Google Exec: AI Wrappers Face 'Check Engine Light'

The generative AI landscape is shifting, with business models centered on AI wrappers and aggregators facing scrutiny. Darren Mowry, who leads Google's global startup organization, indicates that startups merely wrapping existing large language models (LLMs) like Claude or GPT are signaling risk.

These 'thin intellectual property' wrappers are no longer sufficient for growth. Mowry emphasizes the need for startups to possess 'deep, wide moats,' either through horizontal differentiation or specialized vertical market solutions. Successful examples include Cursor, a coding assistant, and Harvey AI, a legal assistant.

AI aggregators, which combine multiple LLMs, are also advised to be avoided by new startups. Mowry suggests users now seek built-in intellectual property to ensure optimal model routing, rather than relying on aggregators that manage access or compute constraints.

This situation echoes the early days of cloud computing. Startups that resold AWS infrastructure ultimately faded when AWS developed its own tools and customers became self-sufficient. Similarly, AI aggregators face margin pressure as LLM providers enhance their enterprise features.

Mowry is optimistic about vibe coding and developer platforms, noting significant investment in companies like Replit and Cursor. He also anticipates growth in direct-to-consumer tech, citing Google's Veo AI video generator as an example for creative applications. Beyond AI, Mowry identifies biotech and climate tech as burgeoning sectors with substantial data opportunities.