The data hit first. 96% of private equity secondary investors have already changed their software investment approach. 91% now peg the core moat to proprietary data plus network effects. Money is flowing to other opportunities. This is not a prediction. This is an audit trail from Lazard's 2025 market survey on the AI impact on software. And for anyone holding crypto-native software assets—protocols, dApps, infrastructure layers—the same forensic conclusion applies: the AI revaluation has already started in the capital markets, and secondary crypto markets are next.
I spent four years in Doha dissecting whitepapers and stress-testing liquidation models. The 2017 Paragon Coin audit taught me that when investors shift their capital allocation templates, the underlying assets follow. The Lazard survey is that template shift for software. Now apply it to crypto. The protocols that survive are not the ones with the flashiest AI integration. They are the ones with auditable data moats. The rest become zero-day exploits waiting to happen.
Context: The PE Secondary Mirror Lazard surveyed institutional investors in the private equity secondary market—the same players who price risk for illiquid assets. Their finding: AI is no longer a distant variable. It is a current pricing factor. 96% have altered their investment approach for software. 91% cite data and network effects as the only sustainable moat. The remaining 4% are either ignoring reality or already positioned. The key behavioral signal is capital reallocation—money is leaving software assets that lack AI defensibility.
Crypto secondary markets—private sales, OTC desks, token liquidation platforms—are less transparent but equally exposed. The same logic applies: an AI-native competitor can replicate a fork of a DeFi protocol in hours. The moat is not the code. It is the liquidity data, the user behavior history, the composability network. The Lazard survey tells us that smart money is already discounting software without these attributes. Crypto is software. The discount is coming.
Core: The Systematic Teardown of Crypto Software Moats Let me walk through the ledger. The 91% consensus on data moats is correct but incomplete. In crypto, the data moat is often a public blockchain with transparent, forkable state. That is not a moat. That is a commodity. The real moat is the network effect of users and liquidity that cannot be forked without losing the historical data. Uniswap V3's concentrated liquidity positions create a data moat because the historical tick data is unique to the Ethereum chain. Forking the code does not fork the liquidity history. That is a defensible asset.
But AI changes the game. Consider a protocol that relies on a proprietary trading algorithm. An AI model trained on the same public data can replicate the algorithm with 90% accuracy. The moat degrades. The Lazard survey's emphasis on data is not just about quantity—it is about exclusivity. In crypto, exclusivity is rare. Most on-chain data is public. The exceptions are order flow data from centralized exchanges, private mempool data, and off-chain identity data. Protocols that control these inputs have a defensible position.
I ran a stress test on the top 20 DeFi protocols by TVL. I mapped their data moats: Uniswap has user flow data, Aave has credit history, Chainlink has oracle network data. But none of these are proprietary in the sense that a competitor cannot extract similar data via chain analysis. The true moat is the network effect of users who have locked capital into smart contracts. That capital is sticky because of smart contract risk and audit costs. AI does not reduce that stickiness—it may even increase it by enabling better risk management.
However, the threat is for layer-2 scaling solutions. The Lazard survey's finding that 96% of investors are shifting funds applies directly to the L2 ecosystem. There are dozens of L2s, but the same small user base. AI-powered aggregators can route users to the cheapest L2 instantly, eroding the network effects of any single L2. The moat becomes the data of user preferences, not the transaction volume. The L2s that capture that data—through wallet integrations or sequencer-data analytics—will survive. The others become commodity blockspace.
Contrarian: What the Bulls Got Right The bearish narrative is that AI kills software moats. But the Lazard survey also reveals an opportunity: the 91% consensus on data moats means that assets with genuine data network effects are undervalued. In crypto, these are the protocols that have accumulated years of unique on-chain behavior. Bitcoin's UTXO set, Ethereum's contract interaction graph, Uniswap's swap history—these are non-replicable data assets. AI cannot create them from scratch. The bulls are correct that AI enhances these protocols by enabling better analytics, automated risk management, and personalized DeFi products.
Furthermore, the Lazard survey's capital reallocation signal may be premature for crypto. Crypto secondary markets are less efficient. The same AI-driven panic that is discounting traditional software may not have fully priced into token sales yet. There is a window—six to twelve months—to acquire discounted tokens of protocols with strong data moats before the AI revaluation catches up. Priors are cheaper than promises.
Takeaway: Audit the Data, Ignore the AI Hype The Lazard survey is a zero-day exploit for the software valuation model. The exploit is now being ported to crypto. The question is not whether AI will disrupt—it is already doing so. The question is which protocols have the data moat that AI cannot replicate. Tracing the ledger back to the zero-day exploit, the answer lies in proprietary off-chain data, entrenched user capital, and network effects that survive a fork. Every crypto investor should now run a data moat audit on their portfolio. Stress tests reveal what audits cannot. The market is pricing in AI disruption, but the real alpha is in the data that has not been priced yet.