YC's Request for Startups Decoded: What Silicon Valley's Smartest Money Is Betting On in 2026
March 11, 2026 · Nakshatra
By Nakshatra, Founder of Novara Labs | Published March 2026 | Last updated: March 11, 2026
Y Combinator just broke a 20-year rule. For two decades, the conventional wisdom in venture capital was simple: do not invest in services businesses. Services don't scale. Margins are thin. Revenue grows linearly with headcount. Venture-scale returns are almost impossible. Then, in 2025, YC added AI-native agencies to its Request for Startups — and told the world's most ambitious founders to build them.
That single move signals something bigger than a new category. It signals that the economics of professional services have been rewritten by AI, and that the smartest money in Silicon Valley is now chasing companies that look like software but sell like agencies. If you're a founder, operator, or investor, understanding what YC is actually betting on — across every RFS batch from Summer 2024 through Spring 2026 — is the difference between building for a wave or building for a footnote.
This guide decodes the full RFS map: the mega-themes, the thesis break on agencies, the "10-person $100B company" bet, and what you should do next.
Table of Contents
- What the Request for Startups Actually Is (And Why It Matters)
- The Five Mega-Themes Across Every YC Batch
- The Rule That Just Died: AI-Native Agencies
- The 10-Person, $100B Company Thesis
- The Full RFS Map: Spring 2026 → Summer 2024
- What YC Is Not Asking For
- What to Do Next: Apply, Build, or Partner
- FAQ
What the Request for Startups Actually Is (And Why It Matters)
The Request for Startups (RFS) is not a casual list of interesting ideas. It's a curated signal to the global founder community about where Y Combinator believes the largest, most transformative opportunities exist. When YC adds a category to the RFS, it's telling thousands of the world's most ambitious builders: build here.
Historically, the RFS has identified categories years before they became mainstream — fintech infrastructure, developer tools, biotech platforms, spatial computing. The inclusion of a category doesn't guarantee funding, but it means YC partners will be especially receptive to pitches in that space. Downstream investors use the RFS as a signal for where to develop their own theses. Talent follows. Capital follows.
YC has published RFS updates in Spring 2026, Fall 2025, Summer 2025, Spring 2025, Winter 2025, and Summer 2024. The themes that repeat across batches are the ones YC is most committed to. The themes that appear once and vanish are experiments. What we're decoding here is the persistent signal — the bets YC keeps making, in different language, across multiple years.
The Five Mega-Themes Across Every YC Batch
Across six RFS batches, five mega-themes show up again and again. If you're choosing where to build, these are the currents YC is riding.
1. AI as the operating system, not a feature
YC is not funding "AI-powered" versions of old businesses. It's funding AI-native ones. The distinction: in an AI-native company, AI is the primary production engine. If you removed AI tomorrow, the business would not function. That applies to agencies (AI does 70–90% of the work), enterprise software (Cursor for sales, HR, accounting), hedge funds (agents combing 10-Ks and making trades), and scientific tools (AI solving physics and chemistry at inference time). The repeated language — "AI-native," "built with AI as the foundation," "treat AI not as a feature but as the foundation" — is deliberate. YC wants companies that couldn't have existed two years ago.
2. Software margins in non-software businesses
The AI-native agency thesis is the sharpest expression of this, but the pattern appears everywhere: use AI to deliver outcomes that used to require human labor, at a cost structure that looks like software. Traditional agencies: 20–35% margins. AI-native agencies: 65–80%. Traditional consulting: headcount-limited. LLMs doing government work: Deloitte/Accenture replacement at a fraction of the cost. The bet is that every knowledge-work industry can be re-architected so that AI does the execution and humans do strategy, quality control, and relationships — and that those re-architected companies will have venture-scale economics.
3. Small, high-agency teams at massive scale
Fall 2025's "The First 10-Person, $100B Company" (Aaron Epstein) is the clearest statement: the best startups of the future will optimize for revenue per employee. Cloud eliminated the need for huge server spend; AI is eliminating the need for huge headcount. YC wants to fund "high-agency founders" who can build multi-billion-dollar companies with as little as $500K from YC — 10 people generating the revenue of 100-person companies, without the politics, meetings, and bloat. This theme connects to AI-native agencies (10 people serving 100+ clients), AI-native enterprise software (small teams shipping Cursor-like tools for every function), and infra for multi-agent systems (small teams operating fleets of agents).
4. Government and regulated industries as the next frontier
Government software (Harj Taggar), AI for government (Tom Blomfield), using LLMs instead of government consulting (Gustaf Alströmer), infra for government fraud hunters (Garry Tan) — selling to government appears in multiple batches. The pitch: government is the biggest customer on earth, spending trillions; it's slow, manual, and desperate for efficiency. AI can automate forms, applications, compliance, and fraud recovery. The first movers who crack distribution will have sticky, expandable contracts. It's hard — but the prize is huge. Same pattern for healthcare AI, compliance and audit, and regulated fintech (stablecoins, FedRAMP).
5. Physical world meets AI: robotics, manufacturing, spatial, guidance
AI is not staying in the browser. YC is betting on AI that touches the physical world: modern metal mills (Zane Hengsperger), AI guidance for physical work (David Lieb — "turn off that valve," "use the ⅜ inch wrench"), software tools to make robots (Diana Hu), large spatial models (Ryan McLinko), spatial computing (Summer 2024), manufacture in the USA (Jared Friedman), AI-aided engineering tools (Diana Hu). The thesis: the next decade will see AI move from text and code into factories, field service, healthcare at the bedside, and real-world environments. The companies that define that transition will be infrastructure (models, tooling) and vertical applications (mills, robotics, physical-work coaching).
The Rule That Just Died: AI-Native Agencies
Why VCs have avoided services for 20 years: Services don't scale. Margins are 20–35%. Revenue scales linearly with headcount. Moats are shallow (your best people can leave). Key-person risk is enormous.
What changed: AI breaks the linear relationship between headcount and revenue. In an AI-native agency, the 11th client adds minimal marginal cost because AI systems handle execution. The human team does strategy, quality control, and client relationships. Gross margins of 65–80% are achievable — comparable to SaaS. A 10-person AI-native agency can generate the revenue of a 100-person traditional agency. That's the economic unlock.
Who's saying it: Aaron Epstein, YC Group Partner and co-founder of Creative Market (acquired by Autodesk), has become the most vocal advocate. His thesis: "Agencies of the future will look more like software companies, with software margins." YC is not funding tools that help traditional agencies work slightly faster. It's funding new companies that replace the traditional agency model — delivering the same or better outcomes with fundamentally different cost structures.
What counts as an AI-native agency (per YC and the market):
- Marketing and content — AI produces 50–100 SEO articles, social campaigns, ad creative per month; humans strategize and QA.
- Sales development — AI researches prospects, personalizes outreach, runs sequences; a tiny team manages the system.
- Legal — AI does contract review, due diligence, compliance; attorneys verify and advise.
- Accounting and back-office — AI categorizes, reconciles, reports; CPAs oversee and handle edge cases.
- Design and creative — AI generates variations; human creative directors curate and maintain brand.
The window is closing. First-movers are accumulating data, client relationships, and proprietary workflows. The best time to build was six months ago; the second-best time is now. For a full breakdown of the model, see our MVP development cost guide — we're an AI-native agency ourselves, and we've written extensively on how the economics work.
The 10-Person, $100B Company Thesis
In Fall 2025, Aaron Epstein wrote: "Thanks to new AI tools, we believe it's now possible for small, high-agency teams – even solo founders – to build multi-billion dollar companies with as little as just $500k in funding from YC."
That's not a metaphor. YC is explicitly optimizing for revenue per employee. The argument: 15 years ago, cloud eliminated the need for massive server spend. Now AI is eliminating the need for massive headcount. Small teams have advantages — no politics, fewer meetings, faster execution. The best startups will be 10-person companies doing what used to require 500.
This ties directly to:
- AI-native agencies — 10 people serving 100+ clients at 65–80% margins.
- AI-native enterprise software — "Cursor for sales, HR, accounting" built by small teams.
- Infrastructure for multi-agent systems — small teams operating fleets of agents that do the work of hundreds of humans.
- Full-stack AI companies — don't sell AI to law firms; start a law firm run by AI and compete with incumbents.
If you're building a startup in 2026, the question YC wants you to ask is: How do we 10x output per person instead of 10x headcount?
The Full RFS Map: Spring 2026 → Summer 2024
A condensed map of every RFS category across the last six batches. Repeated themes = persistent bets.
| Theme | Spring 2026 | Fall 2025 | Summer 2025 | Spring 2025 | Winter 2025 | Summer 2024 |
|---|---|---|---|---|---|---|
| AI-native services | AI-Native Agencies | — | Full-stack AI companies, Voice AI, Healthcare AI | AI Personal Staff, Vertical AI Agents | — | — |
| Product & PM | Cursor for Product Managers | — | — | — | — | — |
| Finance & crypto | Stablecoin Financial Services | — | AI for Personal Finance | B2A, Compliance & Audit | Stablecoins 2.0, Fintech 2.0 | Stablecoin Finance |
| Government & regulated | AI for Government, Infra for Gov Fraud Hunters | Using LLMs Instead of Gov Consulting | — | — | Government Software, Public Safety | — |
| Physical world | Modern Metal Mills, AI Guidance for Physical Work | — | Software Tools for Robots | — | Manufacture in USA, AI-Aided Engineering | ML to Simulate Physical World, Robotics |
| AI infra & agents | Make LLMs Easy to Train | Infra for Multi-Agent Systems | Internal Agent Builder, AI Research Labs, Devtools for Agents | Devtools for AI Agents, Future of Software Engineering, Inference Infra | — | — |
| Models & frontier | Large Spatial Models | Video Generation as a Primitive | — | — | LLMs for Chip Design | Small Fine-tuned Models, Explainable AI |
| Enterprise | — | AI-Native Enterprise Software | — | — | — | AI to Build Enterprise Software, New ERP, LLMs for Back Office |
| Workforce & education | — | Retraining Workers for AI Economy | AI Personal Tutor, Future of Education | — | One Million Jobs 2.0 | — |
| Other | AI-Native Hedge Funds | First 10-Person $100B Company | More Design Founders, AI Personal Assistant, AI Residential Security, Voice for Email | Secure AI App Store, Datacenters, Browser Automation, AICOSS, AI Coding for Hardware | New Space Companies | New Defense, Commercial Open Source, Spatial Computing, Dev Tools from Internal Tools, Cancer, Foundation Models for Bio, MSO Healthcare, Eliminating Middlemen |
Use this map to see where YC's attention has been consistent (agencies, infra, government, physical world, fintech) versus one-off (e.g. cancer, space). The persistent columns are where the capital and talent will concentrate.
What YC Is Not Asking For {#what-yc-is-not-asking}
Clarity on what YC wants also requires clarity on what it's not asking for.
- Not "AI-powered" incumbents. YC is not funding a traditional agency that uses ChatGPT. It wants companies that are structurally AI-native — where removing AI would collapse the business.
- Not tools that make old models slightly better. Dashboards for agencies, plugins for consultants — no. Replace the model; don't optimize it.
- Not ideas without a path to software-like margins. If the unit economics still depend on linear headcount growth, it's not the bet.
- Not pure research without a commercialization path. AI Research Labs (Summer 2025) is the exception — YC will fund deep research — but the bar is OpenAI-level ambition and team.
The RFS is a filter. If your idea fits the themes above and doesn't fall into the "not" list, you're in the zone YC is actively looking for.
What to Do Next: Apply, Build, or Partner
If you're a founder considering applying to YC:
Use this map to position your company. Frame your startup in the language of the RFS: AI-native, outcome-based, software margins, high agency, non-linear scaling. If you're building an AI-native agency, say so — and show early traction (revenue, clients, or measurable output). YC has broken the "no services" rule; your job is to show you're the new kind of services company they're betting on.
If you're building in one of these categories but not applying:
The RFS is still useful as a validation signal. If YC is asking for it, downstream VCs are likely developing theses there. Use the themes to sharpen your narrative and target investors who care about the same mega-trends.
If you need to ship fast in one of these spaces:
AI-native agencies (including ours) exist to help founders move at the speed of the RFS. We're built on the same thesis — AI does the production work, humans do strategy and quality — and we help startups ship MVPs, content, and automation in days instead of months. If you want to test an idea in a YC-hot category without hiring a full team, talk to us.
FAQ
What is Y Combinator's Request for Startups?
The Request for Startups (RFS) is a public list of categories where Y Combinator is actively looking to fund new companies. It represents YC's collective view on where the largest opportunities exist. The RFS is updated periodically (roughly every batch or so). Being on the list doesn't guarantee funding, but it means YC partners will be especially receptive to pitches in that category.
Why did YC add AI-native agencies to the RFS?
YC added AI-native agencies because the economics of professional services have changed. Traditional agencies have 20–35% margins and scale linearly with headcount. AI-native agencies use AI for 70–90% of production work and can achieve 65–80% gross margins while serving many more clients with the same team size. That makes them venture-scale for the first time. YC Group Partner Aaron Epstein has been the main advocate; his background at Creative Market gives him deep experience in how creative and professional services are delivered at scale.
Can I apply to YC with an AI-native agency?
Yes. YC has explicitly included AI-powered professional services in the RFS. Strong applications show early traction (revenue, clients, or measurable output), a clear vertical (e.g. content, sales, legal, accounting), and a credible plan for how AI creates structural advantages over traditional delivery. A working product and real users significantly increase your chances, as with any YC application.
What's the difference between an AI-native agency and an AI SaaS tool?
An AI SaaS tool is software the customer operates themselves. An AI-native agency delivers finished outcomes — the client defines what they need and receives the result. Most businesses don't want to become experts in AI tooling; they want campaigns created, books balanced, and contracts reviewed. The agency model sells the outcome; the SaaS model sells the tool.
How do I use the RFS if I'm not applying to YC?
Use it as a thesis map. The categories that repeat across batches (AI-native services, government, physical world, infra, fintech) are where YC believes the biggest opportunities are. That influences where other investors look, where talent flows, and where incumbents will be disrupted. Aligning your company narrative with these themes can help with fundraising, hiring, and positioning — whether or not you ever apply to YC.
What does "10-person $100B company" mean?
It's a YC thesis from Fall 2025: the best startups of the future will be built by small, high-agency teams that generate outsized revenue per employee. AI and cloud let 10 people do what used to require hundreds. YC wants to fund founders who optimize for that — revenue per employee, speed, and leverage — rather than headcount growth. AI-native agencies are one expression of this; so are AI-native enterprise software companies and infra for multi-agent systems.
The Bottom Line
YC's Request for Startups is a compressed map of where Silicon Valley's best accelerator is placing its bets. The rule that just died — "never invest in services" — died because AI rewrote the economics. The themes that keep showing up — AI-native everything, software margins in non-software businesses, small teams at massive scale, government and physical world — are the themes that will define the next wave of venture-scale companies.
If you're building in one of these spaces, you're building with the wind at your back. If you're not, the RFS is still the best public document for understanding what the smartest money in startups thinks is coming next.
Ready to build in the direction YC is betting? Novara Labs is an AI-native agency — we live the thesis every day. We help founders ship MVPs, content, and automation in days. Start a conversation.
This guide is maintained by Novara Labs, the AI-native studio for founders who refuse to wait. We build MVPs, AI systems, and automation pipelines in days — not months.