Google VP warns that two types of AI startups may not survive

Google Vice President Darren Mowry warned that two categories of artificial intelligence startups face uncertain prospects for long-term survival. Mowry, who leads Google's global startup organization across Cloud, DeepMind, and Alphabet, made the comments on this week's episode of Equity. He said startups built as LLM wrappers or AI aggregators now show warning signs. LLM wrappers add a product or user experience layer around existing large language models such as Claude, GPT, or Gemini to address a specific need. Mowry said that simply placing thin intellectual property around models like Gemini or GPT-5 fails to create meaningful differentiation. Startups must develop deep, wide moats that are either horizontally differentiated or specific to a vertical market, he said. Cursor, a coding assistant, and Harvey AI, a legal assistant, represent the type of wrappers that meet this standard. AI aggregators form a subset of wrappers. These companies combine multiple large language models into a single interface or application programming interface layer and add orchestration features such as monitoring or evaluation tooling. Perplexity and OpenRouter illustrate the approach. Mowry advised new startups to stay out of the aggregator business. Users now seek intellectual property built into the product to ensure routing to the appropriate model, rather than relying on compute or access constraints managed behind the scenes, Mowry said. He compared the current situation to the early period of cloud computing. At that time, many startups that resold Amazon Web Services infrastructure were displaced once Amazon introduced its own enterprise tools, unless the resellers added distinct services such as security or migration support. AI aggregators encounter similar margin pressure as model providers develop their own enterprise capabilities, Mowry said.