Google VP warns that two kinds of AI startups could not survive

The generative AI growth minted a startup a minute. However because the mud begins to settle, two once-hot enterprise fashions are wanting extra like cautionary tales: LLM wrappers and AI aggregators. 

Darren Mowry, who leads Google’s world startup group throughout Cloud, DeepMind, and Alphabet, says startups with these hooks have their “verify engine mild” on.

LLM wrappers are basically startups that wrap current giant language fashions, like Claude, GPT, or Gemini, with a product or UX layer to unravel a selected drawback. An instance can be a startup that uses AI to helps students study.

“If you happen to’re actually simply relying on the again finish mannequin to do all of the work and also you’re nearly white-labeling that mannequin, the trade doesn’t have quite a lot of endurance for that anymore,” Mowry stated on this week’s episode of Fairness. 

Wrapping “very skinny mental property wrapped round Gemini or GPT-5” indicators you’re not differentiating your self, Mowry says. 

“You’ve obtained to have deep, extensive moats which can be both horizontally differentiated or one thing actually particular to a vertical market” for a startup to “progress and develop,” he stated. Examples of the deep moat LLM wrapper sort embody Cursor, a GPT-powered coding assistant, or Harvey AI, a authorized AI assistant.

In different phrases, startups can not count on to slap a UI on prime of a GPT and get traction on their product, like they may, maybe, in mid-2024 when OpenAI launched its ChatGPT retailer. The problem now could be to construct sustainable product worth. 

AI aggregators are a subset of wrappers — they’re startups that combination a number of LLMs into one interface or API layer to route queries throughout fashions and provides customers entry to a number of fashions. These corporations sometimes present an orchestration layer that features monitoring, governance, or eval tooling. Assume: AI search startup Perplexity or developer platform OpenRouter, which gives entry to a number of AI fashions through a single API. 

Whereas many of those platforms have gained floor, Mowry’s phrases are clear to incoming startups: “Keep out of the aggregator enterprise.”

Usually talking, aggregators aren’t seeing a lot development or development as of late as a result of, he says, customers need “some mental property inbuilt” to make sure they’re routed to the fitting mannequin on the proper time based mostly on their wants — not due to behind-the-scenes compute or entry constraints.

Mowry has been within the cloud sport for many years, chopping his tooth at AWS and Microsoft earlier than establishing store at Google Cloud, and he’s seen how this performs out. He stated the scenario right this moment mirrors the early days of cloud computing within the late 2000s/early 2010s as Amazon’s cloud enterprise began taking off.

At the moment, a crop of startups sprang as much as resell AWS infrastructure, advertising and marketing themselves as simpler entry factors that offered tooling, billing consolidation, and assist. However when Amazon constructed its personal enterprise instruments and prospects realized to handle cloud companies straight, most of these startups have been squeezed out. The one survivors have been those who added actual companies, like safety, migration, or DevOps consulting. 

AI aggregators right this moment face comparable margin strain as mannequin suppliers increase into enterprise options themselves, probably sidelining middlemen. 

For his half, Mowry is bullish on vibe coding and developer platforms, which had a record-breaking yr in 2025 with startups like Replit, Lovable, and Cursor (all Google Cloud prospects, per Mowry) attracting main funding and buyer traction.

Mowry additionally expects sturdy development in direct-to-consumer tech, in corporations that put a few of these highly effective AI instruments into the fingers of consumers. He pointed to the chance for movie and TV college students to make use of Google’s AI video generator Veo to convey tales to life.

Past AI, Mowry additionally thinks biotech and local weather tech are having a second — each when it comes to enterprise funding going into the 2 industries and the “unimaginable quantities of information” startups can entry to create actual worth “in methods we’d by no means have been capable of earlier than.”