Only 2 Competitive Moats Hold Up Against AI, Data Shows

Quick Facts

• A Mighty Capital analysis of 576 venture-backed AI B2B companies that raised $50M+ since early 2025 found only two durable competitive moats in the AI era.
• Companies with counter-positioning — just 5% of the dataset — command a median enterprise value of 5.3x per dollar raised.
• Technology features and scale economics are identified as false moats, with model-based advantages proving fragile as foundation models converge in capability.

SC Moatti, founding managing partner of Mighty Capital, has a blunt message for founders: if your pitch leads with “we use AI,” you are describing infrastructure, not a business.

Moatti’s firm applied Hamilton Helmer’s 7 Powers framework to Crunchbase data covering 576 venture-backed AI B2B companies that raised rounds of $50 million or more since the start of 2025. The conclusion: AI-native startups can credibly claim only two of the seven powers as genuine moats.

The Two Moats That Hold

The first is counter-positioning. This occurs when a startup builds a business model so structurally different that an incumbent cannot copy it without damaging its own economics.

The clearest current example is pricing. New AI companies charge per task or per value delivered. Legacy SaaS companies are locked into per-seat pricing. If their AI works well, it reduces the number of seats needed and cannibalizes their own revenue. The incumbent’s success becomes the obstacle to competing.

Only 5% of companies in the dataset use counter-positioning, yet those companies attract a median enterprise value of 5.3x per dollar raised. That scarcity is precisely what investors are paying for.

The second durable moat is network economies. True network effects mean a product becomes more valuable as more people use it, and that improvement compounds into a financial advantage. In 2025, the most powerful version of this is the data loop.

A vertical AI company serving hospitals gets smarter with each new hospital added. More data improves the model. A better model attracts more hospitals. That cycle is difficult to replicate from a standing start, regardless of capital or model quality.

Moatti’s firm draws a hard line between this and simple user growth. A SaaS product used by 100 companies does not automatically improve for the 101st customer. A marketplace with thin liquidity is a distribution channel, not a network effect. Many founders conflate the two in pitch decks.

The Two Traps

Scale economics and technology features are labeled false moats in the analysis. On scale, the argument is direct: believing your unit economics improve with growth is not the same as building a structural advantage. The gap between the assumption and the reality is measured in billions of dollars most startups will never raise.

On features, the case is starker. A small team can now replicate a specific capability faster than at any prior point in software history. Startups that built their competitive position on a particular foundation model’s output in 2024 and 2025 have already seen that position erode as foundation models converge in capability.

An API price drop or a new model release can eliminate an advantage built on model quality overnight. That is not a moat. It is a timing advantage with an uncertain expiration date.

The Question Founders Cannot Avoid

Moatti frames the test for any competitive position in one question: what about your business would survive a competitor who starts today with more capital and a better model?

The analysis draws on Moatti’s Products That Count network of more than 600,000 product leaders and her prior work building products at Meta and Siebel Systems. Mighty Capital’s portfolio includes Amplitude and Netskope.

The firms commanding 4x to 5x multiples, Moatti argues, have built something structural underneath the AI layer — in the business model design or the network architecture — that a model cannot generate on its own. Founders who cannot answer the durability question in one sentence are building a product. The ones who can are building a power.

Read more: The Only 2 Moats That Actually Work In The AI Era