YC Requests for Startups (Winter 2027)

Updated 2026-10-27

With each batch, Y Combinator publishes its Requests for Startups (RFS): ideas its partners want founders to work on. The Winter 2027 list has 11 themes, and its common thread is that AI’s bottlenecks have moved from the models to everything around them: hardware, infrastructure, financing and the software that brings AI into factories, data centres and government. The original is on YC’s site. Below is our short summary of each theme in our own words, what it looks like from Europe, and whether it fits what AVA Ventures backs.

YC is clear that the list is a fraction of what it funds. Treat it as a signal of where experienced investors see open space, not as a list of the only fundable ideas.

The 11 themes at a glance

Themes that fit software founders in Europe

AI that turns workers into experts. YC’s point: wherever a credential is the bottleneck, a smart generalist with good AI tools can do work that used to need a specialist. Think insurance claims, tax and benefits paperwork, medical billing, immigration and trade compliance. European markets are fragmented by language and regulation, which makes these jobs more labour-intensive here, and the opportunity bigger. This is close to what we call vertical AI: software that knows one industry’s workflow end to end. The test investors will apply is whether customers pay for outcomes, and whether gross margin holds after inference costs.

Wall Street for the physical buildout. Robots, GPU clusters and data centres need capital up front, and banks still underwrite a robot like a forklift. YC wants lenders, leasing companies and hedging products built for these assets. Some of the plumbing is growing fast in Europe: venture debt passed €21 billion by September 2026, on track for a record year, much of it for AI infrastructure (PitchBook). For B2B fintech founders who understand both the machines and the money, this is a real gap.

Frontier AI inside the walls. Many of the most valuable datasets sit with banks, insurers, hospitals and defence suppliers that cannot send data to an outside API. YC wants on-premises models, deployment tooling and security accreditation. In Europe, data protection rules and the push for digital sovereignty make this demand stronger than in the US. Sales cycles are long, so founders need a few design partners who will pay before the product is complete.

Own your intelligence. Open-weight models now compete with frontier models on many tasks, and Europe has one of the leading open-weight developers in Mistral. YC wants companies that help businesses turn their workflows into evaluations and training data, and keep improving their own models as base models change. The opportunity is services-like at first, so the question for a seed investor is how much of it becomes repeatable software.

Themes with a European twist

AI for City Hall. YC estimates that faster permit approvals would add billions in tax revenue to US cities. Permits are slow in Europe too, and municipalities are buried in the same repetitive checks. Estonia shows that public services can run digitally end to end, which makes the Baltics a good place to build and test. Selling to government is slow, and we look at govtech case by case.

Verification layer for government benefits. YC wants a “verify once, apply everywhere” layer, the way Plaid connected bank accounts. Europe is ahead here: the EU Digital Identity Wallet is being rolled out under eIDAS 2.0, and Estonia’s X-Road already lets agencies share data with the citizen’s consent. The European opportunity is less about building the pipes and more about the products that use them.

Models people like talking to. YC argues that the big labs optimise for code and agents, and their models now write worse than older versions. A startup that cares about voice and style has room. Smaller European languages are a niche the large labs serve poorly. We invest in consumer apps once they have paying users and good retention.

Specialized foundation models. Small models for manufacturing, energy and biology that work from a few examples, at a fraction of the compute of general models. Europe’s industrial base is a natural source of data and customers. We do not back research labs, but we do back applications built on top of such models once they have revenue.

Themes outside our focus

Space, robot hardware and the data centre supply chain are capital-intensive hardware businesses. Europe is active in all three: European space tech raised a record $2.4 billion in the first half of 2026 (PitchBook), and the EU’s Critical Raw Materials Act targets the same magnet and materials dependence YC describes for the US. They need specialist hardware investors, and we are not the right first call.

Using the list as a founder

• Treat the list as validation, not a brief. If your company already fits a theme, use that in your pitch: it shows you are working on a problem experienced investors care about.
• Translate the theme to your market. Most of the RFS is written for the US. The same problem often looks different in Europe because of regulation, language and how public services work, and that difference can be your edge.
• Revenue still decides. At pre-seed and seed in Europe, investors look for paying customers more than a good theme. Our guide to pre-seed, seed and Series A valuations shows what rounds look like in 2026.

Building in one of these areas?

AVA Ventures invests $100k–500k in pre-seed and seed software companies in Europe: AI agents and applied AI, vertical software and vertical AI, B2B SaaS, B2B fintech and consumer apps with paying users. We usually invest from about €5k of monthly revenue, and we reply to every deck within two days.