Y Combinator took a third of the Slush 100 semifinals
Also see more events coverage: /events/
Slush shortlisted 100 startups from more than 1,400 applications this year. Wave Ventures, the Helsinki firm whose investment team is all 25 or under, cut those 100 down to 20 semifinalists. They pitched at Startup Stage on 18 November for a EUR 500,000 investment that costs no equity. Goldfish won it.
Eleven of the 100 had been through Y Combinator. Six of the 20 semifinalists had. A random draw of twenty from a hundred would be expected to contain 2.2.
Get the Weekly brief
The week’s rounds, last calls and events, every Monday.
That is the only pattern in this cut that survives arithmetic. Not geography, not sector, not how much anyone has raised.
For anyone outside that world: Y Combinator is the San Francisco accelerator founded in March 2005 by Paul Graham, Jessica Livingston, Robert Morris and Trevor Blackwell. It has backed more than 5,600 companies worth a combined $600 billion, among them Airbnb, Stripe, Dropbox, Coinbase, DoorDash, Instacart and Twitch. A YC batch has long been read as a predictor of fundraising: its companies tend to raise fast and at strong valuations.
What this cut suggests is that the signal travels further than money. It also predicts who gets picked on the conference circuit, which is not what the program was built for, and that is the interesting part.
The cut, measured against chance
Only Y Combinator beat the draw
Expected versus actual semifinalists, twenty drawn from one hundred.
Only Y Combinator beat the draw. Paired bars. Y Combinator: 2.2 then 6 (+3.8); United States: 3.0 then 5 (+2.0); United Kingdom: 3.4 then 4 (+0.6); Switzerland: 2.8 then 2 (-0.8); Germany: 2.6 then 1 (-1.6).
Only Y Combinator sits clear of its expectation. A random draw of twenty from a hundred would turn up six or more Y Combinator companies about eight times in a thousand. The four countries land where chance would put them, give or take a company and a half.
The story that looks true and isn’t
Germany had 13 companies in the Slush 100 and got 1 semifinalist. Switzerland had 14 and got 2. Written that way it reads as a verdict on European deep tech.
It does not survive the arithmetic. With twenty drawn from a hundred, Germany’s expected share is 2.6 and Switzerland’s is 2.8. Landing on 1 and 2 is the kind of thing a fair draw does routinely. There is no finding here, only a small sample behaving like a small sample.
The same caution killed a second story. Before we corrected one record, the United States looked over-represented at 6 of 16. Morpheus turned out to be a London company that our database had matched to an American firm of the same name. At 5 of 15 the American result is unremarkable, and the “US founders travel better” reading goes with it.
What is left is Y Combinator, at nearly three times its expected share.
Nor is it money
The obvious explanation is that Y Combinator companies are simply further along. We cannot test that from our own data, and the outside data does not support it either.
Dealroom lists eight Slush 100 companies among the 28 startups it flags as hot at Slush this year. Four reached the semifinals and four did not. Its ranking, built on funding, team and growth data and entirely independent of ours, does not predict this cut. pyannote raised the largest disclosed round in the cohort, EUR 8.1 million, sits at 87 on Dealroom’s signal, and is not in the twenty.
The four largest disclosed rounds in the original 100 all belong to companies with no Y Combinator connection, and three of those four did not reach the semifinals.
Hardware raised less than software at both cuts
Carried over from our September piece on the full cohort of 100. USD millions.
Hardware raised less than software at both cuts. Paired bars. Median disclosed round: $2.2M then $3.1M (-0.9); Median total raised: $2.0M then $2.8M (-0.8).
Read the descriptions
Slush wrote a single line about each semifinalist. Set the Y Combinator ones beside the rest. “Retrofitting robots to be expert welders.” “Making oxygen from water.” “Robots as cheap workers.” “Making pesticides obsolete.” Verbs and outcomes, every one.
Now the others. A team from EPFL and Harvard. An ETH Zurich spinout. A team from Amazon, Apple, Wolt and CERN. Founders with EUR 200 million raised before. These are credentials, and credentials are what you lead with when the product is hard to say quickly.
We only see the second step. Slush does not publish what the 1,400 applications looked like, so we cannot tell whether Y Combinator was already over-represented on the way into the 100. What we can see is that choosing twenty companies to put on a stage for eight minutes each is partly a question of who will be legible to a room.
None of this is a trick. Saying what you do in one line is a hard skill and it is worth as much in front of a customer as in front of a jury. Y Combinator teaches it deliberately and at scale. The finding is not that its companies work the format. It is that the thing YC trains turns out to be the thing a selection round rewards. We cannot separate that from the simpler reading, which is that YC picks well in the first place.
The twenty
| Rank | Company | Based | Sector | Stage | Raised |
|---|---|---|---|---|---|
| 01 | AlludeTeam from EPFL and Harvard making AI agents and robots in factories possible. Backed by Kima and Target Global. | France | AI | no data | amount not disclosed |
| 02 | BTRYETH Zurich spinout building ultra-thin solid-state batteries for next-generation electronics. Backed by HTGF and Redstone. | Switzerland | Hardware | no data | amount not disclosed |
| 03 | CURAFounders with more than EUR 200M raised before, now going after the 8 percent of global CO2 emissions that comes from cement production. | Canada | Climate and energy | no data | amount not disclosed |
| 04 | DAIVIN!Y Combinator startup making oxygen from water, so humanity can breathe from the depths of the ocean to the reaches of space. | United States | Hardware | no data | amount not disclosed |
| 05 | Foundation ClinicTeam from Amazon, Apple, Wolt and CERN, building the harmonization layer for European health data. | Finland | Health | no data | amount not disclosed |
| 06 | GoldfishWINNER. A personal AI memory that helps you write. Used more than a million times since early access opened in June. | United States | AI | no data | amount not disclosed |
| 07 | GraceXSerial founders raised EUR 5M in the company's first year to build a company brain, with privacy built in. | Italy | AI | no data | amount not disclosed |
| 08 | InLoop RoboticsY Combinator team from Apple, Bosch, BMW, TUM and Stanford, building robots as cheap workers. | United States | Robotics | no data | amount not disclosed |
| 09 | MarentisAn intelligence system for industrial assets, built so every engineer and operator can do more. | Germany | AI | no data | amount not disclosed |
| 10 | MonakoThiel Fellow and ex-LVMH team building 49-gram AI glasses that hide a full spatial XR Linux computer. | China | Hardware | no data | amount not disclosed |
| 11 | MorpheusExited founder and former VC, with the former head of AI platform at M&G, building the context layer for capital markets. | United Kingdom | AI | no data | amount not disclosed |
| 12 | MorphoAIY Combinator team building hardware foundation models, commercializing MIT and Harvard research. | United States | AI | no data | amount not disclosed |
| 13 | NextesyUnicorn founder, an AI PhD and an ERP expert building AI-native business administration. | Switzerland | AI | no data | amount not disclosed |
| 14 | NitroduckY Combinator backed, building a new Monsanto, this time to make pesticides obsolete. | Czechia | Climate and energy | no data | amount not disclosed |
| 15 | PallyY Combinator personal AI assistant. The founder left high school and became CFO of a 200-person company at 22. | United States | AI | no data | amount not disclosed |
| 16 | QuadriviaSerial unicorn founder building the AI roll-up for the USD 3 trillion smaller healthcare providers sector. | United Kingdom | Health | no data | amount not disclosed |
| 17 | Sqwish LabsFINALIST. Building System 1 for AI: decision models ready to use or tuned on your own data. | United Kingdom | AI | no data | amount not disclosed |
| 18 | Tekton DynamicsY Combinator team retrofitting robots to be expert welders. | United Kingdom | Robotics | no data | amount not disclosed |
| 19 | TerraSparkFINALIST. Team that has built unicorns and sent hardware to Mars, now building the future of energy infrastructure for Earth, in space. | Luxembourg | Climate and energy | no data | amount not disclosed |
| 20 | Upside RoboticsEx-founders and robotics leaders building autonomous robots for agriculture. | Canada | Robotics | no data | amount not disclosed |
| No company matches that filter. | |||||
Six carry Y Combinator: DAIVIN!, InLoop Robotics, MorphoAI, Nitroduck, Pally and Tekton Dynamics. The full cohort of 100 is on the Slush 100 2026 program page.
The final went elsewhere
Slush cut the twenty to three after the Startup Stage pitches: Goldfish, Sqwish Labs and TerraSpark. None of the three is a Y Combinator company. Goldfish, a personal AI memory for writing, took the EUR 500,000 on the Founder Stage on 19 November.
That does not reverse the finding above, and it is too small a sample to try. Three drawn from a pool of twenty holding six YC companies come up with none about a third of the time. It is still worth saying plainly: the group that cleared the first cut by a distance did not clear the second.
TerraSpark is the one finalist Dealroom’s ranking also flagged. At 94 it is joint top of that list, tied with HyLight.
What to do about it
The opening for European programs. Venture Kick, Venture Leaders, KlimUp and the long tail behind them put 38 of the 100 companies on that shortlist between them, which is real sourcing and it works. The piece they could add is narrative training, because a conference reads the narrative before it reads the science.
If you are a founder out of a lab. The gap between your bench and a stage is a skill, not a talent, and your program probably does not teach it. Six Swiss companies in this cohort have Venture Kick money on record and one reached the semifinals.
The open question. A pitch competition picks for the pitch, which is honest enough. But Slush 100 is read as a ranking of European early-stage companies, and the two things are drifting apart. Does Slush want the best companies or the best eight minutes, and should anyone expect those to be the same twenty?
Half the Slush 100 builds hardware
Fifty-two of the 100 companies Slush selected for 2026 sell something that exists in the physical world: a battery, a robot, a membrane, a sensor, a…
Ben CostantiniMethod and limits
The cohort of 100 is the dataset behind our September piece, Half the Slush 100 builds hardware, corrected since. The semifinalists, the finalists, their descriptions and the application and prize figures come from Slush. The hot startups ranking comes from Dealroom’s Slush 2026 page, as it stood on 13 November 2026.
Expected values are hypergeometric: the chance of drawing at least the observed number of a group when twenty are taken from a hundred at random. The limits are real. One jury, one year, twenty companies. Our Y Combinator count for the 100 is 11 or 12, because Slush calls Nitroduck YC-backed and our September record did not.
Congratulations to Goldfish, and to all twenty.
Sources
- Slush, “Meet the S100 semifinalists”, November 2026
- Slush, Slush 100 finalists announcement, 18 November 2026
- Dealroom, Slush 2026, read 13 November 2026
- Y Combinator, Wikipedia, read 13 November 2026
- Slush 100 2026 program page, Sesamers
- Half the Slush 100 builds hardware, Sesamers, 16 September 2026