91% of Lazard's surveyed PE investors now agree: proprietary data and network effects are the only moats. Only 4% stuck to their old playbook. This isn't opinion—it's a paradigm shift in software valuation.
I've seen this pattern before. Back in 2020, when I traced flash loan arbitrage paths on Uniswap V2, the consensus shift happened in hours—not months. But this Lazard survey is different. It's a structural signal, not a fleeting trade. The survey, released in mid-August (likely 2023–2025 range), captures the moment when institutional capital stopped debating 'if AI will disrupt software' and started pricing 'how disruption happens, which companies survive.'
Context: What Lazard Actually Found
Lazard, a bulge-bracket investment bank with a strong PE secondaries advisory practice, polled its limited partners, general partners, and intermediaries. The headline: 91% of respondents now view 'proprietary data combined with network effects' as the core moat for software companies. Just 4% have not changed their investment approach. The other 95% are either actively shifting capital to other opportunities or adopting a 'wait-and-see' posture. This isn't a 50-50 split—it's a landslide. In normal markets, investor consensus on a new thesis hovers around 60–70%. Hitting 91% means the market has already priced in a regime change.
Core: The Valuation Vacuum
The old valuation framework for software—EV/Revenue multiples based on growth rate, gross margin, and net dollar retention—is collapsing. The new framework is still being built: base multiple × AI exposure discount × moat quality premium. The 91% consensus on 'data + network' is essentially the market assigning a massive weight to moat quality. But here's the rub: there's no standardized way to quantify moat quality yet. That creates a valuation vacuum—a gap between the old pricing mechanism (which no one trusts) and the new one (which isn't standardized). In my experience running a crypto news aggregator, this vacuum is exactly where alpha lives. It's an arbitrage of information asymmetry.
Let me break down the mechanics. Traditional SaaS metrics (CAC, LTV, churn) are losing explanatory power because the cost structure is shifting. If a software company adds AI features, its gross margin may drop from 80% to 60% due to inference costs. The same revenue growth might now produce lower profits. The old multiples don't capture that. The new model must incorporate: (1) data asset uniqueness, (2) network density and defensibility, (3) AI integration depth, and (4) inference cost impact on unit economics. Arbitrage isn't just liquidity waiting for a mirror. It's the gap between what the market pays for a software asset now and what it will pay once the new framework solidifies.
Contrarian: The Consensus Is Too Clean
Every 91% consensus should be stress-tested. Here's what Lazard's survey doesn't tell you: synthetic data, federated learning, and the relentless expansion of context windows (from 4K to 1M+ tokens) may erode the 'proprietary' nature of data over time. If a competitor can infer your proprietary data through API interactions or model distillation, your moat is a mirage. Also, the survey glosses over the 'reliability moat'—in B2B software, determinism still matters. Launch day is a promise; the code is the betrayal. AI models hallucinate; traditional software doesn't. Investors may be underestimating the value of reliability in enterprise workflows. That's a blind spot.
Another angle: the 91% consensus may be a 'groupthink' artifact. If everyone agrees that data is the moat, they'll all pile into the same data-rich assets, creating a crowded trade. The real alpha might lie in identifying software companies that can build network effects on top of AI—not just as a feature, but as a platform. Think of AI-native collaboration tools where the network itself generates training data, creating a flywheel that's harder to replicate than a static dataset. Influence flows where attention bleeds. If attention is bleeding from legacy SaaS to AI-native platforms, the moat shifts from data-custody to data-generation velocity.
Takeaway: What to Watch Next
Over the next 12–18 months, expect a wave of software M&A—the 'middle tier' ($10M–$100M ARR) will be crushed. Large platforms (Microsoft, Google, Adobe) will acquire data-rich vertical SaaS to fuel their AI stacks. Secondaries investors will demand higher discounts for software assets without clear data moats. The key signal: watch for a 'catalyst event'—a major software company issuing a profit warning due to AI substitution, or an AI-native product hitting a million users overnight. That's the moment the valuation vacuum collapses into a new equilibrium.
For now, the playbook is simple: identify companies with genuine data barriers (not just marketing claims) and the ability to deploy capital to activate that data (compute, fine-tuning, RAG). Buy the fear, sell the consensus. The 91% is a warning: the herd is already moving. The only question is whether you're ahead of it or behind it.
