The generative AI boom created a startup a minute. But as the dust begins to settle, two once-hot business models are looking more like cautionary tales: LLM wrappers and AI aggregators. Darren Mowry, who leads Google’s global startup organization, says startups built on these models have their “check engine light” on.
LLM wrappers are startups that wrap existing large language models, like Claude or GPT, with a product or user experience layer to solve a specific problem. An example would be a startup using AI to help students study. According to Mowry, the industry has lost patience for business models that simply white-label a backend model. He states that wrapping very thin intellectual property around a model like Gemini signals a lack of differentiation.
For a startup to progress and grow, Mowry argues it must build deep, wide moats. These moats should be either horizontally differentiated or highly specific to a vertical market. Examples of LLM wrappers with such substantial moats include Cursor, a GPT-powered coding assistant, and Harvey AI, a legal AI assistant. The challenge now is to build sustainable product value, as startups can no longer expect to gain traction by simply placing a user interface on top of a large language model.
AI aggregators are a subset of wrappers. These startups aggregate multiple LLMs into a single interface or API layer, routing user queries across different models. They typically provide an orchestration layer that includes monitoring and governance tooling. Think of AI search startup Perplexity or developer platform OpenRouter.
Despite some platforms gaining ground, Mowry advises incoming startups to stay out of the aggregator business. He explains that aggregators are not seeing much growth because users want intellectual property built into the routing decisions. They want to be sent to the right model based on their needs, not due to behind-the-scenes compute constraints.
Mowry, a veteran of the cloud industry, sees a parallel to the early days of cloud computing. Then, many startups emerged to resell AWS infrastructure, offering easier entry points with consolidated billing and support. When Amazon built its own enterprise tools and customers learned to manage services directly, most of those resellers were squeezed out. Only those adding real services, like security or consulting, survived. AI aggregators now face similar margin pressure as model providers expand their own enterprise features.
Looking beyond these cautionary tales, Mowry is bullish on other areas. He highlights developer platforms and vibe coding, which had a record-breaking year. He also expects strong growth in direct-to-consumer tech that puts powerful AI tools into customers’ hands, such as film students using AI video generators to bring stories to life.
Beyond AI, Mowry believes biotech and climate tech are having a moment. He points to significant venture investment flowing into these industries and the incredible amounts of data startups can now access to create real value in unprecedented ways.