Hook
Over the past quarter, a quiet signal emerged from the private equity secondary markets: 96% of investors have altered their approach to software investing. For the crypto-native reader, this is not just a traditional finance tremor—it's a blueprint for how AI will dismantle the value propositions of many blockchain-based software projects. The data comes from a Lazard survey of PE secondary market participants, but its implications echo directly into the decentralized world. When 91% of those investors identify 'proprietary data advantages and network effects' as the sole moat worth defending, we must ask: Does this consensus actually hold for DeFi protocols, Layer2s, and crypto infrastructure? Or is it a narrative trap that will lead to mispriced assets?
Context
The Lazard survey, conducted in mid-2025, captures the mindset of capital allocators who trade in illiquid software stakes. The key finding: 96% of respondents have already changed their investment approach due to AI, and money is flowing out of traditional software into other opportunities. Only 4% remain unchanged. The consensus belief is that the only durable moat is a combination of hard-to-replicate data and network effects. But this framework was built for centralized SaaS giants—Salesforce, ServiceNow, Microsoft. Applying it to blockchain-based software requires a fundamental rethinking. Crypto projects are by design permissionless and often open-source. Their data is public, their network effects are transparent, and their moats are rarely what the venture capital playbook defines. Tracing the ghost in the machine, I've watched this narrative unfold since the Ethereum 2.0 speculation sprint of 2017—back then, the threat was 'scalability'; now it's 'AI replaceability.' The market is once again anchoring on the wrong variable.
Core: The Three Signals and Their Crypto Translation
Let's dissect the three data points through a crypto lens.
Signal 1: 96% changed investment approach. In the crypto world, this translates to a massive re-rating of protocol tokens and equity in DAO-backed startups. The 'AI factor' is now a mandatory input in valuation models. Projects that lack a clear AI integration narrative—or worse, those that can be fully replicated by a GPT-4 agent—are seeing their implied discounts widen. Based on my audit experience during the 2022 bear market, I saw how protocol teams scramble to add 'AI' to their whitepapers without real technical depth. The same is happening now, but with capital at stake. The money flowing out of traditional software is flowing into AI infrastructure (GPU clouds, data labeling) and away from generic SaaS. In crypto, the equivalent is capital rotating from generic DeFi protocols (like simple AMMs) to AI-co-processed chains (like those with built-in inference oracles) or data-heavy Oracle networks (like Chainlink).
Signal 2: 91% say moat = proprietary data + network effects. This is where the crypto community must resist the temptation to nod along. In a permissionless system, 'proprietary data' is an oxymoron. On-chain data is public. The only genuinely proprietary data in crypto is off-chain—like order flow from a centralized exchange, or KYC data from a regulated platform. Network effects are real (Uniswap's liquidity, Ethereum's developer community), but they are fragile. AI can spawn synthetic liquidity pools or generate code that mimics smart contract interactions. The real moat, as I argued in my 'DeFi Digest' days, is regulatory compliance and workflow embedding. A protocol that has integrated with real-world banking rails (like a tokenized RWA platform) has a moat that no AI can replicate—because it's built on legal contracts, not code. Yet the survey shows investors undervaluing this. The 91% consensus is a classic case of 'narrative capture'—everyone agrees, so the alpha is elsewhere.
Signal 3: Only 4% unchanged. This tiny minority is perhaps the most interesting. They are the contrarians—the ones who believe that software's inherent value (e.g., its ability to enforce trustless logic) is immune to AI. In crypto, this could be Bitcoin maximalists who argue that a simple, immutable ledger needs no AI enhancement. Or it could be projects that are so deeply embedded in legacy infrastructure (like a tokenized bond platform) that AI cannot replace the institutional trust. This 4% may be holding the assets that will be the safest havens in the coming AI-driven rotation.
Contrarian: The Consensus Moat Is Already Priced In—Look Elsewhere
The contrarian angle is that the 91% consensus is a lagging indicator. When everyone agrees that data and network effects are the moat, those assets are already bid up. The real alpha lies in the overlooked moats that the survey didn't capture: - Regulatory Moats: In the crypto world, being compliant with MiCA, the SEC, or local securities laws is a massive barrier to entry for AI agents that arbitrarily generate financial products. - Institutional Embedding: A protocol that handles real-world asset settlement (e.g., tokenized Treasuries) has a moat of legal agreements and banking relationships that no AI can replicate. - AI Inference Cost Structure: The cost of running AI models on-chain is still prohibitive. Projects that optimize for on-chain inference (like zk-rollups with AI coprocessors) will have a cost advantage that is not zero in the eyes of investors.
Furthermore, the risk of 'data moat erosion' is real. Synthetic data generation is improving rapidly; soon, a DeFi protocol's historical transaction data can be simulated by a GPT model, making the 'proprietary data' claim hollow. The 91% consensus may be building on sand. Artifacts of a new digital renaissance—like decentralized AI marketplaces—will emerge, but they will be built on trust, not just data.
Takeaway
The market is mispricing crypto software assets that have strong but non-obvious moats. Investors should look for projects that are 'AI-enhanced' rather than 'AI-replaceable'—those that use AI to deepen their compliance or workflow integration, not just to add a chatbot. The next narrative cycle will be about 'AI-native DeFi' vs 'AI-enhanced DeFi,' and the winners will be the ones that the 96% are too busy fleeing to notice. Following the thread from code to culture, I'm watching for the moment when the market realizes that the true moat is not the data you own, but the trust you've earned.