91% of institutional investors now believe that 'proprietary data + network effects' are the only moats for software companies. That's a consensus so strong it's a red flag. Chaos is opportunity. Compile the data.
Lazard's latest survey on PE secondary markets dropped a bombshell: the entire software investment thesis is being rewritten. Only 4% of respondents haven't changed their approach. The rest are either shifting capital to other assets or waiting for clarity. This isn't a hedge—it's a rout. The bear market in software valuations has just begun.
Let's break down what this survey actually reveals. The headline: 91% agree that 'proprietary data + network effects' define the new moat. Sounds like a smart-money consensus. But digging deeper, the technical logic is flawed. These investors are treating LLM capabilities as a public good—which is true today. But they're ignoring the compounding effect of synthetic data, federated learning, and context windows that will soon absorb even niche datasets. The moat is not static; it's a moving target.
Context: Lazard's survey targets PE secondary market players—the folks who buy and sell stakes in private software companies. Their collective stance signals a paradigm shift in valuation. Traditional metrics (MRR multiples, NDR, gross margins) are being replaced by an 'AI exposure discount' framework. The market is pricing in a future where most software becomes a commodity, and only platforms with defensible data assets survive. This is the same logic that drove the 2022 crypto crash when investors realized most DeFi protocols had no real revenue. Narrative broken.
Core Analysis: The technical underpinnings of this shift are straightforward. LLMs have reached a point where generic code generation, document processing, and customer support are near-zero marginal cost. The value has moved upstream to the data layer. But here's the catch: proprietary data is only a moat if the model cannot infer it. With context windows expanding from 4K to 1M+ tokens, and agents that can query APIs, the assumption that 'data is hard to replicate' is an artifact of 2023 thinking. I've seen this before. In 2021, everyone thought NFT minting bots were the edge—until the mempool became transparent. Same pattern.
My experience auditing EigenLayer's restaking mechanism in 2023 taught me a lesson about consensus. Everyone agreed that restaking was safe—until the first slashing event. The 91% figure here is a statistical anomaly. Typical investor surveys show 50-70% agreement on trends. 91% means the herd has already moved. That's when the true contrarian bets pay off. Shorting the dip.
Let's quantify the risk. The survey implies that software companies without proprietary data or network effects will face a 40-70% valuation compression. For crypto-native software—think DeFi frontends, NFT marketplaces, or data analytics platforms—the same applies. Projects that rely on open data (e.g., on-chain analytics) have no moat. Those with exclusive feeds (e.g., institutional trading data) have a temporary edge. But the edge decays as AI models learn to synthesize public data. The real alpha is in identifying which 'data moats' are illusions.
Contrarian Angle: The market is overestimating the stickiness of data moats. Consider the rise of open-source models like Llama 3. Any software company can fine-tune it on their own data. That means the 'exclusivity' of proprietary data is shrinking. The true moat is not the data itself, but the feedback loop between network effects and AI—a dynamic that few companies have mastered. Crypto projects like Chainlink or The Graph have this loop. Most SaaS products do not. The 91% consensus is a bet that the current AI infrastructure will remain static for 3-5 years. That's a dangerous assumption.
From my experience shorting LUNA in 2022, I learned that when a valuation framework collapses, the speed of repricing is brutal. The same is happening now. PE secondary market investors are 'waiting for clarity'—which is a euphemism for 'waiting for someone else to take the first loss.' The arbitrage opportunity is in front of us: short the legacy software names that lack genuine data moats, and go long on AI infrastructure tokens that power the new stack. The window is closing.
Takeaway: The Lazard survey is a confirmation that the software industry's valuation paradigm has shifted. But the consensus is too tight. The real winners will be those who can synthesize public data with proprietary signals—not those who sit on a static dataset. For crypto traders, the play is clear: monitor the discount rates on PE secondary software deals as a leading indicator. When the discount widens past 30%, the bottom is near. Until then, Liquidity dries up. Watch the spreads.
Chaos is opportunity. Compile the data.