The surge in generative AI has led to the creation of a startup every minute. However, as the situation stabilizes, two once-popular business models are appearing more like warnings: LLM wrappers and AI aggregators. 

Darren Mowry, who heads Google’s global startup unit encompassing Cloud, DeepMind, and Alphabet, mentions that startups employing these strategies have their “check engine light” activated.

LLM wrappers are essentially startups that encase established large language models, such as Claude, GPT, or Gemini, in a product or user experience layer to address a specific issue. An instance would be a startup that harnesses AI to assist students in their study efforts.

“If you’re merely relying on the backend model to do all the heavy lifting and you’re effectively white-labeling that model, the industry has lost patience for that approach,” Mowry commented on this week’s Equity episode. 

Wrapping “very minimal intellectual property around Gemini or GPT-5” indicates a lack of differentiation, according to Mowry. 

“You need to establish substantial, wide moats that are either horizontally distinctive or particularly tailored to a specific vertical market” for a startup to “advance and thrive,” he explained. Examples of the deep-moat LLM wrapper type encompass Cursor, a GPT-driven coding assistant, or Harvey AI, a legal AI assistant.

In essence, startups can no longer merely add a UI on top of a GPT and expect to gain traction on their product, unlike the situation in mid-2024 when OpenAI unveiled its ChatGPT store. The current challenge is to create lasting product value. 

AI aggregators represent a category of wrappers — they are startups that compile various LLMs into a single interface or API layer to direct queries across models and provide users access to multiple models. These companies typically offer an orchestration layer that includes monitoring, governance, or evaluation tools. Consider AI search startup Perplexity or developer platform OpenRouter, which grants access to multiple AI models through a unified API. 

Although numerous platforms have made strides, Mowry’s guidance to emerging startups is clear: “Avoid the aggregator business.”

Generally, aggregators are currently not experiencing significant growth or development, as he states, users desire “some form of intellectual property incorporated” to guarantee they are directed to the appropriate model at the right moment based on their requirements — not due to behind-the-scenes computing or access limitations.

Mowry has extensive experience in the cloud sector, having begun his career at AWS and Microsoft before establishing a presence at Google Cloud, and he has observed how this scenario unfolds. He mentioned that today’s environment mirrors the early phases of cloud computing in the late 2000s and early 2010s when Amazon’s cloud division began to flourish.

At that time, numerous startups emerged to resell AWS infrastructure, branding themselves as simpler solutions that offered tools, billing consolidation, and support. However, when Amazon developed its own enterprise tools and customers adapted to managing cloud services directly, most of those startups were eliminated. Only those who provided genuine services, such as security, migration, or DevOps consulting, managed to survive. 

AI aggregators today are under similar margin pressure as model providers introduce enterprise features themselves, potentially marginalizing intermediaries. 

Mowry, for his part, is optimistic about vibe coding and developer platforms, which experienced a record-setting year in 2025 with startups like Replit, Lovable, and Cursor (all customers of Google Cloud, according to Mowry) drawing substantial investment and customer engagement.

Mowry also anticipates robust growth in direct-to-consumer technology, in companies that deliver some of these powerful AI tools directly to consumers. He highlighted the potentials for film and TV students to utilize Google’s AI video generator Veo to bring their stories to fruition.

Beyond AI, Mowry perceives that biotech and climate tech are currently significant — both in terms of venture capital inflow into these sectors and the “immense amounts of data” accessible to startups to create real value “in ways we couldn’t have done before.”