Google VP warns that two types of AI startups may not survive

The rapid expansion of generative AI has led to a surge of startups, but as the market stabilizes, two business models are raising red flags: LLM wrappers and AI aggregators. Darren Mowry, who oversees Google's global startup initiatives spanning Cloud, DeepMind, and Alphabet, warns that these types of startups may be in precarious positions. LLM wrappers, which are essentially startups that utilize existing large language models such as GPT or Claude while adding a user experience layer, are facing particularly tough scrutiny. For instance, a startup that employs AI to assist students in studying would fall into this category. Mowry emphasizes that relying solely on the backend model without significant differentiation is no longer viable. “If you're merely white-labeling a model, the industry has little patience for that anymore,” he stated during a recent episode of the Equity podcast. The need for startups to carve out unique value propositions is more crucial than ever. Mowry illustrates this point by highlighting examples of successful deep-moat LLM wrappers like Cursor, a coding assistant powered by GPT, and Harvey AI, which serves the legal sector. He asserts that simply layering a user interface over a model like GPT-5 won’t suffice to gain traction in today’s competitive landscape. Similarly, AI aggregators, which combine multiple LLMs into a single interface or API, are facing challenges. These platforms are designed to route queries across different models and often include features like monitoring and governance. However, Mowry advises new entrants to steer clear of the aggregator space. "Users today are looking for integrated intellectual property that ensures they are guided to the most suitable model based on their specific requirements," he said. Drawing parallels to the cloud computing landscape of a decade ago, Mowry noted that many startups aimed to resell AWS infrastructure but ultimately failed when Amazon expanded its enterprise services. The survivors were those who offered genuine added value, such as security and consulting. The current landscape for AI aggregators mirrors this earlier phase, as model providers enhance their own enterprise features, often sidelining intermediaries. On a positive note, Mowry remains optimistic about the future of vibe coding and developer platforms, highlighting the success of startups like Replit and Cursor, which have attracted significant investment. In addition to AI, Mowry sees promising opportunities in biotech and climate tech, as increasing venture capital flows into these sectors, supported by the wealth of data available to create substantial value in innovative ways.