The AI Moat Doesn’t Exist Yet
Moats form when companies change how industries compete
Where’s the moat in AI? Foundation models? Proprietary data? The application layer? A lot is at stake in getting the answer right. What the pattern of past technologies tells us is that the moat does not exist yet.
When companies use new technologies to build moats they change how industries compete. Amazon made retail a competition over online selection, personalization, checkout, shipping and returns. Retailers before Amazon built their advantages around physical location, foot traffic, price, and supplier relationships. Tesla built a moat around batteries, vehicle software, and online ordering. Incumbent auto manufacturers struggled to compete.
With new technologies the moat must be invented. To shift frictionless buying from the store to the screen Amazon had to invent one-click checkout. None of the expertise American, Japanese, or European auto manufacturers built prepared them for cars where engines mattered less than software and batteries.
AI will be no different. Challengers will build new moats by changing how industries compete. Consider data. Many industries generate valuable proprietary data that looks like potential moats. As customer behavior changes, the data that matters will change too. Incumbent data will remain valuable to incumbent business models. But incumbent business models are not AI business models.
Mobile entertainment existed for years before TikTok showed what a mobile-native moat looked like. Before TikTok the user’s social graph drove mobile discovery. Meta with its multi-billion user social graph enjoyed a seemingly insurmountable data moat. TikTok’s algorithmic discovery created a data moat beyond the social graph. The social graph then proved inessential for mobile video discovery.
As new technologies spread they divide companies into camps. Some use the new technologies to build new moats. Everyone else adopts the technologies to catch up. Today, every major network and studio offers its customers a mobile streaming app. But none of them rely on mobile discovery and consumption for their moat, like TikTok does. AI will follow this same path. Some will use AI to build new moats. Most will use AI to catch up.
History suggests that incumbents will begin by layering AI on top of their existing business models before they adapt to AI-first competition. Entertainment studios invested in many forms of digital entertainment before eventually converging on Netflix-like streaming. For decades Walmart viewed e-commerce through the lens of its existing business before embracing an omni-channel strategy to catch up to Amazon. Even when category leaders like Netflix or Amazon change customer behavior and build new moats, incumbents struggle to adapt without doubling down on existing capabilities first.
With AI, those category leaders do not yet exist. New technologies on their own do not deliver moats. They must be married to a vision for an industry that competes on new dimensions. We’ll recognize those new visions when customers migrate to AI-first experiences beyond what today’s industries look like. In the end it’s customers, and not companies, that validate moats. With AI, the moat does not exist yet. It must be invented.