Survival Strategies for AI Startups: Insights from a Google VP

The Rise and Fall of LLM Wrappers and AI Aggregators in the Generative AI Boom

The generative AI boom saw a new startup emerge almost every minute, but as the dust settles, two once-promising business models are facing challenges: LLM wrappers and AI aggregators.

Darren Mowry, head of Google’s global startup organization, warns that startups relying on these models have a “check engine light” on.

LLM wrappers are startups that wrap existing large language models like Claude, GPT, or Gemini with a product or UX layer to solve specific problems, such as AI-assisted student studying.

Mowry cautions that simply relying on the backend model without differentiation is no longer sustainable in the industry.

According to Mowry, startups need to have unique value propositions or deep moats to thrive. Examples include Cursor, a GPT-powered coding assistant, and Harvey AI, a legal AI assistant.

Startups can no longer rely on slapping a UI on top of a GPT model for success. Building sustainable product value is now the key challenge.

AI aggregators, a subset of LLM wrappers, aggregate multiple LLMs into one interface or API layer to provide users access to various models. Companies like Perplexity and OpenRouter offer such services.

Despite some success, Mowry advises incoming startups to steer clear of the aggregator business due to lack of growth and user demand for intellectual property in AI solutions.

Mowry draws parallels between the AI aggregator landscape and the early days of cloud computing, where middlemen were squeezed out as major providers expanded into enterprise services.

Looking ahead, Mowry sees potential in vibe coding and direct-to-consumer tech, with companies like Replit, Lovable, and Cursor leading the way.

He also highlights the growth potential in biotech and climate tech, driven by significant venture investments and access to vast amounts of data for creating value.