Google has hell of a moat because they are continuously crawling and indexing the Web for the last 27 years. They have tons of data and information that can be used for training of their AI.
For consumers, i think we could see local first AI boxes (possibly just our phones) + subscription based plan to integrate with set of API providers for info and actions with paid add-ons and updates. The moat might be integration or deals with a set of “atomic”/foundational API providers to invoke results (without a [headless] browser).<p>E.g. book a flight will be thru APIs from airlines partnering with consumer AI box companies with some “best pricing” layer on top so user can book optimal flights, user can generate high quality infinite doomscrolling content on demand, provided by TikTok with ad and ad-free tiers. Law enforcement will crack down on unregistered, “dark” boxes and require API providers to talk with registered boxes only so dark boxes will have to rely on shady, limited APIs.
Hot take but 99% of all profitable businesses have no moat, and there’s absolutely nothing wrong with that. You can still make lots of $$$.<p>Moats matter the most to investors and maybe the C suite. For everyone else, it’s just an intellectual exercise
The VCs thinked that the scaling law may build a moat for very large computing powers like Google and OpenAI, and VCs rushed into the area the foundation models.<p>The emerging of DeepSeek V3 and DeepSeek R1 somehow make the moat vulnerable with their much smaller training cost. The fact verified the claim of Google.
Frankly I hope this pops the AI bubble back to reality. We don't really need a 500B investment in AI. We do need a 500B investment in cleaning up our emissions. Something AI would only make worse.