Breaking. Lazard's latest survey of private equity secondary investors dropped a bomb that most crypto natives missed. 96% of respondents have already changed their approach to software investing. Not 'considering' — already changed. 91% now define the only viable moat as 'proprietary data advantage + network effects'. The remaining 4%? They haven't touched a thing. That's the signal. Capital is fleeing traditional SaaS. The old playbook — ARR growth, gross margin, feature depth — is being rewritten in real time. But here's what the survey doesn't say: the same capital rotation is already hitting crypto, and most projects are still pretending it's 2021.
Context: The PE Secondary Market as a Leading Indicator
Lazard's survey targets the private equity secondary market — the space where institutional investors trade stakes in private companies before an IPO. It's a shadow market, but it's where the smartest money positions for structural shifts. The survey asked ~200 investors how AI is changing their view of software companies. The result: a near-unanimous repricing of the entire asset class. The 'software premium' is evaporating. Investors are moving capital to 'other opportunities' — and that includes AI infrastructure, data providers, and yes, crypto-native AI protocols.
Why should crypto care? Because the same pattern is playing out on-chain. The DeFi summer of 2020 was about liquidity mining. The NFT summer of 2021 was about digital scarcity. 2024-2025? It's about data — who owns it, who trains on it, who can prove it. The Lazard survey confirms that the traditional software world is undergoing the same shift that crypto has been dancing around for years. The difference is speed. The PE secondary market moves at the pace of quarterly LP meetings. On-chain, it moves at the speed of blocks.
Core: The Data Moat Mirage and the Real Valuation Shift
Let's dive into the numbers. 96% of investors changed their approach. That's not a trend — it's a stampede. The capital that used to back Salesforce alternatives is now flowing into AI-native sales tools, or worse, into AI infrastructure that doesn't even have a product yet. The 91% consensus on data moats sounds like a clear signal, but it's a crowded trade. When everyone agrees on a single factor, that factor is already priced in. The real alpha comes from the factors that aren't yet consensus.
I've seen this before. In 2020, when I audited the Uniswap V2 factory contract, the market was obsessed with liquidity depth. Every analyst talked about TVL. But the real innovation was the constant product formula that enabled direct ERC-20 swaps without ETH — a structural change that most people missed until it was too late. Today, the market is obsessed with 'data moats'. But the structural change is something else: the commoditization of model inference and the rise of verifiable data provenance.
Here's what the Lazard survey implies for software valuation: the traditional DCF model is dead. You can't project future cash flows when the product itself might be replaced by a chatbot in 18 months. Instead, investors are moving to a 'data asset multiple' framework — a rough measure of how much proprietary, continuously updated data a company owns, and how defensible that data is from AI-generated synthetic substitutes. This is almost identical to how the crypto market is starting to value data availability layers like Celestia or EigenDA. The parallel is exact.
But the survey also reveals a blind spot. 91% of investors focus on data + network effects, but they ignore the cost of compute. Every software company integrating AI will face a margin compression crisis — inference costs eat into the 80% gross margins that SaaS investors worship. The ledger never sleeps, only updates. And the update shows that AI-native companies with lower capex for inference will have a structural advantage. In crypto, that means projects that build on efficient L2s or use cost-effective model architectures (like Mistral or Llama) will outrun those that rely on expensive API calls.
Contrarian: The Consensus Trap Nobody Is Talking About
Here's the contrarian angle that Lazard's survey doesn't surface. The 91% consensus on data moats is itself a risk. When everyone wants the same asset, the asset becomes overpriced. The real opportunity is in the 9% of investors who are looking at other moats — compliance expertise, workflow embedding depth, or even regulatory licenses. In crypto, that translates to projects that have secured actual legal clarity for their AI-agent operations, or those that have built deep integration into enterprise middleware.
There's another layer. The survey shows capital flowing 'to other opportunities.' But where? My suspicion is that a significant portion is moving to AI infrastructure — compute, data centers, networking. This is exactly the same rotation we saw in crypto during 2023 when capital flowed from L1s to L2s and data availability layers. The pattern is fractal. The question is: which crypto protocols are positioned to capture this compute and data demand?
Moreover, the survey's emphasis on 'proprietary data' ignores the emerging threat of synthetic data. If AI models can generate high-quality training data that mimics real-world distributions, the moat around proprietary data erodes. I've analyzed dozens of NFT metadata contracts — the Bored Ape Yacht Club case taught me that market narratives often diverge from technical reality. The same is true here. The narrative says 'data is the new oil.' The technical reality is that synthetic data is the new fracking. It can crack the data moat open.
Chaos is just data waiting to be indexed. But the market is indexing the wrong data. While everyone piles into data-rich software companies, the real value is being created in the infrastructure layer that makes data verifiable, composable, and provably unique. That's where crypto-native AI projects shine — zk-proofs for data provenance, decentralized compute networks, and on-chain model registries. These are the assets that the PE secondary market hasn't discovered yet.
Takeaway: The Borderless War for Data and the Speed Moat
What does this mean for the next 12 months? The capital rotation from software to AI is real, and it's accelerating. But the easy money — buying data moats — is already gone. The next wave will be about identifying which data moats are truly defensible and which are synthetic. In crypto, that means projects that can prove data uniqueness on-chain will command a premium. Projects that rely on scraped or public data will be arbitraged to zero.
Speed is the only moat in a borderless war. The PE secondary market is moving slower than the on-chain data flow. By the time the Lazard survey results are fully priced into traditional software valuations, the crypto-native AI stack will already have captured the next cycle. The question is not whether to allocate capital to AI — it's whether you're allocating to the right data layer. The block holds the truth. The ledger doesn't lie. But you have to read it fast enough.
Adapt or get front-run by your own assumptions.