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 reactor. Fifty-six of them carry an AI tag in our database. Those two sets overlap on 13 companies.

Read that again. The list that a decade of tech press treated as a software index now splits down the middle, and the AI label sits mostly on one half of it. We pulled the whole 2026 cohort into the Sesamers database this week, enriched it company by company, and classified each one by what the customer actually buys. Below is where they come from, what they build, what they have raised, and who backed them before Slush did. The geography is the part that will surprise you.

Some context on how narrow the gate was. Slush took more than 1,400 applications for the 2026 edition, a record, and shortlisted 100 of them. That is a selection rate of about 7 percent. Wave Ventures, the Helsinki firm whose investment team is all 25 or under, cuts the 100 down to 20 semifinalists, and those 20 pitch at Startup Stage in November for a €500,000 investment that costs no equity. Slush published the application figure and Wave Ventures’ role on LinkedIn in September 2026.

Half the list makes atoms

We used two buckets. Physical means the customer buys something made of matter: a device, a material, a chemical process, a robot, a clinic. That is 52 companies. Software means the product ships as code, and that is the other 48.

The physical half spans several industries at once. It runs from BTRY (Switzerland), solid-state batteries with ceramic electrolytes aimed at European carmakers, to Divea (Switzerland), monolayer graphene membranes with angstrom-scale pores that capture CO2 at the flue, to Rhonexum (Switzerland), integrated circuits that keep working near absolute zero so quantum computers can lose some cabling. Add ROTOBOOST (Finland) cracking methane into hydrogen and solid carbon, Willo (Finland) delivering wireless power to devices that tilt and rotate, and WeldNova (Germany) holding a weld pool with a magnetic field so thick plate can be welded in a single pass.

Twenty of the 100 sit in industry, robotics or factory software. Twenty carry a health or biotech tag, 19 a climate or decarbonization tag, 10 fintech, nine agrifood, six defense or space.

The defense and space entries are worth naming because they were almost absent from this list five years ago. Navictus (Portugal) sells autonomous surface vessels for coastal security as a fleet service. Arqus Aerospace (Germany) builds directed-energy systems for space defense. Project-S (Germany) works orbital debris from three sides: space-based radar, avoidance software, robotic removal. TerraSpark (Luxembourg) is doing radio-frequency energy transmission for space-based solar, with terrestrial demos first, on a pre-seed of more than €5 million with Daphni among the investors.

Two years ago the same competition looked different. OASYS NOW won Slush 100 in 2024 with a platform matching patients to clinical trials, and the other two finalists were software as well: Mohana Health, a perimenopause care platform, and DevAlly, a compliance tool for the European Accessibility Act. The four trends we took back from Helsinki that December were second-hand commerce, medtech, Roblox and EU regulation. None of the four was about manufacturing, materials or energy.

AI stopped being a category

Fifty-six companies carry an AI tag. Forty-three of the 48 software companies carry one. When a label covers nine out of ten companies in a segment, it has stopped sorting anything.

I’m not arguing that the AI-first software companies here are weak. pyannote (France) commercializes more than a decade of CNRS speaker-diarization research and raised the largest disclosed round in the cohort, €8.1 million co-led by Crane Venture Partners and Serena. Neuphonic (United Kingdom) generates speech word by word in under 25 milliseconds, on a £3 million pre-seed led by Moonfire. Literal Labs (United Kingdom) builds models on Tsetlin machine architecture that are explainable by construction, on a £4.6 million pre-seed led by Northern Gritstone and Mercuri. Those are real companies with real differentiation.

The point is what the tag no longer tells a jury, an investor or a buyer. Where AI does still sort the list is in combination with something physical. Xatoms (Canada) uses quantum chemistry and AI to discover visible-light photocatalysts for water purification, with pilots in Texas, Kenya and South Africa, on a $3 million pre-seed led by Quantacet. Qalam Health (Canada) built a handheld probe that uses spectroscopy and explainable AI to tell malignant from healthy bone at the surgical margin. Theseus Robotics (Switzerland) sends autonomous robots down supermarket aisles to restock shelves.

Thirteen companies of a hundred sit in that overlap, by our classification. If you are looking for the scarce profile in this cohort, that is it.

Switzerland beats France, and it isn’t close

Here is the top of the table: United Kingdom 16, United States 16, Switzerland 14, Germany 13, then France, Canada and Finland tied at six.

Six of a hundred are Finnish. At a Finnish event, that is a jury refusing the home-field discount.

The 2026 cohort is truly global: 71 companies from Europe, 16 from the United States, and the rest from Canada, China, India, Israel, Türkiye, Colombia, Armenia and Georgia. The Nordics account for eight of the 100 in total, six Finnish and two Swedish.

Switzerland at 14 is the number worth sitting with, because of what those 14 are. Nine of them build something physical. Six have money from Venture Kick, the Swiss pre-seed program whose first stage is a grant, on record, and several are ETH Zurich or EPFL spinouts: MYNERVA, Manukai and UNOMR out of ETH, Divea and Rhonexum out of EPFL. Germany’s 13 lean the same way, nine physical. The two Anglo blocs lean the other way: 12 of the 16 American companies and 11 of the 16 British ones ship software.

So the cohort has two distinct pipelines. One runs through federal research institutes and non-dilutive grants in German-speaking Europe and produces hardware. The other runs through accelerators and angel money in London and San Francisco and produces software. Slush picked roughly equal numbers from each.

Y Combinator is the cohort’s biggest single backer

Eleven of the 100 have been through Y Combinator, across batches running from S23 to Summer 2026. No other program comes close. Nine of the 11 are American.

That is the real story of the American entries: Rimba, Karumi, Pally, Stoa, Frizzle, SpaceFlow, MorphoAI, InLoop Robotics, DAIVIN!. Nine of the 16 US companies in the cohort share one alma mater. Y Combinator has become a distribution channel into European stages.

The answer on this side of the Atlantic is fragmented and mostly non-dilutive. Venture Kick appears five times as a program, Venturelab’s Venture Leaders three, KlimUp in Zurich twice, then a long single-mention tail including Techstars Berlin Deep Tech, Wyss Zurich, ESA BIC Bavaria, EIT Manufacturing, Entrepreneur First, HAX, YES!Delft in Delft, Station F in Paris, Lloyd’s Lab in London, Zinc, Alderley Park, and Centech and LE CAMP in Quebec. Thirty-eight companies have at least one program on record, and 54 have at least one named investor.

Every one of those programs does useful work. Not one of them puts nine companies on this list.

The leads are European, and the American money arrives through YC

Twenty-six of the 40 disclosed rounds name a lead investor, and that list holds the one pattern in this cohort nobody put in a press release. Almost every lead is a European fund, and no American venture firm leads a single round.

The leads read like a map of the continent’s early-stage market. Amadeus Capital Partners and Hoxton Ventures in London, Cherry Ventures and Redstone in Berlin, Serena and Iris Capital in Paris, Kfund in Madrid, Koinos Capital in Milan, Butterfly Ventures in Oulu, byFounders in Copenhagen, 3xP Global in Lisbon, RockawayX in Prague, Uni.fund in Athens. Regional and institutional money is in the mix as well: Northern Gritstone out of Manchester co-led Literal Labs, Ground State Ventures in Amsterdam, the quantum fund formerly known as QDNL Participations, led Rhonexum, and High-Tech Gründerfonds and NBank Capital back alteva and Digity in Germany.

Two leads sit outside Europe: Quantacet in Montreal, which backs quantum companies only, and cyber.Fund. The two American names that do appear, Sequoia and a16z, reach the cohort as scout cheques in Tekton Dynamics. Neither of them leads a round here.

So the split is clean. European funds write the leads; American capital reaches this cohort as a batch place at Y Combinator, then a seed round raised in San Francisco. One of those routes gives a founder a term sheet. The other gives them a new address.

Hardware is raising less than software

Forty companies have a disclosed round in the database: 18 pre-seed, 18 seed, one grant, three unlabeled. Converted to dollars at mid-September 2026 rates, the median disclosed round is about $3 million, and roughly half of the 39 we could convert came in under that.

Split that by what the company builds and the gap runs the wrong way for physics. Median disclosed round for the physical half: about $2.2 million. For the software half: about $3.1 million. The 49 companies carrying a total-raised figure point the same way, a median of $2.0 million for physical companies against $2.8 million for software ones, on a field that is patchier and worth less weight.

Hardware in this cohort raises less than software at every cut we ran. Whatever else the shift to physical products represents, the capital has not shifted with it.

Hardware in this cohort raises less than software at every cut we ran.

The exceptions are instructive. Upside Robotics (Canada) raised a $7.5 million seed led by Plural for solar-powered field robots that place fertilizer precisely. AgriPass Robotics (Israel) raised $7.5 million for towed machinery that weeds without chemicals. Milvus Advanced (United Kingdom) raised £5.14 million led by Hoxton Ventures for copper, nickel and iron nanoalloys meant to replace platinum in electrolyzers. All three sell into an industry with a procurement budget already open.

Fifteen of them came out of a lab

Fifteen companies in the cohort describe themselves as a university or research-institute spinout: ETH Zurich, EPFL, Aalto, MIT, Harvard, TU Delft, Newcastle, Southampton, Sant’Anna in Pisa, INESC TEC in Porto, and Germany’s BAM materials institute. Cornea Sense (Finland) came out of Aalto with terahertz sensing for corneal hydration. CIWI (Netherlands) came out of TU Delft making water treatment chemicals on site from steel and salt. Brainalyze (United Kingdom) came out of Southampton reading a dementia fingerprint from one drop of blood.

Who fills the hardware funding gap

For investors: the scarce profile in this cohort is a company that puts a model on top of a physical process, and there are 13 of those. Go find them before November.

For event organizers: your startup program’s country mix is a claim you should be able to defend with a table. Slush’s 2026 cohort has six Finnish companies in it. Can you say what yours has, and would you publish it?

For founders building hardware: the median disclosed pre-seed for a physical company in this cohort is about $2.0 million across 10 rounds, and six of the Swiss companies have Venture Kick money on their record. Check what your national equivalent will write before you take a term sheet.

And for the rest of us, one open question. If half of the most-watched startup list in Europe now makes physical things, and the money has not moved, who fills that gap first: the national deep-tech funds, the corporates, or Y Combinator?

The 100 companies

The full 2026 cohort, classified by what the customer buys. Filter by physical or software, or search by name, country or what they build.

Photo: Konsta Linkola, via Sun Effects, which designs the lighting and staging for Slush.

Some of these records will be wrong. We built the cohort fast, from company sites, press and automated enrichment, and corrected a dozen entries by hand before publishing. If your company is in the 100 and its entry is off, tell us and we will fix it.

Method: cohort records built on sesamers.com and enriched through mytradeshow.ai, as recorded on 15 September 2026. A disclosed round is on file for 40 of the 100 companies, a named investor for 54, and a program or accelerator for 38. Physical and software are our classification. Application and semifinalist figures come from Slush’s own LinkedIn account, September 2026. Currency converted at EUR 1.08, GBP 1.27, CHF 1.25 and SEK 0.095 to the dollar.