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

Google VP Darren Mowry cautions that AI startups relying on simple “LLM wrappers” or acting as “AI aggregators” are facing increasing challenges. LLM wrappers, which add a user interface to existing large language models like Gemini or GPT, are seen as lacking sufficient differentiation. Similarly, AI aggregators that combine multiple LLMs into a single interface are losing favor as users desire more built-in intellectual property to optimize model selection. Mowry draws parallels to the early days of cloud computing, where resellers of AWS infrastructure were squeezed out by Amazon’s own enterprise tools.

He emphasizes the need for startups to build “deep, wide moats” – substantial and unique value propositions – to succeed. Examples of companies with strong moats include Cursor and Harvey AI. Mowry is optimistic about “vibe coding” and developer platforms, as well as direct-to-consumer AI applications, and also highlights the potential of biotech and climate tech due to increased data availability. He believes startups must offer real services beyond simply providing access to models to ensure long-term viability, as model providers expand their own feature sets and potentially cut out intermediaries.

(Source：Turks)