We asked AI which YC startups will make it big. Here are its 10 picks
We showed an AI the launch pitches of 33 brand-new Y Combinator startups and asked which will make it big. Its #1 pick has already raised money.
The question“Which company is more likely to succeed, meaning to become a large, lasting business?”
- 🥇
Stoa Markets
A marketplace for GPUs and AI servers
AI's takeA huge, urgent market, and $300M in quote requests in its first month.
- 🥈
Discovered Materials
AI agents to discover new materials
AI's takeTackles GPU heat with patentable materials and deep technical founders.
- 🥉
Salem Robotics
Software for industrial inspection robots
AI's takeReal domain expertise in high-stakes inspection, plus paying pilots.
- 4
Bloomy
AI-powered mastery learning for K-12
AI's takeA big, well-understood problem, and an early pilot showing 1.8× growth.
- 5
Mireye
Infrastructure for physical-world AI agents
AI's takeA data moat: 366 fields of real-world facts that AI agents need.
- 6
Speko
OpenRouter for voice AI
AI's takeBenchmarked routing for voice agents, with usage already growing.
- 7
Agnost AI
User feedback from agent conversations
AI's takeLive customers and about 1M messages a day already flowing through it.
- 8
Simantic
Firmware simulation for AI agents
AI's takeA from-scratch simulation engine, and a satellite customer saving time and money.
- 9
Traceforce
Security monitoring for AI apps
AI's takeAlready deployed on 1,000+ devices across 10 organisations.
- 10
Coasty
An API for computer-use agents
AI's takeTakes on healthcare prior authorisations, a painful, API-less workflow.
⚔️ Head-to-heads
Simantic’s solution is more convincing and more defensible: a custom Rust engine and 24/7 hardware validation rigs. Stoa’s defensibility rests largely on relationships and a brokerage-style model.
Traceforce addresses a large, immediate enterprise problem with concrete deployment evidence. Discovered Materials targets a deep but slower, more uncertain R&D problem.
Bloomy targets K-12 tutoring, a crowded market with a weaker moat, and its evidence is a pilot the authors themselves caveat.
Concrete early traction: a named customer case, about 1M messages a day, public pricing. Pomerado’s evidence is thinner.
Y Combinator is the startup school behind Airbnb, Stripe and Dropbox. Its 2026 classes are packed with AI companies, and dozens introduced themselves to the world with a launch post on Hacker News.
We gave all 33 of those pitches to an AI and asked one simple question: which of these will become a big, lasting business? It compared them two at a time, 320 times over, until a clear top 10 emerged.
💸 Some are already raising money
Before we looked at the AI’s picks, we checked which of these startups had announced new funding since joining YC. Four had:
| Startup | AI’s rank | New funding |
|---|---|---|
| Stoa Markets | #1 🏆 | Raised a round in November 2026 (source) |
| Salem Robotics | #3 | $1.8M (listing) |
| Speko | #6 | $1.1M (source) |
| Context.dev | #20 😬 | $9M (source) |
Three of the four made the AI’s top 10, including its #1. The biggest round so far, though, went to a startup the AI ranked 20th. Nobody’s perfect.
⏳ Check back in 2028
We’ve locked the AI’s full ranking of all 33 startups so it can’t be changed. In November 2028 we’ll see which ones went on to raise big follow-on rounds, and whether the AI’s favorites beat the rest.
🤖 The fine print
- The AI only read the pitches. It didn’t know the founders, the investors or the startups’ real numbers, which is what investors care about most.
- It couldn’t cheat. These startups launched after the AI was built, so it had no idea how they’re doing.
- This is the AI’s hot take, not investment advice.