Gas fees don't lie. But investor sentiment does. A recent survey by investment bank Lazard of private equity secondary market participants reveals a stark truth: 96% of investors have already changed how they approach software investments. The trigger? Artificial intelligence. The crypto industry, built on software that promises decentralization and data sovereignty, should pay close attention. Because the same forces that are reshaping traditional SaaS valuations are now grinding against the blockchains we trade on.
Context: The Survey and Its Signal
Lazard's survey, conducted in H1 2025, polled institutional investors active in the secondary market for private equity stakes. The headline finding is a 96% behavioral shift—meaning nearly all respondents have altered their investment strategy for software companies due to AI. A further 91% identified “proprietary data advantages and network effects” as the core moat for software firms in an AI-dominated world. Only 4% said their approach remained unchanged.
Let’s cut through the banker-speak. This is a capital reallocation event. Money is moving from “software” as a generic asset class toward specific data-rich, network-secured platforms. The rest—feature-driven products, thin integration layers, even well-designed code—are being discounted. For crypto, this is a mirror. The software layers of blockchain—wallets, DeFi front-ends, analytics dashboards, NFT marketplaces—are all facing the same AI-driven valuation scrutiny. But the crypto industry has a unique variable: on-chain data is public, transparent, and replicable. That changes the game.
Core: The Systematic Teardown of Crypto Software Moats
Code is truth. Intent is fiction. And the ledger keeps score. But what happens when AI can write the code, read the ledger, and replicate the intent?
Let’s break down the three moats that the Lazard survey highlights—data, network effects, and feature functionality—and see how they hold in a crypto context.
1. Data Advantages: The False Idol
Traditional software companies build moats by accumulating proprietary user data—behavioral patterns, transaction histories, proprietary databases. In crypto, all transaction data is public. Every swap, every mint, every liquidation is permanently recorded on a public ledger. This is beautiful for transparency. It is terrible for creating a proprietary data moat. An AI model can scrape the entire Ethereum blockchain, train on every DeFi interaction, and generate insights that rival any single protocol’s internal analytics. The “data advantage” that 91% of investors rely on? In crypto, it’s often an illusion. Projects that claim a unique data moat are either building on top of aggregated public data or relying on off-chain data that is itself easily sourced. The result: the AI-driven commodity pressure on software features is even more acute in crypto because the raw material is freely available.
2. Network Effects: Stronger, but Not Unbreakable
Network effects in crypto are real. Metcalfe’s law applies. A larger user base of a DEX generates deeper liquidity, which attracts more traders, which generates more fees. But AI-native competitors can attack from the outside. Consider a new AI-driven aggregator that uses reinforcement learning to find optimal routing across multiple DEXs. It doesn’t need its own liquidity; it rides on top of existing networks. The network effect of the underlying DEX becomes a commodity. The aggregator captures the value. This is the same pattern Lazard identifies: traditional software companies with strong network effects (like Salesforce) are vulnerable to AI overlays that bypass the interface. In crypto, the same dynamic applies to wallets, interfaces, and even smart contract platforms. The network effect of the base layer is valuable, but the application layer is increasingly contestable.
3. Feature Functionality: The Easiest to Replicate
Minted nothing, promised everything. Many crypto projects built their entire value proposition on a clever feature—a new bonding curve, a novel voting mechanism, a unique NFT metadata standard. In the AI era, generating Solidity code that implements a bonding curve takes seconds. I’ve seen it. During my audit days in 2020, I spent hours manually reviewing reentrancy vulnerabilities. Now, an AI assistant can generate a secure contract in minutes. The feature is no longer a moat. It’s a table stake. The premium goes to the project that can execute, distribute, and build a community—not just code. The Lazard survey’s 96% behavioral shift reflects this: investors are no longer paying for feature innovation; they are paying for data defensibility and distribution. In crypto, most projects have neither.
Contrarian: What the Bulls Got Right
Before I descend into full cynicism, the contrarian angle: the Lazard survey might be overstating the threat for crypto. The 91% consensus on data moats could be a herd mentality, and herds often misprice uncertainty.
First, on-chain data is not entirely fungible. While transactions are public, the context—the user identity, the off-chain metadata, the governance patterns—is often proprietary. Projects that build strong user relationships (e.g., through custodial wallets, KYC, or social graphs) can create a semi-proprietary data layer that AI cannot easily replicate. The “data moat” is not dead; it’s just shifted from raw data to curated data with context.
Second, network effects in crypto are reinforced by token incentives. Traditional software relies on switching costs and habit. Crypto adds financial stake. A user who has locked liquidity in a Curve pool or staked tokens in a governance protocol faces a real cost to switch. AI can’t break that. It can only optimize within the existing network. The AI threat is real, but it operates at the application layer, not the base layer. Protocols with strong token-based network effects may be more resilient than the traditional SaaS companies the survey worries about.
Third, the survey’s 96% figure might be a lagging indicator. Investors have already adjusted their behavior, but the market has not yet fully priced in the new reality. This creates opportunity for those who can identify projects with genuine, sustainable moats—projects that combine on-chain data curation, token-incentivized network effects, and AI-native features. The herd is selling the entire software category. The contrarian buy is the projects that are actually AI-resistant, not AI-addled.
Takeaway: The Ledger Keeps Score
The Lazard survey is a cold dose of reality for any software investor, including those in crypto. The 96% figure is not a warning—it’s an obituary for the old valuation framework. Feature-based software is becoming a commodity. Data-based moats are being eroded by public ledgers. Network effects are still strong but are being bypassed by AI overlays.
But the ledger keeps score. The projects that will survive are those that can prove their value empirically: through on-chain activity, through user retention, through demonstrable data curation that cannot be replicated by a generic AI model. The market will reward those who can show that their code is not just beautiful but uniquely defensible. As for the rest—minted nothing, promised everything—the AI era will ruthlessly expose them.
Gas fees don’t lie. But investor sentiment does. The ledger keeps score. And the score says: adapt or discount.