The generative AI boom has spawned a flurry of startups. But as the dust begins to settle, two once-popular business models are looking more alarming: 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 a “check engine light” on.
LLM wrappers are basically startups that wrap existing large language models like Claude, GPT, Gemini, etc. in a product or UX layer to solve a specific problem. One example is a startup that uses AI to help students learn.
“If you’re just letting your backend model do all the work, and you’re pretty much white-labeling that model, the industry doesn’t have a lot of patience for that anymore,” Morley said on this week’s episode of Equity.
Wrapping “very thin intellectual property wrapped around Gemini and GPT-5” shows a lack of differentiation, Morley says.
For startups to “progress and grow,” he said, “they need deep and wide moats, either horizontally differentiated or vertically market-specific.” Examples of deep LLM wrapper types include Cursor, a GPT-powered coding assistant, and Harvey AI, a legal AI assistant.
tech crunch event
boston, massachusetts
|
June 9, 2026
In other words, startups can no longer expect their products to attract attention by layering a UI on top of GPT, as they did in mid-2024 when OpenAI launched its ChatGPT store. The current challenge is to build sustainable product value.
AI aggregators are a subset of wrappers and startups that aggregate multiple LLMs into one interface or API layer to route queries between models and provide users with access to multiple models. These companies typically provide an orchestration layer that includes monitoring, governance, and evaluation tools. Consider AI search startup Perplexity or OpenRouter, a developer platform that provides access to multiple AI models through a single API.
While many of these platforms are well established, Mowry’s words are clear for up-and-coming startups. “Stay out of the aggregator business.”
Generally speaking, he says there hasn’t been much growth or advancement in aggregators recently, not because of background computing or access constraints, but because users want “embedded intellectual property” to ensure they’re routed to the right model at the right time based on their needs.
Mowry has been in cloud gaming for decades, cutting his teeth at AWS and Microsoft before starting with Google Cloud and seeing how this plays out. He said the current situation reflects the early days of cloud computing in the late 2000s and early 2010s, when Amazon’s cloud business was taking off.
At the time, a slew of startups emerged to resell AWS infrastructure, pitching themselves as an easy entry point that offered tools, billing integration, and support. But as Amazon built its own enterprise tools and allowed customers to manage cloud services directly, most of those startups were wiped out. The only ones that survived were those that added real services like security, migration, and DevOps consulting.
AI aggregators now face similar margin pressures as model providers themselves expand into enterprise capabilities and potentially sideline intermediaries.
Mowry is bullish on Vibe coding and developer platforms, which had a record year in 2025 with startups like Replit, Lovable, and Cursor (all Google Cloud customers, according to Mowry) attracting significant investment and customer traction.
Mowry also expects strong growth in direct-to-consumer technology, with companies putting some of these powerful AI tools in the hands of customers. He noted that film and television students have the opportunity to bring their stories to life using Google’s AI video generator Veo.
Beyond AI, Mowry also sees opportunities in biotechnology and climate change technologies. Both in terms of venture investments in these two industries and the “incredible amount of data” that startups can access to create real value “in ways that weren’t possible before.”
