Key startup directions according to Y Combinator for Fall 2025

Requests for Startups is a long-standing Y Combinator tradition in which a team shares ideas they would particularly like to pursue. This list is not a constraint but rather a guide.

Published 2 min read

In the 2025 Fall list, the focus has shifted from demonstrating AI capabilities to its practical application. It is no longer enough to simply impress; the goal is to build—products, infrastructure, and new approaches. Y Combinator highlights areas where AI is not a supplementary function but the foundation upon which next-generation solutions can be created. It is in these sectors that the team sees the most significant opportunities for startups right now.

Reskilling workers for AI infrastructure

The AI revolution requires not only engineers but also real people who will build the infrastructure—from data centers to factories. Y Combinator is looking for startups capable of reskilling specialists—welders, electricians, technicians—using AI, AR/VR, and multimodal models. The idea is to create a new wave of vocational training, where AI becomes a personal mentor and a scalable teacher for the accelerated preparation of future workforce.

Video generation as a technological foundation

Video as a building material—this is how YC proposes viewing the development of generative models. Neural networks already create realistic clips, opening the path to new applications: from personalized films to custom online shopping and gamified simulations.

Small team—giant business

AI tools are radically changing the equation: now, one founder with a team of 10 people and $500,000 in funding can build a company valued in the tens of billions. YC is betting on such highly agentic founders—those who make quick decisions, operate without bureaucracy, and use AI not just for optimization, but as the engine of the entire business.

Tools for working with AI agents

The future of AI lies in systems where thousands of agents work in parallel, coordinate, and learn from each other. Such architectures are already used for complex tasks, but they are currently inconvenient to manage. YC is looking for founders ready to build tools and platforms for developing, scaling, and debugging multi-agent AI systems, making them accessible not only to top-tier laboratories.

AI-native enterprise software

Once upon a time, Salesforce and ServiceNow took off thanks to cloud technology—now it is the turn of AI-native startups. This is not about SaaS products with GPT bolted on, but about products where AI is built into the architecture itself. Y Combinator is seeking teams ready to reimagine B2B solutions: with maximum automation, intelligent user support, and a minimal barrier to entry.

Using LLMs instead of consulting for government agencies

The US government spends hundreds of billions annually on inefficient consulting, while LLMs are already capable of performing many of these tasks faster and cheaper. YC is interested in startups that will replace Accenture and Deloitte in the realm of government services—from auditing to law evaluation.