Tracing the ghost in the blockchain’s memory – except here the ledger is not a distributed ledger, but the private equity secondary market’s pricing book. A recent survey by investment bank Lazard dropped a seismic signal: 96% of investors in software-focused PE secondary funds have already altered their investment approach because of AI. And 91% now believe the only durable moat is proprietary data plus network effects. The numbers are stark, but the story beneath them is far more treacherous.
Where liquidity flows, stories drown. The PE secondary market has long been a quiet bellwether for institutional sentiment. When LPs sell their stakes in software companies at a discount, it’s not just a liquidity event – it’s a narrative vote. The survey reveals that capital is actively fleeing software allocations, seeking shelter in other opportunities. This isn’t a tactical rotation; it’s a structural reassessment of what software is worth when AI can replicate its core functions at near-zero marginal cost.
Parsing truth from the noise of new value – the 91% consensus around data moats feels like a safe harbor. But in my years consulting on narrative strategy for institutional clients, I’ve learned that when everyone agrees on a single factor, that factor is already priced in. The real alpha lies in what the consensus ignores: the fragility of data exclusivity, the rising cost of AI inference, and the cold-start problem facing AI-native startups.
The Hook: A 96% Behavioral Shift
Let’s sit with that number. Ninety-six percent of surveyed investors have changed how they approach software investments. That’s not a marginal adjustment; it’s a near-universal pivot. The survey, conducted by Lazard in early 2025, polled a cross-section of PE secondary market participants – fund managers, institutional LPs, and intermediaries. The key finding: AI is no longer a distant technological variable; it’s a capital pricing variable embedded in every deal.
Context: The Historical Narrative Cycle of Software Valuation
Software valuation has always been a narrative game. In the 2010s, the story was “SaaS multiples are justified by recurring revenue and high gross margins.” ARR growth was the god. Then came the 2022 correction, and the narrative shifted to “profitable growth.” Now, AI has injected a new layer of uncertainty. The old story – that software companies enjoy network effects and switching costs – is being challenged by a new one: AI can dissolve those moats by making interfaces conversational and code generation automated.
The Lazard survey captures this transition mid-flight. Investors are not waiting for proof; they are acting on anticipation. The 96% figure is a behavioral signal that the market has already priced in a significant probability of disruption. But what kind of disruption? The survey doesn’t specify timeframes, and that’s where the nuance lives.
Core: The Narrative Mechanism and Sentiment Analysis
The 91% Consensus: Data + Network Effects as the Only Moat
At first glance, this seems logical. AI models are becoming commoditized – open-source models like Llama and Mistral are closing the gap with GPT-4. So the differentiation shifts to data: who has the unique, high-quality, domain-specific data that can fine-tune models? And network effects: who has the user base that generates more data, creating a flywheel?
But let’s apply the Skeptical Storyteller lens. The 91% consensus is a narrative in itself – a story that investors tell each other to justify their risk models. The danger is that this story becomes a self-fulfilling prophecy, causing capital to flow only to companies that fit the “data moat” template, while ignoring software companies that have other, less obvious advantages.
The Hidden Signal: Capital Is Fleeing, Not Just Rotating
The survey mentions that investors are “moving capital to other investment opportunities.” This is not just a sector rotation from software to AI infrastructure; it’s a broader risk-off move. In my experience advising fund managers during the 2022 crypto winter, similar language preceded a multi-year repricing of digital assets. The current signal suggests that software as a whole is being viewed as a “toxic exposure” until the AI narrative clarifies.
But here’s the contrarian twist: when capital flees en masse, it creates mispricings. The PE secondary market, with its illiquidity premium, is ripe for investors who can separate the AI-panicked sellers from the genuinely vulnerable companies.
The Mechanism: AI as a Forced Revaluation of Intangible Assets
Software companies have long traded on intangible assets: code, customer relationships, brand. AI threatens to commoditize the code layer, but it also creates new intangibles: AI training data, model fine-tuning expertise, and compliance trust. The Lazard survey suggests investors are already weighting these new intangibles, but the framework is immature. There is no standardized “data asset multiple” yet. This uncertainty is what drives the 96% behavior change – investors are guessing, not calculating.
Contrarian Angle: The Blind Spots in the Consensus
Blind Spot #1: Synthetic Data and the Fragility of Data Moats
The 91% consensus assumes proprietary data is a durable barrier. But what happens when synthetic data generation matures? Companies like OpenAI and Anthropic are already using synthetic data to train models. If high-quality synthetic data can replicate the patterns of proprietary datasets, the moat evaporates. I’ve seen this pattern before in the crypto world – when on-chain data was thought to be a moat for DeFi protocols, but oracles and cross-chain bridges eroded that advantage within two years.
Blind Spot #2: The Cost of AI Inference
Software companies integrating AI face a hidden cost: inference compute. A traditional SaaS product might have 80% gross margins. Add AI features, and those margins can drop to 50-60% due to GPU costs. The Lazard survey doesn’t mention this, but it’s a critical factor. Companies with strong data moats but high inference costs may see their profitability compress, making them less attractive in a high-interest-rate environment.
Blind Spot #3: The Cold-Start Problem for AI-Native Startups
While investors fear AI-native startups disrupting incumbents, they overlook the cold-start problem. New AI companies have no proprietary data and no network effects. They rely on generic foundation models and hope to build a data flywheel. But that takes time – and capital. The real threat to incumbents is not the startup but the incumbent’s own ability to integrate AI. The market is pricing disruption as if it’s inevitable, but execution risk is high.
Blind Spot #4: Regulation as a Moat
In regulated industries (healthcare, finance, legal), compliance know-how and trust are moats that AI cannot easily replicate. The survey’s 91% consensus ignores this. A software company with decades of regulatory filings and auditor relationships has a defensible position, even if its code is generic. This is where my own cybersecurity background kicks in: I’ve audited smart contracts for DeFi protocols that had strong code but weak compliance – and they failed. The same will happen in AI-augmented software.
Takeaway: The Next Narrative Cycle
The Lazard survey is not a prediction; it’s a snapshot of a narrative in motion. The 96% behavior change tells us that the market is already acting on a story of AI-driven disruption. But stories have endings, and the next chapter will be written by those who see beyond the consensus.
The next narrative will be about “AI-adjusted DCF” – a valuation framework that incorporates inference costs, data moat durability, and regulatory risk. Investors who can build that framework now will be the ones buying software stakes at a discount today and selling them at a premium when the AI panic subsides.
Minting moments that outlast the cycle – that’s the real work. The chaos of the current repricing is the curriculum. The question is: are you parsing truth from the noise, or just following the herd?