Why Software Thesis Assumptions No Longer Hold in PE

The software thesis that drove private equity returns from roughly 2015 to 2022 was built on four assumptions: that software businesses could grow quickly at reasonable capital efficiency; that multiple expansion would continue to reward growth over profitability; that interest rates would remain low enough that leverage was accretive; and that SaaS business models were structurally defensible against competitive disruption.

All four assumptions have been reset.

Apollo's research is direct: software investing has been reset, and most of the market hasn't fully priced it yet. Higher rates and AI together have rewritten software valuations, growth assumptions, and leverage models. A meaningful share of past software PE returns came from multiple expansion rather than operational improvement. Sponsors holding software assets bought during the multiple expansion era should be running a hard internal review on whether their thesis still holds.

Most haven't.

ST · 01What Has Actually Changed

The four assumption failures are distinct in their mechanisms and require different responses.

Multiple expansion is over. From 2015 to 2021, software multiples expanded dramatically — driven by falling interest rates, increasing market confidence in SaaS economics, and a broadening investor base that included public market crossover funds willing to pay growth rates forward. At peak, software businesses were being valued at 20x, 30x, and higher multiples of revenue. Those multiples reflected expected future cash flows discounted at near-zero rates. When rates rose, the discount rate embedded in those multiples repriced dramatically. Software businesses bought at 2021 multiples need to generate extraordinary operating performance to produce the returns the entry price requires — not because the businesses are bad, but because the financial math of high-rate environments is simply different from the financial math of zero-rate environments.

AI is restructuring software's competitive dynamics. The May 2026 PE CxO Report's most consequential finding for software investors: the same AI that helps one portfolio company can kill another's revenue. Vista Equity's Agentic AI Factory, operating across one-third of its 100 portfolio companies, is already replacing SaaS workflows with AI-native processes. Horizontal SaaS doing rules-based work — the category that constituted a large share of software PE deal flow from 2015 to 2022 — is most exposed. A shakeout that would have taken five years in the open market could happen inside diversified PE portfolios in 18 months.

NRR compression is structural, not cyclical. ICONIQ's State of Software analysis found that net dollar retention has compressed from 110.5% to 107.1% since Q1 2023. 58% of SaaS companies report lower NRR than two years ago. This compression is not a market cycle artifact — it reflects the increased pricing power of enterprise buyers, the availability of competitive alternatives including AI-native tools, and the end of the "land and expand" dynamic that allowed SaaS companies to grow revenues faster than their contract footprints.

Leverage economics have reset. Software businesses bought with 2019-to-2022 leverage structures are servicing debt at rates that were not underwritten at acquisition. The interest coverage ratios that looked comfortable at 3% base rates are stressed at current rates. The cash that was available for growth investment is now servicing debt. The operating model has effectively been levered up — not through intentional decisions, but through rate changes that the original underwriting didn't stress test.

ST · 02What Buyers Are Now Requiring

The buyer universe for software assets has adjusted its diligence framework to reflect the reset assumptions. Sponsors whose exit plans depend on buyers applying pre-2022 diligence criteria will find those plans don't survive contact with the market.

Buyers are now requiring operating evidence rather than category positioning. A software company with a compelling narrative about market opportunity, product roadmap, and competitive positioning but inconsistent NRR, margin compression, and revenue growth below plan is not a premium asset in the current buyer market — regardless of how credible the narrative is. The ICONIQ framework is direct: the new playbook is prove efficiency now, show the AI path to asymmetric upside next.

Retention economics and customer unit economics have become the primary valuation inputs for horizontal SaaS. NRR, gross revenue retention, CAC payback period, and customer lifetime value are being scrutinized with a precision that buyers weren't applying in 2019. The reason is straightforward: in a world where SaaS retention is under structural pressure from AI alternatives, a business's existing customer economics are its primary moat. Software businesses without strong, defensible customer economics are priced accordingly.

AI strategy is now a standard diligence input. The May 2026 PE CxO Report documented the framework buyers are beginning to expect: buy general AI tools, redesign the work around them, then build something proprietary on top. Software businesses without a credible AI story — one that demonstrates how the product uses AI to create defensibility rather than being displaced by AI — face a valuation discount that reflects the risk of AI-native competition.

ST · 03How Sponsors Should Re-Test Their Software Theses

The re-testing discipline for software theses bought in the 2019-2022 cycle requires asking four questions that the original underwriting didn't need to ask.

What is the AI replacement risk? For every software asset, the question is whether the workflows the software automates are candidates for AI-native displacement. Horizontal SaaS doing rules-based workflows — data entry, report generation, schedule management, compliance tracking — faces high replacement risk. Vertical SaaS with proprietary data, high switching costs, and deeply embedded workflows faces lower near-term replacement risk.

What does the business need to be worth at the current multiple environment? Run the exit math at current software multiples rather than at the multiples embedded in the original underwriting. If the exit thesis requires multiple expansion back to 2021 levels to generate target returns, the thesis is broken. The operating performance required to generate target returns at current multiples is the actual operating target — and for many 2019-2022 software investments, that target is significantly above the current trajectory.

Is the NRR trend structural or cyclical? Diagnose whether NRR compression reflects market cycle factors (enterprise buyer caution, IT budget scrutiny) or structural factors (AI competitive alternatives, workflow replacement, pricing power erosion). Structural NRR compression requires a different strategic response than cyclical compression — and requires different communication to buyers.

What is the AI path to asymmetric upside? For software assets that have been challenged by the thesis reset, the most important question is whether there is a credible AI-enabled next chapter: a product evolution, a workflow integration, or a data moat that uses AI to create defensibility rather than being displaced by it. This is the exit narrative that commands a premium in the current buyer market.