• Global sponsor software exits down 69% YoY, buyouts down 54%
• Proving terminal value trumps immediate earnings or growth
• Successful deals require strong AI story and AI readiness

Every software outcome hinges on its “AI story,” according to Houlihan Lokey’s global head of technology Sascha Pfeiffer. The so-called SaaS apocalypse has definitively fragmented the industry into tiers in the minds of private markets investors, with artificial intelligence (AI) acting as the ultimate sorting mechanism.

Software businesses are no longer underwritten on growth alone. Instead, investors are asking a more existential question: how defensible is this company in an AI-driven world?

“We don’t even take processes with an AI risk,” Pfeiffer said, an observation that underscores how the old software playbook has faded, replaced by a harsher calculus around survivability.

This shift is driven in part by rapid advances in AI itself. The arrival of autonomous coding agents such as Anthropic’s Claude Code and OpenAI’s Codex poses structural challenges, raising the prospect that core functions of traditional software businesses could be eroded or replaced in five or 10 years.

“The core threat is structural, we have to acknowledge that,” said Daisy Cai, a general partner at B Capital, an early to growth-stage investor active in technology. “Coding agents are becoming mature enough to generate code as well as humans. They can commoditise a lot of software development.”

The sharp repricing of software assets has therefore hinged more on uncertainty regarding their terminal value rather than this year’s earnings or growth. “Most of these companies are still hitting their numbers,” said Benedikt Joeris, a partner at Hg Capital. The question is how long they continue to exist in their current form.

That uncertainty already seems to be rattling private equity M&A activity, as buyers hold back and reassess not just pricing but long-term viability.

Both sponsor exits and buyouts have suffered in this market, with the deal volume for the former down globally 69% YoY to USD 102bn and the latter down 54% to USD 52bn. On both sides, volume is tracking at its most depressed since 2020.

Broader global software M&A data does tell a different story, with USD 623bn of transactions recorded this year through June 2026, up roughly 65% YoY. Volume is the highest since 2022.

Deal value is supported by a small number of outsized transactions, mostly AI-related. These include funding rounds for OpenAI and Anthropic of USD 122bn and USD 65bn, respectively. The OpenAI deal also stands as the largest on record. Overall, investors have become highly selective, concentrating on fewer, higher-conviction opportunities with a bias towards perceived AI champions.

Tiers for fears 

Software once accounted for an estimated 20%-30% of PE activity. Now the market is in limbo: deals launched and pulled, processes dragging out, or investment committees backing away at the last minute even after reaching exclusivity, according to one M&A lawyer. Private equity is entering a “no man’s land,” they said, where conviction is to proceed with deals is increasingly difficult to build.

But out of the uncertainty, a clearer structure is beginning to emerge. Sponsors are assessing the extent to which individual business models are exposed to AI disruption – whether positive or negative – and reclassifying the industry into distinct tiers.

Specialist software investors such as Netherlands-headquartered Main Capital Partners have formalised this into internal frameworks. Late last year, it developed a five-tier system ranking SaaS companies from most insulated to most vulnerable to being replaced by AI, according to Bram Kaashoek, a partner and COO at the firm.

Those in tier one own proprietary data and core system records, and their services are typically adopted by customers on a company-wide basis. In tier two, services enable company-wide procedures, but SaaS providers do not own or control them. For example, they might be delivering asset management solutions or IT operations management.

Tier three representatives support specific department-level tools – for legal, compliance, product development. These types of software are crucial, but systems failures do not prevent entire organisations from operating, Kaashoek said.

Tier four has productivity tools for generic work processes such as analytics or forecasting, while tier five would include point solutions with narrow functionality, such as scheduling.

“Tiers three to five are under the biggest threat for AI to replace them,” Kaashoek added.

Public markets have already validated this stratification. Companies in the first two tiers – particularly enterprise resource planning (ERP) software providers, critical data, or security infrastructure providers – have held up relatively well, while valuation compression has been most severe among lower-tier assets, he said.

Other advisers frame the same insight differently but reach similar conclusions. Houlihan Lokey, for example, groups software into three buckets: businesses that are effectively “doomed” by AI, incumbents that will survive, and companies for which AI will be tailwind.

Luckily, only a minority, perhaps 10%-15%, may fall into the first category, said Pfeiffer, but they do present a challenge for GPs that own them.

Crucially, the distinction cuts across traditional classifications. Vertical software is often perceived as safer – it pushes deeper into organisational structures than horizontal products that wrap around existing systems – but penetration varies within sectors, said Main’s Kaashoek. A core human resources platform might sit in tier one, while a simple feedback tool falls into tier five.

What matters is not the label, but the depth of integration, ownership of data, and whether the software delivers mission-critical outcomes, he explained.

Selling the story

For decades, software investing followed a relatively consistent formula: prioritise recurring revenue, pursue growth at scale, and rely on multiple expansion on exit. That is now being rewritten. AI proof points have gone from a “nice to have” in the equity story to something buyers expect.

Hg’s Joeris referenced GTreasury, which his firm backed in 2023 and sold to US-based Ripple last October for USD 1bn. The company, a provider of treasury management software used by CFOs, developed an agentic AI tool called GSmart AI that executes administrative tasks previously performed manually and proactively identifies risks and variances.

GTreasury was able to show solid AI proof points, with tangible increases in commercial traction – 40% of all new customer inbounds came from GSmart AI-related content. Additionally, within two months of launch, the AI tool achieved 100% weekly active usage of across the existing install base, Joeris said.

In practical terms, that means diligence has expanded. AI is now a standalone workstream in almost every deal, with buyers probing not only technical capabilities but also competitive positioning against future AI-driven entrants. Moreover, selling a story has become as important as analysis.

“Nobody really knows how AI will impact each business,” said Houlihan Lokey’s Pfeiffer. “You can speak to 20 experts and get 20 different answers. So, crafting a credible story – where AI is an opportunity, not a threat – is critical.”

That creates a new burden on sellers. It is no longer enough to demonstrate strong key performance indicators (KPIs), they must also convince buyers the business will remain relevant in an AI-enabled economy.

Despite a broader slowdown, there is still demand for assets with the right profile. Advisors note continued activity in cybersecurity, industrial software, healthcare technology and life sciences technology. Businesses with strong regulatory moats, mission-critical functions, or tangible, real‑world applications also attract interest.

Notable transactions include Thoma Bravo’s recent LBO of occupational healthcare software provider Padoa for around EUR 600m, Apax’s acquisition of supply chain compliance platform Sedex Information Exchange, and Nord Holding’s investment in healthcare IT managed services provider VisionmaxX.

The ongoing auction of UK buildings maintenance software provider SFG20 has also been attracting a wide range of PE suitors, as reported.

More broadly, buyers are gravitating toward companies with three defining characteristics: ownership and control of valuable data, solutions deeply embedded in customer workflows, and an ability to deliver deterministic, reliable outcomes where errors are not acceptable. Where those are present, advancements in AI will strengthen competitive advantage, not erode it, argued Kaashoek.

Uncertainty in private equity markets tends to favour strategic buyers. For the year through June, there were 42 sponsor-to-sponsor exits worth an aggregate USD 6.5bn. This compares to sponsor-to-strategic volume of USD 31.9bn across 81 deals over the same period in 2025, according to Mergermarket data.

Strategic acquirers are often more comfortable underwriting AI risk and can justify acquisitions as a way to embed capabilities internally or mitigate disruption, Pfeiffer observed. This dynamic is also driving further industry consolidation, both within traditional software segments and among AI vendors themselves.

Rewriting value creation

As underwriting has changed, so too has value creation. For GPs, improving margins or accelerating sales is no longer enough – they must make portfolio companies AI-ready or, ideally, AI-native. That requires investment in additional operational capabilities.

Hg, for example, has a team of around 150 AI specialists, including engineers and product leaders. Through its in-house AI product engineering team, Catalyst, the firm embeds tiger-teams inside portfolio companies to accelerate the development and launch of AI products. Joeris described it as a version of Palantir’s forward deployed engineering model.

Hg will also help companies reimagine and rebuild how their core functions operate in an AI-first way, rather than bolting AI onto old processes.

“Our strategic advantage is that we do this systematically, across around 60 companies at once. We’re not running a one-off project in a single business and hoping it works. A product build that works in one company becomes the playbook we run in the next, and the learning compounds across the portfolio,” Joeris said.

“That’s the hard part, and it’s what separates managers now. Can you genuinely help a company build new products, repeatably? That’s not a capability you can build overnight. We started early, and we’re still investing heavily.”

Main outlines three primary levers that can be pulled to move companies through the AI-risk tiers: embedding AI into existing products to enhance functionality and monetisation; expanding into emerging AI categories, such as managing AI agents; and transitioning from tools to execution platforms, where software performs tasks rather than supports them.

The last has the potential to be the most transformative, Kaashoek explained. By automating workflows directly, companies can move beyond IT budgets and tap into broader operational spend, with the potential to even replace certain roles or tasks within customer organisations.

Rewiring, repricing laggards

The valuation corrections that have routed listed software companies are visible in private markets as well. High-quality assets previously commanded EBITDA multiples well above 20x and annual recurring revenue (ARR) multiples of 10x or more, Houlihan Lokey’s Pfeiffer noted. Now, high teens would be the ceiling for EBITDA multiples, with ARR multiples in the 4.5x to 8x range.

For more challenged businesses – particularly those in tiers four and five under Main’s system – the outlook is unclear. Some will be repositioned, moving up the value chain through product expansion or consolidation. Others may survive as part of broader platforms, but some may simply lose their relevance.

Private equity owners may look to buy more time for turnarounds through continuation vehicles (CVs), but these are unlikely to provide a way out; secondary investors would be equally discerning over the health and terminal value of assets.

In some cases, GPs may simply accept lower valuations and exit. In others, writing down positions could become unavoidable, said Chris Townsend, a partner at Ropes & Gray.

This marks a departure from previous cycles, where underperforming assets could often be restructured or held until market conditions improved. Waiting is not an option amid concerns of AI displacing the underlying product.

Repricing is particularly problematic for assets acquired between 2020 and 2024, when entry multiples were significantly higher. Unless those businesses have delivered exceptional growth, many cannot be exited today at acceptable returns. However, some private equity firms are now beginning to sell regardless as they feel the pressure to return capital from older funds.

“A lot of funds are already behind in their deployment plans. They’ve already lost a year and cannot afford to lose another two years because of the five-year deployment phase,” said Pfeiffer, who expects sponsor-backed M&A from the 2020-2024 vintages to pick up from early 2027.

Not all industry participants consider AI disruption to be catastrophic. Indeed, some argue that adverse reactions have been amplified by the public markets shock rather than shaped by clear-eyed assessment of the long-term impact on businesses. “The noise around software is overblown,” one LP said. “The market will adapt, as it always does.”

There is also recognition that AI will inevitably expand – not contract – investment opportunities and outcomes. The automation of cognitive work, for example, could push the software industry far beyond its traditional boundaries.

“Software has been used to organise work,” Hg’s Joeris said. “Now it can automate it. That takes the opportunity from the roughly USD 1tn software market toward the USD 60 trillion-plus cost of cognitive labour.”

What is clear is that the industry has entered a new phase, where players can no longer rely on rising multiples or predictable playbooks. Instead, they must navigate a landscape in which success depends on AI positioning as much as financial performance. The consequence is a sharper divide between winners and losers, and less room in the middle.

“To win today, you need three things: a credible AI story, real AI readiness, and strong KPIs. That’s it,” said Pfeiffer. In a market defined by uncertainty, that may be the closest thing to a new rulebook.