Hook
A freshly released survey from Lazard reveals that 96% of private equity secondary investors have already altered their approach to software investments—and the same force is now silently fracturing the digital asset market. The crypto sector, long insulated by its own narrative of decentralization, is next in line. The proof is in the flows, not the promises.
Context
Lazard’s 2025 survey targeted institutional investors in the PE secondaries market, a cohort that trades illiquid stakes in private companies. The headline finding: 96% of respondents have changed how they invest in software, with many redirecting capital to other opportunities. The core driver is artificial intelligence—specifically, the belief that AI will commoditize traditional software features, erode gross margins, and invalidate legacy valuation models. Among the 96%, a staggering 91% cited “proprietary data advantage + network effects” as the new moat.
For crypto, the parallels are uncomfortable. The sector’s “software” layer—smart contract platforms, DeFi protocols, infrastructure tools—relies on similar network effects and data accumulation. Yet the same AI-driven commoditization that threatens Salesforce or ServiceNow is now targeting Ethereum’s execution layer, Solana’s validator set, and every rollup that claims unique functionality. Based on my audit experience, the theoretical gap between code and reality is where the real risk lives.
Core: Systematic Teardown of the Crypto Software Moat
I applied the Lazard framework to three crypto software categories: general-purpose L1/L2 execution environments, vertical DeFi protocols, and data infrastructure platforms. The results are sobering.
1. General-Purpose L1/L2 Execution Environments
Take Ethereum’s rollup ecosystem. The promise is that each rollup provides a unique execution environment—Optimism’s OP Stack, Arbitrum’s Nitro, zkSync’s ZK Stack. But AI-generated code can now replicate Solidity-based logic in minutes. GitHub Copilot and similar tools have reduced the cost of building a basic DEX or lending protocol to near zero. The functional differentiation between rollups is collapsing. The true moat lies not in the execution environment but in the network effects of liquidity and user base. Yet AI agents can fragment liquidity further by routing through aggregators, making even the “network effect” moat thinner.
2. Vertical DeFi Protocols
Uniswap V4’s hooks turn the DEX into programmable Lego, but the complexity spike will scare off 90% of developers. This is exactly the kind of “feature-based” value that AI can commoditize. An AI-native DEX could dynamically adjust liquidity curves, optimize swap routes, and even simulate MEV attacks in real time—all without a human developer touching the hooks. The data moat here is the order flow and historical price data, but synthetic data generation is advancing rapidly. The question is: how long until a rival protocol trains a model on Uniswap’s public data and offers a better swap experience?
3. Data Infrastructure Platforms
The Graph, Chainlink, and similar oracles provide data feeds that are essential for DeFi. Their moat is the network effect of data providers and the trust in their validation. But AI can now generate synthetic price feeds, predict volatility, and even detect manipulation. The cost of building a custom oracle using a fine-tuned LLM is dropping fast. The Lazard survey’s 91% consensus on data moats may be correct today, but the half-life of that moat is shrinking.
Contrarian: What the Bulls Got Right
Despite the bleak picture, the Lazard survey also reveals a counterintuitive insight: only 4% of investors have not changed their approach. That means the vast majority have already priced in the AI disruption. In crypto, this panic is already embedded in the capital flows. The contrarian angle is that the most obvious risks—commoditization of smart contracts, AI replacement of developers—are fully discounted. The true alpha lies in the overlooked assets: protocols with unique, non-trainable data assets, such as on-chain identity or reputation systems. For example, a decentralized identity protocol that holds verified KYC data cannot be replicated by a model trained on public blockchain data. That data is inherently exclusive. Yields are just risk wearing a tuxedo, but the risk of over-pricing the AI threat is also real.
Takeaway
The Lazard survey is a mirror for crypto: the same forces that are reshaping traditional software are now at the gates of the blockchain stack. The market has already started adjusting—the 96% number is a signal, not a prediction. The question is not whether AI will disrupt crypto software, but whether the current consensus on “data moats” is correct or if it will become a trap. Assume malice, verify everything, trust nothing. The coming 18 months will reveal which protocols have real, defensible data assets and which are just dressing up their code with AI buzzwords. Complexity is the camouflage for incompetence; the proof is in the logic, not the promise.