AI aggregators — startups that combine multiple LLMs into one interface or API — may be facing similar pressure.

Platforms like Perplexity AI and OpenRouter provide orchestration layers that route user queries across different models, often bundling monitoring, governance, and evaluation tools.

But Mowry’s advice to new founders is blunt: “Stay out of the aggregator business.”

He argues that users increasingly expect embedded intellectual property — systems that intelligently determine which model to use based on context and need, rather than simply offering access to multiple models. As major AI providers roll out their own enterprise features, standalone aggregators could face margin compression and reduced relevance.

Mowry compares the situation to the early days of cloud computing. In the late 2000s and early 2010s, startups emerged to resell Amazon Web Services infrastructure, offering easier onboarding and consolidated billing. But as AWS expanded its native tools and customers became more sophisticated, many of those middlemen disappeared — leaving only companies that provided real value through security, migration, or DevOps expertise.

Despite his caution around wrappers and aggregators, Mowry remains bullish on AI developer platforms and direct-to-consumer tools. Startups like Replit and Cursor are seeing strong traction, and he expects continued growth in sectors where AI unlocks entirely new capabilities — including biotech and climate tech.

The AI gold rush may not be over, but the easy wins appear to be.