The software industry is currently navigating a “dramatic” period of repricing and scrutiny, driven by the emergence of large language models (LLMs) and a fundamental shift in how the market values technology. As the era of “software eating the world” transitions into AI potentially “eating” software, investors are grappling with a sharp bifurcation between companies with defensible moats and those facing terminal risks.

In a conversation on 28 July, Claus Mansfeldt explored the transition in public and private companies from seat-based to outcome-based pricing, the impact of AI on cost structures, and the importance of a disciplined, sector-agnostic approach in a market where even the most “unassailable” incumbents are being challenged.

Sylvain Barrette: What is your assessment of the current ‘software crisis’ in both public and private markets?

Claus Mansfeldt: The situation is dramatic. We are seeing a significant repricing of software companies and an intense debate regarding the durability of their business models in the face of AI. This has led to a clear bifurcation between the ‘haves’ and the ‘have-nots.’

While the market often gets things wrong with hindsight, it is currently reacting appropriately—through pricing adjustments—to fundamental changes in the technological landscape. Software has been underperforming for the last three to four years as winners shifted towards semiconductors and equipment providers that support AI.

We are seeing a deceleration in Annual Recurring Revenue (ARR) growth; companies that once grew at 15% to 20% are now expanding at 8% to 10%. Changing the growth rate by one or two per cent for the next ten years has a dramatic impact on value when using a discounted cash flow approach.

How is AI specifically threatening established software business models?

C.M.: There is a major distinction between vertical and horizontal software. Vertical software—which is specialised for industries like healthcare or finance and deeply entrenched in client workflows—tends to have a better ‘moat,’ a defensive competitive advantage.

Horizontal software, which is more of a ‘jack of all trades,’ such as digital centres, accounting software, or human resources specialists like Workday, is more easily replaced by LLMs. We are even seeing individuals recreate expensive tools using AI; one investor recently built a personal clone of a Bloomberg terminal for just €2,000 a year by using an AI interface and a cheap data feed, rather than paying $30,000 to $50,000.

While this might not yet be replicable at an enterprise scale, the threat to pricing power is real. If a software company asks for its usual 7% price increase, clients may now point to their need to reallocate budgets towards hardware, consultants, or AI tokens.

Do you expect AI to improve the margins of software houses even if the top line slows down?

C.M.: AI will certainly bring productivity gains, particularly for software engineers. On the cost side, software houses could be the main beneficiaries of AI. However, the question remains whether these margin improvements can offset the potential top-line slowdown caused by budget reallocation and increased competition from AI-native startups.

How do you see pricing models evolving for Software as a Service (SaaS)?

C.M.: Traditionally, SaaS was priced per ‘seat’—a subscription for every employee with access. We are now moving towards ‘outcome-based’ pricing. If AI agents are doing the work, a company might only need one general subscription rather than ten, and they will pay for the result or the output rather than the number of users. This is where the industry believes we are going to end up. This shifts the economics significantly.

With software stocks like Salesforce and Adobe seeing significant drops, do you see that as a development signalling attractive software companies in the private market?

C.M.: It is very difficult to call the bottom of a correction. While some suggest software growth has never been cheaper, we prefer to remain disciplined and sector-agnostic. Year-to-date, 17 of the 30 worst performers in the S&P 500 have been software houses, with some stocks down 30% to 40% in a rising market.

We treat software companies like any other company through the review of their client concentration, their technology position versus peers, and their relative share in the market. We also rely on external experts to perform the technological due diligence.

We are also focusing on traditional metrics like Ebitda, free cash flow, and operating leverage, while taking a new look at Capex. They are less asset-light than in the past. This new focus does not only apply to public market companies; the same is true in the private market. Everything becomes obsolete very quickly.

What is the software exposure in your funds?

We maintain a 10–15% allocation, which is lower than the 20–30% typical in the US. We prioritise diversification rather than concentrated exposure. A €300m fund might hold 20–30 software firms among 300 total investments, with single-name exposure capped at €10m (around 3%).

Currently, the market is discounting much higher risk for these assets than in the past. However, ignoring the sector would be a mistake, as it still offers the potential for ‘home runs’ that can lift an entire portfolio's performance.

There has been concern about private credit funds overextending themselves in deals like Medallia, a software company acquired by private equity firm Thoma Bravo. Should the limited, if any, financing coming from banks for the Medallia deal have alerted private credit funds?

C.M.: The boom in private credit fundraising, particularly from retail and high-net-worth investors, has created giant machines under pressure to deploy capital. This can lead to a ‘herd mentality’ where funds participate in syndicated deals just to move volume.

We have learned that there is no safety in numbers or ‘big brands’—everyone can make mistakes. It is tempting to invest alongside large players. That said, the strategy is not fundamentally broken. Even if there are defaults, the net returns for credit funds can still be a healthy 7% to 8%—down from high single digits to 10% (see Chart 1)—should recovery rates remain high. The interest rates and the margins they could charge were very ‘juicy’, and they have not suffered from massive write-offs.

Are you tempted to allocate a greater exposure nowadays to AI firms at the expense of software companies?

C.M.: We don't really do venture; we don't have those mandates, and our investors are not asking us to look into that. We are in the buyout game, which means we typically invest in profitable companies. Alternatively, we may consider companies active in AI infrastructure that demonstrate recurring revenue and are profitable with a strong moat.

Finally, what is the state of the secondary market and the ‘exit crisis?’

C.M.: The idea of the ‘distressed seller’ is still largely a myth. Most sellers are simply looking to optimise their portfolios. Buyout assets are currently trading at roughly a 13% discount to their last reported Net Asset Value (NAV), a level much lower in venture or growth funds. For good funds, you still have to pay par or close to par.

While there is a perceived lack of exits, the nominal volume of deals last year was quite high. The issue is that the ‘exit magnitude’ is small relative to the massive inventory built up in previous years. We prefer being a buyer—while not instigating a transaction—in the secondary market rather than a seller, as valuations have become more attractive and there is less competition for deals. Investors have become more disciplined and learned the lessons of the Medallia debacle.