AN-035 Company lists · list
Deep Tech Startups: What Counts, and 50 Examples
Deep tech startups defined by a stated rule, with 50 examples from the HardwareMap index: chips, photonics, fusion, fission, launch, neurotech and robots.
by the HardwareMap editors3 min read
From the HardwareMap index, REV 26.09, refreshed 3 Oct 2026
Contents
Deep tech startups are companies whose product rests on a scientific or engineering advance that is not yet a commodity, so they spend years and large sums on research before selling at scale. From the HardwareMap index, REV 26.09, this page picks 50 examples across computing, energy, space, neurotech and robotics, and states the rule used. It sits under the hardware startups index.
What counts as deep tech
The label is used loosely, so here is the rule this list applies. A company counts when all three hold:
- New science or engineering in the product. The core part does not exist off the shelf: a wafer-scale processor, a superconducting magnet, a stage that lands, an implanted electrode.
- Technical risk before market risk. The open question is whether it can be built and made to work at scale, not whether anyone wants it.
- Long, capital-heavy development. Years of R&D and prototypes before meaningful revenue.
What does not count: good products assembled from mature parts. Repairable laptops, e-bikes, smart rings and 3D printers are hardware, often excellent, but not deep tech by this rule. Most of the 219 companies in the index are hardware startups; a minority are deep tech.
The 50 below are picked to cover each field, not ranked overall. Inside each table, companies sort by disclosed capital raised. "Raised" means disclosed equity only.
Deep tech startups in computing
Eighteen examples, from AI accelerators to light-based and brain-inspired processors.
Cerebras Systems builds processors the size of a whole wafer. PsiQuantum pursues a fault-tolerant quantum computer built from photons on silicon chips. Lightmatter, Ayar Labs, Q.ANT and Lumai move or compute data with light instead of electrons. Mythic and EnCharge AI compute with analog charge inside memory, and Innatera and Prophesee build brain-inspired chips and event-based sensors. Pragmatic Semiconductor makes flexible integrated circuits on thin substrates in its own UK fabs. The AI chip startups list covers accelerators in more depth.
Deep tech startups in energy
Thirteen examples: fusion, fission, geothermal, storage, hydrogen and carbon.
Four fusion approaches are here. Commonwealth Fusion Systems uses high-temperature superconducting magnets to shrink a tokamak and is building SPARC in Devens, Massachusetts, to demonstrate net fusion energy (CFS). Helion Energy pursues pulsed magneto-inertial fusion and aims to recover electricity directly from the plasma. Proxima Fusion designs stellarators, and Marvel Fusion uses lasers. On fission, Radiant Industries builds a helium-cooled microreactor that fits in a shipping container, and Oklo a sodium-cooled fast reactor. Form Energy stores grid power by reversibly rusting iron. See fusion companies for that field alone.
Deep tech startups in space
Ten examples: launch, in-space transport, reentry and stations.
Stoke Space is building Nova, a rocket designed so both stages land and fly again (Stoke). Firefly Aerospace's Blue Ghost made the first fully successful commercial Moon landing in March 2025. Isar Aerospace's Spectrum reached orbit on its second flight in September 2026. Varda Space Industries processes drug formulations in orbit and brings them back in capsules. The space startups list covers the rest.
Neurotech, medical imaging and robotics
Nine examples where the hard part is the interface with a body or the physical world.
Neuralink, Precision Neuroscience and Synchron take three routes to a brain-computer interface: threads implanted in the brain, a thin film through a slit in the skull, and a stent in a blood vessel. Butterfly Network put an ultrasound array on a chip; Hyperfine made MRI portable. Figure AI, Agility Robotics and Neura Robotics build humanoids, and Machina Labs uses paired robot arms to form sheet metal without dies. Investors who back this kind of company are covered in deep tech VC firms.
What the 50 deep tech startups have in common
Counted from their datasheets at REV 26.09:
- Large rounds. 49 of the 50 disclose a raise, and the median is about $734 million. 18 have disclosed $1 billion or more. Software startups rarely need that much before revenue; these companies need it for fabs, test stands, launch sites and clinical trials.
- Founded around 2018. The median founding year is 2018, and the oldest, Rocket Lab, dates from 2006. A decade from founding to scaled product is normal here.
- Some reach public markets first. Seven are public: Cerebras, Fervo, Oklo, Firefly, Rocket Lab, Butterfly Network and Hyperfine. Listing before profit is common in deep tech because the capital need outgrows private rounds.
For founders, the read is that the money and time are front-loaded. The seed fundraising guide covers how the first round is structured when the product is years away.
Frequently asked questions
What is a deep tech startup?
A company whose product depends on a scientific or engineering advance that is not yet a commodity, and that needs years of research and capital before it can sell at scale. Fusion reactors, photonic processors, reusable rockets and brain implants qualify. A laptop or a smart speaker built from standard parts, however good, does not.
What is the difference between deep tech and hard tech?
The terms overlap and are often used interchangeably. Hard tech usually means any company building physical products, including consumer electronics and vehicles. Deep tech is narrower: the product rests on new science or engineering, so technical risk comes before market risk. Every deep tech startup in this list is hard tech, but most hardware startups are not deep tech.
Which deep tech startups have raised the most?
Among the 50 examples, Cerebras Systems has the largest disclosed raise at about $10.2 billion, then Commonwealth Fusion Systems at about $4.0 billion, Fervo Energy at about $3.2 billion and Tenstorrent at about $3.0 billion, all as of REV 26.09. Cerebras and Fervo are now public. Raised means disclosed equity rounds only.
Is deep tech a good investment?
This page records companies, it does not rate them. Deep tech rounds are larger and take longer to return capital than software rounds, and technical milestones can slip by years. Several companies here are public, which brings filings and audited numbers. Not investment advice: read each company's own disclosures before deciding anything.
Where are most deep tech startups based?
Of the 50 examples, the SF Bay Area holds the most, mainly in chips and photonics. Los Angeles leads in space, Boston and Munich in fusion, and Europe is strong in photonics and launch. The hardware startup hubs note maps the full index by city, so you can see which hub specializes in what.
HardwareMap catalogues deep tech alongside the rest of hardware, datasheet by datasheet. Open any company above for its stage, disclosed raise, team and status.
The index
All 219 startups, filterable by category, hub, stage and funding.