Y Combinator Portfolio Shows Shift to AI Infrastructure
A recent analysis of Y Combinator's 5,668 startups reveals a significant decline in pure consumer-focused companies, with AI becoming a core infrastructure layer across all verticals. The trend indicates that successful new companies are embedding AI agents deep into specific industry workflows rather than building general-purpose AI tools. This shift aligns with recent YC Demo Day presentations that featured agentic commerce and multi-agent orchestration platforms.
- The percentage of AI-tagged companies in Y Combinator's W24 batch reached 63%, a significant jump from 51% in the S23 batch and 29% in the batch prior, indicating a rapid acceleration of this trend. More than half of all new YC companies in 2025 are focused on building agentic AI solutions. - This focus mirrors the broader venture capital landscape, where AI startups captured nearly a third of all global funding in 2024, amounting to over $100 billion. The average funding round for a generative AI company soared to $407 million in 2024, nearly double the total investment value from 2023. - In the real estate sector, YC-adjacent startups are applying this trend directly. Tidalwave, which builds agentic AI to automate the mortgage process, raised a $22 million Series A and integrates directly with Fannie Mae and Freddie Mac. Meanwhile, Ridley, an AI-powered platform for commission-free home selling, secured a $6.4 million seed round led by Fifth Wall. - YC startups are building the infrastructure for "agentic commerce," where AI agents manage tasks like customer support and abandoned checkouts. Companies like Terminal Use are creating orchestration platforms for multiple background agents to work together on complex tasks like coding. - Top venture firms are investing with the belief that AI will create a $10 trillion opportunity by shifting work from "certainty to leverage." Sequoia Capital's current thesis is that vertical AI agents, trained on specific end-to-end workflows, represent a massive opportunity for startups and can outperform the best human experts in fields like security and DevOps. - The push for more powerful, embedded AI is driving developments in edge computing. Google's LiteRT framework, an evolution of TensorFlow Lite, enables high-performance on-device AI by providing faster GPU performance and unified access to neural processing units (NPUs), which can speed up models by up to 25x compared to CPUs.
