YC just told you where AI money is going in 2026.

Y Combinator’s Spring 2026 “Requests for Startups” isn’t just a wishlist. It’s a market signal from the organization that has seen more AI startups up close than anyone else.

When the world’s biggest accelerator says “build this,” that’s not opinion. It’s pattern recognition at scale.

YC has a portfolio spanning dozens of different AI companies. That exposure gives them a ground-level view of what’s actually working and what’s not.

Cursor for Product Managers

Coding agents are already handling implementation. The bottleneck is now WHAT to build, not HOW.

YC sees a clear gap: there’s no AI system for the full product discovery loop — user interviews, feedback synthesis, feature prioritization.

Business implication: the next wave of AI ROI isn’t in writing code. It’s in making better product decisions, faster.

AI-Native Hedge Funds

YC believes we’re watching the same shift in investing that we once saw with quant trading.

It’s not about layering AI on top of existing strategies. It’s about rebuilding the entire research and decision-making process from scratch.

Business implication: competitive advantage in finance will come from AI architecture, not just AI access.

AI-Native Agencies

Traditional agencies scale by hiring more people. Thin margins, slow output. AI flips this model entirely: you sell the finished product, not the hours.

It’s the perfect blend of a services business and a product business. A design agency that shows you the final result before you sign. An ad agency that produces video without a physical shoot. A law firm that delivers in minutes, not weeks.

Huge margin potential, but also serious competition on the horizon.

Business implication: service businesses that adopt AI start operating with software-level margins.

AI for Government

AI has made it easier than ever to fill out forms and submit applications. But government is still printing them out and processing them by hand.

YC sees a massive, underserved market here. Hard to break into, but contracts are sticky once you’re in.

Business implication: the least hyped vertical might have the most stable revenue.

Government Fraud Investigation

AI can parse messy PDFs, trace complex corporate structures, and package findings into complaint-ready files in a fraction of the time. No more manual digging.

Business implication: making fraud recovery 10x faster isn’t just a big business opportunity. It returns billions to taxpayers.

Spatial Reasoning Models

The next frontier for LLMs is spatial reasoning: understanding geometry, physical structure, and 2D/3D relationships.

Models can handle basic depth estimation today, but they still struggle to reason about how objects relate in space or how to manipulate them.

Business implication: whoever cracks spatial reasoning owns the interface between AI and the physical world. That’s robotics, manufacturing, design, and medicine.

Making LLMs Easier to Train

Training LLMs sounds straightforward in theory. In practice, you run into bugs, GPU headaches, and terabytes of data to wrangle.

The opportunity: APIs that abstract the training process, databases built for massive datasets, dev environments designed for ML research from the ground up.

Right now we’re focused on general-purpose models. But the direction is clearly toward specialized models — high intelligence, lower data overhead.

AI Guidance for Physical Work

We’re not replacing physical work. At least not yet.

But we can make it dramatically better with AI guidance, and this use case pairs perfectly with wearable devices. Meta is already investing heavily here.

Imagine real-time coaching through a phone or smart glasses. Workers effective from day one, without years of on-the-job training.

Business implication: the skilled labor shortage is an AI opportunity hiding in plain sight.

Every category on this list shares one thing in common.

It’s about rebuilding entire workflows and business models with AI at the center, not bolted on at the edge.

Even working with B2B companies today, the biggest ROI consistently comes when we rethink the process with AI in mind from the start, not after.

Kamil Kwapisz