AI Is Splitting Software Moats: Why PE Losses Signal Disruption

The $5 billion question

When a Blackstone-led group took control of Medallia from Thoma Bravo last week, the private equity firm wrote off close to $5 billion—the second-largest loss in PE history by one widely cited count. Thoma Bravo had taken the customer-experience software company private in 2021 at $6.4 billion, betting on growth that never materialized. Orlando Bravo has since called it a "big mistake."

The conventional explanation points to interest rates. Private equity bought software at pandemic-peak multiples on cheap debt, underwriting growth that assumed 2021 conditions would persist. But rising rates don't explain the full picture.

Consider Chegg. The homework-help company carried no buyout leverage. It was a profitable public business worth roughly $14 billion in 2021, built on a decade of paying contractors to create tens of millions of step-by-step answers. Then ChatGPT arrived—free and better. Within months, the CEO told analysts the chatbot was hurting growth. Google's AI summaries then choked off the search traffic feeding new sign-ups. By late 2025, Chegg had cut nearly half its staff and its market value sat barely above $100 million. That's a 99% collapse.

Why it matters

When generative AI can produce answers, summaries, designs, or code at near-zero cost, recurring-revenue businesses built on producing that output lose their competitive advantage. The market has dubbed this the "Saaspocalypse," but the reality is more nuanced. AI doesn't kill all software—it splits existing moats in half. The replicable half, what frontier models can now do, stops being defensible. The half AI cannot reach becomes more valuable.

The 3Ds framework under pressure

Alex Lazarow, a venture capitalist and author, argues that durable software businesses—particularly in fintech—possess three core advantages: distribution, data, and delivery. AI doesn't retire this framework, but it cuts each leg in half.

Distribution strengthens when it cannot be replicated by a model. Sitting inside a system of record, holding a license or payment rail, or owning trust that allows financial transactions—these forms of distribution endure. Chegg's distribution was Google referrals; when Google changed how it answered queries, that channel disappeared. Rented distribution evaporates when the channel reprices.

A new complication: buyers themselves are changing. Visa, Mastercard, Stripe, and Google have all shipped payment rails for AI agents. Visa frames the coming year as when agents stop assisting purchases and start completing them. McKinsey estimates the potential U.S. market at up to $1 trillion by 2030. When an agent does the buying, traditional customer acquisition tactics matter less.

Data advantages are where founders most often deceive themselves. Historical data piles lose value rapidly. What survives is live, exclusive, typically regulated data flows that keep generating and cannot be purchased elsewhere. Lending based on a borrower's own transaction history remains a real moat—proprietary, real-time, and consented. Local fraud and credit patterns tied to specific markets hold up because foreign models cannot cheaply acquire them.

Delivery is evolving too. Delight is now cheap; an AI-native competitor can build a polished product in weeks. The durable part of delivery is trusted, accountable execution—money actually settles, compliance holds, and someone is liable when the model fails.

Own the flow

The synthesis of these three dimensions points to a single principle: own the flow. Sit inside the transaction, and distribution, proprietary data, and accountable execution arrive together. The flow generates the data, and the flow is where settlement happens. Operate as a thin layer on someone else's rails, and you hold none of it—you become a feature priced at zero when the underlying engine improves.

Medallia's closest comparable, Qualtrics, also went private in a $12.5 billion Silver Lake deal in 2023. Customer-experience and feedback software, built on generating output now easily replicated by AI, has been particularly vulnerable.

The details were first reported by Alex Lazarow in Forbes, who notes that in markets where hundreds of teams chase the same idea in San Francisco while only a few pursue it in São Paulo or Jakarta, the businesses that own their flow—where data and trust are anchored locally—are the ones AI makes stronger rather than obsolete.