K. is an important testing ground for the shift. European direct lending activity in the first quarter of 2026 totaled 223 deals – a 4% increase year over year, according to Octus league tables. U.K. and Ireland accounted for 27.8% of European direct lending activity in the first quarter of 2026.
 

Hirschmann said lenders are increasingly assessing the characteristics that make an individual business defensible, rather than relying primarily on the sector in which it operates. “We’re seeing a move from sector-based underwriting to moat-based underwriting.”

The winners are not necessarily businesses that are least exposed to AI, but those that are most capable of incorporating AI into their products and operations, he said. “So increasingly, we view AI readiness as important as AI risk.”

David Miles, partner and co-leader of global leveraged finance and private capital at law firm Mayer Brown, said that there is a growing bifurcation in the market between businesses that are beneficiaries or resilient to AI and those where the technology could cause disruption.

“This is carefully considered in the commercial diligence phase of a transaction and those outcomes of course drive capital allocation across those categories of businesses,” Miles said.

For businesses where AI is seen as a challenge, to the extent capital remains available, this can result in lower leverage requirements and higher equity checks, coupled with greater information right requests, increased key performance indicator, or KPI, monitoring and reporting, particularly in areas where AI may first impact, he said.

“Each situation is different, though. It’s company and sector specific, and those businesses ‘winning’ with AI are able to attract capital on favorable terms relative to those where AI is seen as a genuine risk to their operation model.”

“As the adoption of AI accelerates, we expect valuation, leverage and refinancing outcomes to diverge more meaningfully between resilient and exposed borrowers,” Hirschmann added.
 

The shift is particularly significant for software, where recurring revenue have traditionally been viewed as a source of defensiveness.

That assumption is increasingly tested in the U.K. market. Sophos, the U.K.-based cybersecurity company backed by Thoma Bravo, has been the subject of heightened lender scrutiny ahead of a March 2027 maturity. Several private credit firms had passed on a proposed $2.5 billion refinancing, with the company subsequently discussing an amend-and-extend with existing lenders around its outstanding $2.1 billion loan. Octus reported in April 2026 that Sophos was weighing options to refinance its 2027 debt, including a debt package with existing lenders in the mid-SOFR+500 bps area or a private credit option with pricing at SOFR+575 bps-600 bps.

“Of course, the old assumption that recurring revenue equalled resilience has been challenged by AI,” Hirschmann said. “We pay much more attention to factors such as workflow embeddedness, data moats, error tolerance, regulation and management’s ability to adapt,” he said.

The counterpoint is visible in new-money financing. Octus reported in March 2026 that Nordic Capital secured a $175 million unitranche from Ares to finance its acquisition of U.K.-based trade surveillance technology provider TradingHub. The facility was sized at about 7x EBITDA. The deal illustrates that lenders remain willing to finance specialist technology businesses.

“We’re more constructive and open to businesses that sit deeper within customer workflows and those that combine software with domain expertise, proprietary data and regulated processes,” Hirschmann added. “In those cases, AI typically becomes an enhancement tool rather than a substitute.”

Miles said businesses where AI could replicate the product currently provided are an area of focus, including customer service platform deals, low complexity application software and content generation.

“Lenders are focused on the degree to which AI could replace the product, whether a business has defensive characteristics around its product and proprietary rights to its product, and to what degree pricing power could be eroded,” he said.

The result is a more granular approach to software credit. AI disruption fears were increasingly reshaping credit structures for software companies, with market participants expecting shorter maturities, higher pricing, greater amortization and tighter documentation, as reported.

The issue is not confined to software, however. “AI risk is widespread and has the potential to disrupt all industries and all businesses,” Hirschmann said. Every business should be subject to scrutiny around this issue, across every sector and every business, he said.

The ability to invest in technology is another consideration, particularly for smaller companies competing against larger businesses. “There is a correlation, in my mind, between AI readiness and capacity to invest,” Hirschmann said.

“A €5 million EBITDA business competing with much larger players clearly faces a risk of lagging behind if it cannot invest as much and therefore cannot incorporate AI into its products and operations.”

Permira Credit’s direct lending strategy generally targets businesses with EBITDA of roughly €20 million to €100 million.
 

AI considerations are increasingly feeding through into financing terms. “For businesses that the market views as AI-exposed, we see lenders focusing on lower leverage multiples, more conservative EBITDA adjustments, greater scrutiny around forecasts and margin assumptions, potentially higher pricing and wider spreads and stronger covenant protection,” Hirschmann said.

Octus data shows that average pricing on software-related direct lending deals fell by almost 50 bps year-over-year in the first half of 2026, to 524 bps from 571 bps, even as pricing increased for business services and industrials and materials. The data are sector-wide, while lenders are increasingly differentiating between individual software companies based on their AI exposure. So software is still attracting competitive financing, but lenders are becoming more selective.
 

By contrast, companies viewed as more resilient can continue to attract strong lender competition.

“They can often achieve leverage, pricing and documentation terms that would have looked attractive a couple of years ago,” he said.

Two businesses that would previously have received similar financing packages may now receive very different outcomes based on the assessment of AI risk, Hirschmann added.

The divergence is also becoming visible in the level of monitoring lenders require after a deal closes. “Lenders want enhanced KPIs to ensure a strategy is delivering on its stated objectives and to help them understand how a business is evolving against its stated business plan and model,” Miles said.

Any strategy is only of use if it is then deployed to deliver the results intended, he added.

The shift is also affecting private equity sponsors. “We’re seeing sponsors concentrate capital behind businesses they believe can benefit from AI or successfully adapt to it, while becoming more cautious around businesses where the long-term outcome is harder to predict,” Hirschmann said.

But Hirschmann doesn’t believe simply putting more equity into a transaction solves the underlying problem. “I don’t think sponsors are using more equity to offset AI risk, because more equity or a higher valuation cannot solve a structurally challenged business model.”

He said the issue also extends to exit valuations. “If you buy a business today at a certain enterprise value, or EV multiple or if you bought a business at a high EV multiple back in, say, 2022, are you able to exit and realize the asset at the same EV multiple today in an AI-disrupted environment?”

Miles pointed out that changes in debt capacity can also affect transaction structures. “The quantum of debt capable of being raised is one aspect which can impact purchase price/valuation realities, but of course equity cheques are also key, as are other ways of structuring purchase price consideration mechanics. You may, for example, see a further growth in deferred consideration, earn out and vendor loan usage.”
 

The implications could become particularly visible as loans originated before the AI boom approach maturity.

A substantial portion of loans maturing in 2028 that originated between 2021 and 2023, before AI disruption became a central underwriting consideration, Hirschmann said. “In many cases, leverage, valuation and growth assumptions were made without necessarily assessing how AI could affect things like customer behavior, pricing power and competitive dynamics.”
 

As a result, some capital structures may need to be revisited through a much more demanding lens when they refinance, he said.

“Again, some of those businesses that have adapted quickly, invested in AI and demonstrated their AI readiness will be successful when they need to find a refinancing solution,” he added.

In terms of 2021 to 2023 deals, the average repayment date is around a six or seven year term, so deals from the 2021 vintage onwards are an area of focus, Miles said. “In my experience, markets generally find a way of tackling refinancing walls to the extent those arise and exist, and that has proven to be the case historically.”

Clearly valuations and debt levels will be considered carefully as part of any refinancing or sale, and debt and equity structures adapted to meet the requirements, he added.

The more immediate question is whether the underlying business still supports the leverage and valuation assumptions used when the original financing was arranged. “Potentially that is a risk for some businesses, but to the extent debt is being serviced this is a positive in the context of discussions around what a refinance may look like and what terms may apply to it,” Miles said.
 

In the U.K. and Europe, direct lenders should focus on avoiding businesses whose competitive position is eroding, Hirschmann said. “In that sense, AI for us is less about finding the winners and more about avoiding the losers.”

The question for lenders is becoming less about what sector a company operates in and more about how vulnerable its business model is to technological substitution, how defensible its customer relationships are and whether management can adapt.

“Loss avoidance remains the key factor for direct lending,” Hirschmann added.

Disclosure: Funds associated with Permira hold a majority interest in the parent company of Octus.