Google VP Warns AI Startups
A Google senior executive warned that not all AI startups will survive, especially those lacking unique data or defensible business models. The advice emphasized focusing on fundamental product-market fit rather than just generative AI hype, as the market faces increasing consolidation pressure.
- The executive who issued the warning is Darren Mowry, a Google VP who leads the company's global startup organization, drawing on his previous experience at AWS and Microsoft. He argues that just as early cloud resellers were squeezed out when Amazon built its own enterprise tools, AI startups that are merely a thin layer on top of existing models will also be made obsolete. - Mowry specifically identified two vulnerable business models: "LLM wrappers," which add a simple user interface to a major AI model like those from Google or OpenAI, and "AI aggregators," which provide a single access point to multiple AI models. He noted that the "check engine light" is on for these types of companies as the underlying platform providers build competing features. - The warning comes as venture capital has heavily concentrated in the AI sector, which accounted for 52.7% of all global VC funding in 2025, totaling $270.2 billion. This represents a significant increase from 34% in 2024 and just 14% in 2020. - This capital, however, is not evenly distributed, with the majority flowing to large, foundational AI model developers in mega-deals. For instance, in 2025, OpenAI and Anthropic alone captured 14% of all global venture investment. - Signs of this consolidation are already appearing, with a 2026 report noting that multiple "AI wrapper" startups are beginning to shut down due to compressed margins. One prominent 2025 failure was Builder.ai, which burned through $445 million before it was revealed that much of its "AI" work was done by human engineers. - Mowry and other industry experts see more promise in startups that have strong defensibility. This includes companies with unique, proprietary datasets, those that are deeply integrated into specific industry workflows, or those building developer and coding platforms. - Examples of companies with more defensible models include Cursor, an AI-powered coding assistant, and Harvey AI, which is tailored for the legal industry. Mowry also expressed optimism for startups in biotech and climate tech, where large, specialized datasets can be a key advantage.
