Hook
Ninety-one percent. That’s the proportion of institutional investors in a recent Lazard survey who now identify “proprietary data + network effects” as the primary moat for software companies. Only 4% have not changed their investment approach. The message is stark: the old playbook of valuing software on MRR multiples and growth rates is dead. But what does this mean for crypto? In a market where the line between “software” and “protocol” is increasingly blurred, this survey is a canary in the coal mine for how we evaluate blockchain projects in the age of AI.
Context
Lazard’s deep dive into private equity secondaries reveals a consensus that has formed with unusual speed. Investors are no longer debating whether AI will disrupt software—they are now asking how and which companies will survive. The same forces are now sweeping through crypto. As a founder of a crypto education platform based in Shenzhen, I’ve watched this shift unfold firsthand. In 2017, during the ICO boom, I translated Tezos’s governance whitepaper for 50,000 readers, believing that self-amending code was the future. By 2022, after FTX collapsed, I spent six months auditing Polygon ID to understand how true sovereignty could be built. The lesson from both eras is the same: the market’s attention is now fixed on structural moats, not speculative narratives.
Core Analysis: The AI-Crypto Crossroads
Lazard’s 91% consensus on “data + network effects” is not just a software signal—it’s a crypto signal. Let me unpack why.
1. The Data Moat in Crypto
In traditional software, proprietary data means user behavior logs, transaction records, or industry-specific datasets. In crypto, it’s on-chain data, MEV patterns, and liquidity profiles. Projects that control unique, high-quality on-chain data—like Dune Analytics, which aggregates query patterns, or Flashbots, which captures MEV flow—are building moats that are hard for general-purpose AI models to replicate. The key insight from the Lazard report is that LLMs excel at public patterns but struggle with private distributions. For crypto, this means that protocols with exclusive access to order flow, validator networks, or cross-chain activity data will command premium valuations. Code over hype.
2. Network Effects as a Dynamic Shield
The survey emphasizes that network effects complement AI rather than being replaced by it. In crypto, this is even more pronounced. A decentralized exchange with deep liquidity benefits from a data flywheel: more traders → more liquidity data → better AI-driven routing → more traders. Uniswap’s v4 hooks, for example, allow custom liquidity strategies that leverage AI to optimize pool compositions. The Lazard data suggests investors will pay a premium for projects that have both network effects and AI integration. However, the report also warns that “AI augmentation” alone is not enough—the network must be structurally defensible. Hold the line.
3. The Valuation Paradigm Shift
Lazard found that only 4% of investors have not changed their methods. This is a flashing red light for crypto analysts who still use traditional metrics like TVL growth or transaction count. The new framework must incorporate an “AI exposure discount” and a “moat quality premium.” For example, a Layer 2 that relies solely on public data for its sequencer (like base EVM) will face a higher discount than one that controls proprietary user data (like zkSync’s zkEVM with private mempool). I’ve seen this shift in my own work: when I audited Polygon ID’s zero-knowledge proofs, I realized that the quality of the identity data—not just the technology—determined the project’s long-term value. Truth decays slowly.
Contrarian Angle: The Crypto Blind Spot
The Lazard survey’s consensus is powerful, but it has a blind spot when applied to crypto. The 91% focus on “data + network effects” implicitly assumes that the underlying model layer will become a commodity. In crypto, the equivalent assumption is that base-layer consensus will become a commodity (e.g., Ethereum as the “world computer”). But what if the model layer itself is decentralized? Projects like Fetch.ai or Bittensor are building decentralized AI networks where the model’s value is not just in data but in the distributed compute and incentive alignment. In such a world, the “moat” shifts from data to the governance of the AI itself. The Lazard survey may underestimate the potential for crypto-native AI to create a new category of moats that are more about robustness than exclusivity.
Another blind spot: the survey’s investors are predominantly from traditional PE secondaries, not crypto-native funds. They may not fully understand that crypto’s “network effects” are often permissionless and composable, which can both amplify and erode moats. For example, a DeFi protocol’s liquidity can be forked overnight, while a software company’s customer base is harder to replicate. The “wait-and-see” approach in PE may be less applicable to crypto, where the pace of innovation is faster and the window for alpha is narrower.
Takeaway: Build Anyway
The Lazard survey is a Rosetta Stone for the next phase of crypto investing. It tells us that the market is coalescing around a new valuation grammar: data moats, network effects, and AI integration. But it also warns that the transition period is a “valuation vacuum” where old metrics fail and new ones are not yet standard. For founders, this means doubling down on proprietary data and network density. For investors, it means developing a framework to quantify AI exposure—fast. The window for mispricing is closing. Build anyway.