For Fall 2025, Y Combinator has published a set of startup ideas that treat AI not as a feature but as a foundation for creating entirely new opportunities. The vision is that the AI revolution extends far beyond models, chips, and software — it will reshape physical infrastructure, unlock new business models, and transform legacy industries. In particular, YC is looking for startups that tackle the skilled-labor shortage, explore new uses for generative video, build high-efficiency small teams, develop multi-agent systems, reimagine enterprise software, and bring AI to the government consulting market.

1. New Vocational Training Schools for the AI Economy 👷‍♀️

If the past few years were about proving what AI can do, the next phase is about actually building things with it. Making the AI revolution real requires large-scale physical infrastructure — data centers, semiconductor fabs, and more — and that creates an urgent need for skilled tradespeople: electricians, HVAC technicians, welders, and so on. As of 2025, the U.S. government's new AI Action Plan addresses this directly with a "workers first" agenda, directing the Departments of Labor and Commerce to fund rapid retraining programs for these physical-trade occupations.

This context creates a major opening for startups. YC wants to invest in companies that build new vocational training schools for the AI economy — training the workforce these jobs require. The bet is that AI-powered, personalized training programs can get people job-ready in months rather than years.

The key question is: how do you teach welding or pipe repair through AI? Unlike coding, these skills can't be learned through a keyboard — they require hands-on practice in real or simulated environments. YC sees multimodal AI as the answer here. Possibilities include voice AI that coaches someone through a task as they perform it, or combinations of AR/VR and AI tutors that let learners practice in simulated environments and receive feedback via visual models.

The business model is straightforward. Employers will pay for well-trained workers. Traditional trade training businesses struggled to scale because the quality of human tutors was hard to replicate — AI solves that problem. Build one effective AI teacher and you can scale it infinitely. This is an opportunity to create a massive business while ensuring that everyone can participate in and benefit from the new AI economy.

2. Generative Video: A New Computing Primitive 🎬

Video generation models are advancing at a remarkable pace. As of 2025, Google's Veo 3 can already produce realistic, sound-on clips of eight seconds for just a few dollars per video — often indistinguishable from real footage. Soon it will be possible to generate near-perfect video on demand at near-zero marginal cost, and video will become a new fundamental building block of software.

This shift makes a huge range of new ideas possible.

• Media and entertainment: Create new seasons of canceled TV shows, make personalized children's cartoons featuring your own family members, or imagine an AI-driven TikTok successor where every video is made for an audience of exactly one.
• Shopping experiences: See yourself wearing clothes or using products before you buy them online. Browse apartments and automatically see your own furniture placed in every listing you click.
• Games and simulations: Video games built without a game engine, APIs that return infinite robot training data, or the ability to have video calls with loved ones even after they're gone.

YC is looking for founders who treat generative video not as an output but as a new computing primitive — and who are building the apps, tools, and infrastructure for a world where infinite, low-latency video exists. 🚀

3. High-Efficiency Small Teams: The Future's $100B Companies 💰

Thanks to new AI tools, YC believes that small, high-efficiency teams — even solo founders — can now build billion-dollar companies with as little as $500K in YC funding. Fifteen years ago, the rise of cloud computing eliminated the need to spend enormous sums on physical server infrastructure, making it far easier to build large companies with much less capital. The same kind of shift is happening again today: new AI tools are making it easier for ambitious founders to scale without large headcounts.

This is why YC expects the best future startups to optimize for a single metric: revenue per employee. Small, efficient teams have enormous advantages over bloated incumbents. They aren't bogged down by the internal politics, excessive meetings, and lack of focus that stall large organizations. They can stay focused on winning through speed and execution. YC wants to support these high-efficiency founders and help build the first ten-person, $100 billion company.

4. Multi-Agent Systems: Automating Complex Workflows 🤖

AI agents are evolving beyond single-threaded loops into distributed workflows that fan out across multiple sub-agent calls within a single execution. These multi-agent systems are useful for everything from long-running workflows to agentic map-reduce jobs, where hundreds of thousands of sub-agents apply human-level judgment to filter and retrieve massive volumes of data in parallel.

But these systems are extremely hard to build. They require solving traditional distributed-systems problems — ensuring high throughput and reliability while controlling costs. They also introduce new challenges that look familiar but need to be solved at a higher level of abstraction, such as:

• How to write effective prompts for agents and sub-agents
• How to handle unreliable context
• How to monitor and debug these agents

YC has experienced these difficulties firsthand in real production environments, and is looking for developers who want to build tools that make these systems easier to create and maintain. If you want to make deploying a fleet of agents as routine and reliable as deploying a web service or running a Spark job, YC wants to hear from you.

5. AI-Native Enterprise Software: A New Era of SaaS 💼

Salesforce and ServiceNow are among the world's largest enterprise software vendors — each with over $10 billion in annual revenue and more than $200 billion in market cap. Both were founded roughly 25 years ago, and that's no coincidence. Salesforce built the first cloud-native CRM; ServiceNow built the first cloud-native ITSM system. They won by riding the wave of cloud computing and outmaneuvering incumbents. Countless other B2B startups from the early 2000s — NetSuite, SuccessFactors, and more — did the same. Their founders recognized that SaaS offered a way to build a 10x better product, and they knew that incumbents would struggle to adapt to a cloud-native world, giving them the window they needed to win.

AI offers today's founders that same generational opportunity. The enterprise software systems of the future won't merely be "systems of record" for tasks humans perform. AI will be deeply and thoughtfully embedded, helping employees work faster and more accurately. Imagine Cursor, but for sales, HR, or accounting. Just as before, today's incumbents will struggle to rebuild their products around this new technology — and that gives startups today the time they need to win. History doesn't repeat itself, but it certainly rhymes. If you're interested in building AI-native enterprise software, YC wants to talk.

6. Bringing AI to the Government Consulting Market 🏛️

The U.S. government spends more than $100 billion per year on consulting. As you might expect, this is not the most efficient or innovative corner of the economy. But there are now several compelling reasons to believe this is about to change.

Most importantly, as of 2025, there is significant political pressure to cut wasteful consulting and spending. Every part of government runs on software — usually custom software built by consulting firms — and anyone who has used it knows it could be done far better. And today's large language models are already capable enough to perform much of what consulting firms currently do.

YC recently invested in a company like FedRAMP that helps businesses get approved to sell products to the government, as well as a company helping government reduce regulation and use LLMs to verify that laws and policies coming out of government are actually legal. YC sees a much larger opportunity in all the work that government consulting firms like Deloitte and Accenture currently do for government, and wants to invest in startups building LLM-powered software to take on that work.

Closing Thoughts

Y Combinator's Fall 2025 startup requests send a clear message: AI is becoming the core driver of a paradigm shift across all of society, not just a single industry. From training the workforce needed to build physical AI infrastructure, to enabling new forms of content and interaction through generative video, to high-efficiency teams scaling to outsized outcomes, to better tooling for complex AI systems, to disrupting legacy enterprise software and government consulting — the opportunities AI will unlock are vast. These are ideal conditions for founders to build transformative businesses with AI at their core, and YC is ready to back the founders with the vision to seize them.