startup ideas · 2026
Trending startup ideas in 2026
1,441 companies from 2026 cohorts are in the MarkosWeb corpus so far - four Y Combinator batches plus Techstars and 500 Global intakes. Trend here means measured movement: which clusters gained share batch over batch, which words appeared in founders' one-liners for the first time, and how the product form shifted. The headline: 30% of 2026 companies ship an autonomous agent, against 11% in 2022, while copilots fell to 3%.
1,441
companies from 2026 cohorts
67%
AI-native in 2026
30%
build an agent
3%
build a copilot
Rising in 2026
Share of classified YC companies, last two batches against the two before. Positive means the cluster is taking a bigger slice of each new batch.
- Robotics and physical world+7.5 pp · 17.1% of F26
- AI infra and compute+2.2 pp · 5.4% of F26
- Data for AI+1.9 pp · 7.2% of F26
- Horizontal AI assistants+1.2 pp · 4.5% of F26
- Agent infrastructure+0.4 pp · 17.1% of F26
- Security and compliance+0.2 pp · 3.6% of F26
Cooling
Losing share is not the same as shrinking: vertical AI agents still lead every batch. It means the marginal founder is choosing something else.
- Vertical AI agents-8.6 pp · 21.6% of F26
- Fintech-1.4 pp · 6.3% of F26
- Defense and space-1.0 pp · 1.8% of F26
- Consumer-0.8 pp · 2.7% of F26
Agents replaced copilots
Product form by cohort year, across all programs: AI at the core of the product, and whether it does the job itself (agent), assists a person (copilot) or is something others build on (infra).
| Cohort year | Companies | AI-native | Agent | Copilot | Infra |
|---|---|---|---|---|---|
| 2022 | 1,535 | 27% | 11% | 3% | 9% |
| 2023 | 1,674 | 45% | 17% | 7% | 10% |
| 2024 | 2,443 | 49% | 18% | 7% | 11% |
| 2025 | 1,908 | 60% | 26% | 5% | 15% |
| 2026 | 1,445 | 67% | 30% | 3% | 21% |
Words being born
Two-word phrases from company one-liners that barely existed before the current wave and are now used by several companies. Vocabulary moves a year before cluster shares do.
- prediction markets7 now · 2 before · first 2023
- claude code7 now · 2 before · first 2022
- ainative insurance6 now · 0 before · first 2025
- continual learning5 now · 0 before · first 2025
- product development5 now · 1 before · first 2024
- knowledge work5 now · 1 before · first 2024
- accounting firm4 now · 0 before · first 2025
- real life4 now · 0 before · first 2025
- cpa firms4 now · 0 before · first 2025
- energy infrastructure4 now · 1 before · first 2024
- medical imaging4 now · 1 before · first 2022
- ainative law4 now · 1 before · first 2022
- learning layer3 now · 0 before · first 2026
- deliver real3 now · 0 before · first 2025
- development lifecycle3 now · 0 before · first 2025
- compliance copilot3 now · 0 before · first 2025
- crowdsourced benchmark3 now · 0 before · first 2025
- data warehouse3 now · 0 before · first 2025
- fortune 5003 now · 0 before · first 2025
- health assistant3 now · 0 before · first 2025
Method and the longer list: Trends.
2026 cohorts in the corpus
| Program | Batch | Companies |
|---|---|---|
| Plug and Play | PnP 2026 | 279 |
| Y Combinator | S26 | 237 |
| Y Combinator | W26 | 199 |
| Y Combinator | X26 | 196 |
| Techstars | TS 2026 | 130 |
| Y Combinator | F26 | 111 |
| a16z speedrun | SR007 | 84 |
| a16z speedrun | SR006 | 63 |
| alchemist | Alchemist Class 41 | 26 |
| 500 Global | - | 20 |
| spc | - | 10 |
| antler | Antler UK 2026 | 10 |
| antler | Antler 2026 | 8 |
| seedcamp | Seedcamp 2026 | 8 |
| antler | Antler Australia 2026 | 8 |
| 500 Global | 500G MENA 11 | 7 |
| 500 Global | 500G Eurasia 10 | 7 |
| SOSV | SOSV HAX Seed 2026 | 7 |
| 500 Global | 500G Lucha - NONE | 6 |
| antler | Antler Canada 2026 | 5 |
| ef | EF London 2026 | 5 |
| antler | Antler US 2026 | 5 |
| antler | Antler Malaysia 2026 | 5 |
| antler | Antler Singapore 2026 | 5 |
The newest YC batch is still filling: Batch Live.
From a trend to an idea
- A rising cluster is a market with a tailwind and more competitors next batch. Check the crowded subniches on AI startup ideas before committing to the obvious one.
- A cooling cluster with a big graveyard is often the better entry: fewer new rivals, and the last wave's mistakes are documented on dead startup ideas.
- Demand side: Gaps compares YC's Requests for Startups with how many companies already build each theme.
More startup ideas by angle
- AI startup ideas →Where AI-native companies are actually being built
- Dead startup ideas worth reviving →Graveyards that founders are re-entering
- Stealth startups →How many accelerator startups are really in stealth
Validate any of them: Idea Check shows who already builds it; Idea Lab writes concepts from a cluster's data; Chat matches ideas to what you can build.