Over the past six months, I have tracked a quiet but forceful shift in the private equity secondary markets. A new survey by Lazard—one of the most respected investment banks—reveals that 96% of investors have already altered their approach to software investing because of AI. Nearly 91% now define a software company’s moat as its proprietary data combined with network effects. And most tellingly, capital is actively moving away from legacy software assets into other opportunities.
These are not abstract signals. They are the first recorded pricing data points of a paradigm shift that is now cascading into crypto-native software. As a researcher who built my career on forensic analysis of crypto protocols—from the 2017 ICO audits to the 2022 Terra collapse hedging—I recognize the pattern. The same forces that are compressing valuations in traditional SaaS are now silently reshaping how tokens, DAOs, and decentralized applications are priced.
Context: The Survey and Its Crypto Parallel
Lazard’s survey targeted secondary market investors in private equity—institutions that buy and sell stakes in private software companies. The key finding: 96% of respondents have changed their investment behavior due to AI. They are not just tweaking models; they are reallocating capital. The 91% consensus on “data + network effects” as the primary moat reveals that the market has already priced in the commoditization of AI model capabilities.
In crypto, the secondary market is less formal but equally revealing. I have seen a similar pattern in OTC token desks and private sale allocations. Protocols that lack a defensible data moat—think simple DEX forks or generic lending platforms—are trading at discounts of 20-40% relative to their total value locked (TVL) adjusted peers. Meanwhile, protocols with unique data pipelines, such as on-chain analytics aggregators or privacy-preserving computation layers, are commanding premiums.
Core: The Three Signals and Their Implications for Crypto Software
Signal 1: 96% behavior change — AI is a mandatory variable, not optional.
In crypto, this means that any protocol whose software layer is vulnerable to AI-driven replacement will see its token valuation compress. Consider the rise of AI-native agents that can execute trades, manage liquidity, or even compose DeFi transactions. Traditional smart contract-based protocols that rely on manual user interaction are at risk. The 96% figure tells me that the window for legacy crypto software to adapt is closing. I have already flagged this in my own reports: the protocols that are integrating AI into their core logic—like AI-optimized MEV searchers or autonomous risk managers—are the ones that will survive the next cycle.
Signal 2: 91% focus on data + network effects — the moat is shifting on-chain.
In traditional software, proprietary data means user behavior, migration logs, or integration configurations. In crypto, the data is often public. Yet the same principle applies: the value lies in the unique curation, labeling, or preprocessing of that data. For example, a DEX aggregator that captures order flow data across multiple chains has a data moat that a simple swap interface cannot replicate. The network effect is even more pronounced in crypto: a protocol with a large, sticky user base creates a liquidity moat that AI cannot easily decompose. However, the 91% consensus is a double-edged sword. If everyone is already pricing in this moat, the excess returns from betting on it are gone. The real alpha will come from identifying moats that are not yet on the consensus radar.
Signal 3: Capital moving to other opportunities — the flight to non-software assets.
In the Lazard survey, investors are shifting capital away from software entirely. In crypto, I see a parallel: capital is flowing from application-layer tokens to infrastructure and AI-focused chains. The recent premium on L1 tokens like Solana or Monad, which are optimized for AI computation, mirrors this trend. The capital that fled software in traditional markets is now entering crypto infrastructure that powers AI. This is a secular shift, not a cyclical one.
Contrarian: The Consensus Is Overpriced — The Real Moat Is Workflow Embedding and Compliance
Here is the counter-intuitive angle that the Lazard survey misses. The 91% consensus that “data + network effects” is the moat is already priced into the market. In crypto, I have seen this pattern before. During the 2020 DeFi summer, everyone believed that TVL was the moat. I published a spreadsheet analysis showing that liquidity mining APY was a subsidy for TVL numbers, and that real users vanished when incentives stopped. The same logic applies here. Data moats in crypto are fragile because synthetic data and zero-knowledge proofs can replicate or anonymize user behavior. The true moat that endures is workflow embedding—how deeply a protocol is integrated into the operational fabric of its users. For example, a cross-chain bridge that is embedded in the routing logic of multiple aggregators has a sticky advantage that AI cannot replace overnight. Similarly, regulatory compliance—such as having a licensed custodial wrapper or a KYC-friendly on-ramp—creates a barrier that AI-native competitors cannot easily cross.
Another blind spot: the Lazard survey does not account for the rise of AI-native crypto software. These are protocols built from day one with AI as the core engine, not as an add-on. They can leverage AI to optimize gas fees, predict congestion, or automate risk management. The 96% of investors who changed their behavior are likely underestimating the speed at which these AI-native crypto protocols will displace legacy software. I have seen early signs in our own research: over the past twelve months, the trading volume of AI-driven MEV strategies has grown from negligible to 18% of total Ethereum blockspace. This is not a niche; it is a structural shift.
Takeaway: Positioning for the Cycle
The Lazard survey is a clear signal that the macro environment is forcing a re-evaluation of software assets. For crypto, this means the next six to twelve months will be a window of opportunity for those who can identify protocols with genuine, hard-to-replicate moats—workflow embedding, compliance, or AI-native design. The safe play is to avoid generic crypto software that relies on feature-based value. Instead, focus on protocols that have a defensible data pipeline, a sticky network effect, and a clear path to AI integration. The capital flight from legacy software will eventually reach crypto, but it will also create mispricings that the disciplined analyst can exploit.
safe. The 96% figure is a warning, but it is also a compass. The protocols that survive will be the ones that treat AI not as a threat, but as the next layer of their own architecture. The rest will be commoditized.
