The generative AI boom minted a startup a minute. But as the dust starts 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 across Cloud, DeepMind, and Alphabet, says startups with these hooks have their “check engine light” on.

LLM wrappers are essentially startups that wrap existing large language models, like Claude, GPT, or Gemini, with a product or UX layer to solve a specific problem. An example would be a startup that uses AI to help students study.

“If you’re really just counting on the back-end model to do all the work and you’re almost white-labeling that model, the industry doesn’t have a lot of patience for that anymore,” Mowry said on this week’s episode of Equity. 

Wrapping “very thin intellectual property around Gemini or GPT-5” signals you’re not differentiating yourself, Mowry says. 

“You’ve got to have deep, wide moats that are either horizontally differentiated or something really specific to a vertical market” for a startup to “progress and grow,” he said. Examples of the deep-moat LLM wrapper type include Cursor, a GPT-powered coding assistant, or Harvey AI, a legal AI assistant.

In other words, startups can no longer expect to slap a UI on top of a GPT and get traction on their product like they could, perhaps, in mid-2024 when OpenAI launched its ChatGPT store. The challenge now is to build sustainable product value. 

AI aggregators are a subset of wrappers — they’re startups that aggregate multiple LLMs into one interface or API layer to route queries across models and give users access to multiple models. These companies typically provide an orchestration layer that includes monitoring, governance, or eval tooling. Think: AI search startup Perplexity or developer platform OpenRouter, which provides access to multiple AI models via a single API. 

While many of these platforms have gained ground, Mowry’s message is clear to incoming startups: “Stay out of the aggregator business.”

Generally speaking, aggregators aren’t seeing much growth or progression these days because, he says, users want “some intellectual property built in” to ensure they’re routed to the right model at the right time based on their needs — not because of behind-the-scenes compute or access constraints.

Mowry has been in the cloud game for decades, cutting his teeth at AWS and Microsoft before setting up shop at Google Cloud, and he’s seen how this plays out. He said the situation today mirrors the early days of cloud computing in the late 2000s/early 2010s as Amazon’s cloud business started taking off.

At that time, a crop of startups sprang up to resell AWS infrastructure, marketing themselves as easier entry points that provided tooling, billing consolidation, and support. But when Amazon built its own enterprise tools and customers learned to manage cloud services directly, most of those startups were squeezed out. The only survivors were the ones that added real services, like security, migration, or DevOps consulting. 

AI aggregators today face similar margin pressure as model providers expand into enterprise features themselves, potentially sidelining middlemen. 

For his part, Mowry is bullish on vibe coding and developer platforms, which had a record-breaking year in 2025 with startups like Replit, Lovable, and Cursor (all Google Cloud customers, per Mowry) attracting major investment and customer traction.

Mowry also expects strong growth in direct-to-consumer tech, in companies that put some of these powerful AI tools into the hands of customers. He pointed to the opportunity for film and TV students to use Google’s AI video generator Veo to bring stories to life.

Beyond AI, Mowry also thinks biotech and climate tech are having a moment — both in terms of venture investment going into the two industries and the “incredible amounts of data” startups can access to create real value “in ways we would never have been able to before.”

View Bio