Deep tech means companies whose product depends on a genuine scientific or engineering advance rather than on software assembled from existing components. The defining test is technical risk: with deep tech there is a real possibility the thing does not work, and proving it works takes years, capital, and specialist people. Robotics, novel materials, quantum computing, fusion and next-generation energy, biotech tooling, semiconductors, space systems, and foundational AI research all sit inside the definition.
Ordinary technology companies take known building blocks and assemble them into a product. The risk is commercial: will anyone buy it. Deep tech companies carry that commercial risk plus scientific risk, longer development timelines, heavier capital requirements, and usually a regulatory path. That combination changes how they are funded, how they hire, and how they go to market.
The four tests of deep tech
A company is usually deep tech if it meets most of these.
1. Technical risk is real. Something fundamental might not work. Not "the integration might be difficult" but "the physics, chemistry, or biology might not cooperate". This is the primary test.
2. The advance is defensible. Patents, proprietary processes, hard-won manufacturing capability, or expertise that cannot be hired quickly. Deep tech moats come from difficulty rather than from network effects or speed.
3. Timelines run in years. Five to fifteen years from research to meaningful revenue is normal. Trials, certifications, pilot plants, and tooling cannot be compressed by working harder.
4. Capital intensity is high. Labs, fabrication, prototypes, clinical or field trials, and specialist staff. Deep tech companies typically raise more, earlier, against less revenue than software companies.
A useful negative test: if a competent engineering team could rebuild the core product in six months using publicly available components and models, it is probably not deep tech, however sophisticated it looks.
The term is also used loosely as a positioning claim, which is worth naming. Plenty of companies describe themselves as deep tech because the label attracts a particular kind of investor and a particular kind of engineer. The four tests above are the ones a technical diligence process will actually apply.
Deep tech companies are usually years from revenue and one conversation away from being misunderstood. That is a marketing problem long before it is a sales problem.
Michael Raven, Founder and CEO, Blazon Agency
Real examples, by field
Robotics and physical automation. Machines that do useful work in unstructured environments. Pillo Health built an FDA-cleared in-home health robot that dispensed medication and managed adherence, a product that had to satisfy medical regulation and consumer usability at the same time, and was later acquired by Stanley Black & Decker.
Biotech and life science tooling. Sequencing platforms, lab automation, novel therapeutics, diagnostic hardware. Long regulatory paths and evidence standards set by clinicians rather than by buyers.
Advanced materials and manufacturing. Batteries, structural composites, low-carbon cement, novel semiconductors. Value is proven in industrial qualification cycles that can run years per customer.
Energy and climate technology. Fusion, geothermal, grid-scale storage, carbon capture, alternative fuels. Capital intensity is extreme and the first customer is often a utility or a government.
Space systems. Launch, satellite platforms, in-orbit servicing, earth observation. Hardware risk plus regulatory complexity plus a small number of very large customers.
Foundational AI and computing. New model architectures, specialised inference hardware, quantum computing. Note the distinction that matters commercially: building a new model architecture or chip is deep tech, while wrapping an existing model in a workflow is a software product with AI in it.
Agriculture and food technology. Precision fermentation, cellular agriculture, engineered crops. Scientific risk plus consumer acceptance plus regulation.
For the sector view, the top deep tech startups to watch list covers current companies across these fields.
Is AI deep tech?
Sometimes, and the distinction is commercially important rather than semantic.
Deep tech AI means the advance itself is scientific: novel architectures, new training methods, specialised inference silicon, robotics that combine learning with hard physical constraints, or AI applied to a scientific domain where the model has to be validated against physical reality. The risk is whether the approach works at all.
AI-enabled software means an existing model, usually accessed through an API, is applied to a workflow. This can be an excellent business, and it is not deep tech: there is no fundamental technical risk, and the moat is distribution, data access, or execution rather than difficulty.
The reason to be precise is that the two go to market completely differently. Deep tech AI sells on proof and benchmarks to buyers who will validate claims technically. AI-enabled software sells on speed, price, and workflow fit. Positioning an AI-enabled product as deep tech invites technical scrutiny it cannot survive, and positioning genuine deep tech as ordinary software throws away its only real advantage.
Why deep tech matters
It addresses problems software cannot. Decarbonisation, disease, food security, and materials constraints are physical problems requiring physical solutions. No amount of application software fixes a battery chemistry limit.
The returns are structural. When deep tech works the advantage lasts, because it cannot be copied over a weekend. That is why investors accept the timelines.
It is where policy and capital are now pointed. Public research funding, industrial strategy, and sovereign technology programs across the US, Europe, and Asia have concentrated on exactly these fields, which changes both the funding environment and the buyer set.
Talent concentrates around it. The people who can do this work are scarce, which makes a company's public narrative a genuine competitive input rather than a marketing nicety.
Why deep tech companies market differently
Five practical consequences, which is where most deep tech go-to-market work actually goes wrong.
The buyer has no category. They are not choosing between vendors, they are deciding whether this class of thing belongs in their organisation at all. That decision is won with proof, not with persuasion.
Two audiences read everything. Customers and investors. The technical claim must survive expert review while the narrative stays legible to a generalist. Most deep tech companies write for one and lose the other.
Credibility is the currency. Peer-reviewed results, pilots, certifications, and named design partners outperform any campaign. The marketing job is to make existing proof findable and repeatable, and to plan the next proof point deliberately.
Cycles outlast campaigns. Six to twenty-four months from first conversation to contract, which means launch-month revenue is the wrong measure. Qualified conversations and pilot commitments are the right ones.
Milestones set the calendar. Funding rounds, clearances, pilot completions, and manufacturing readiness dictate timing, not a marketing plan.
The category label matters less than the question every deep tech founder has to answer: who feels this problem badly enough to change how they work?
Michael Raven, Founder and CEO, Blazon Agency
How deep tech companies are funded
Funding shapes go-to-market in deep tech more than in any other category, so it is worth understanding the pattern.
Grants and public research funding come first. Non-dilutive funding from national research programs, defence agencies, and climate or industrial strategy funds often carries a deep tech company through its earliest technical risk. It is slow, competitive, and does not require a market story, which is why some deep tech companies reach Series A without ever having written one.
Specialist investors, not generalists. Deep tech rounds are led by funds that understand long timelines and technical diligence. They evaluate the science and the team first, then the beachhead and the commercial path. A vague go-to-market answer is survivable at seed and rarely survivable at Series B.
Milestone-based rather than growth-based. Rounds are raised against technical and regulatory milestones rather than revenue multiples. This is why launch timing follows company milestones: the clearance, the pilot result, or the manufacturing readiness is both the fundraising trigger and the marketing moment.
Strategic and corporate investors appear early. Potential acquirers and industrial partners often invest before revenue, which makes credible public positioning valuable years before there is anything to sell. Pillo Health's eventual acquisition by Stanley Black & Decker followed exactly this shape of relationship.
The practical consequence is that deep tech marketing serves three audiences at once, and only one of them is a customer. Blazon's deep tech agency practice is built around that, and the same launch mechanics carry over from the wider product launch agency work.
How deep tech companies reach market
The sequence that works, in short: pick a beachhead rather than a market, translate the technology into an outcome a non-expert can repeat, choose one verifiable proof point, then sequence announcements so each one earns the audience for the next.
Distribution is narrow and precise. Deep tech audiences are measured in hundreds: named accounts, specific job functions, technical communities, and a short list of sector journalists and analysts. Founder-led channels outperform brand advertising because buyers want to interrogate the claim with someone who can answer.
The full sequence is covered in the deep tech product launch guide and the deep tech go-to-market guide. For how far this diverges from software norms, see deep tech versus B2B SaaS marketing. Blazon runs this work through a dedicated deep tech agency practice, and the deep tech marketing cost guide covers what it costs. Blazon has worked on product launches since 2016, 500+ of them, with a minimum engagement across all agency services starting at $15,000.
How to evaluate a deep tech company from the outside
Investors, journalists, partners, and prospective employees all face the same problem: the underlying science is hard to assess without domain expertise. Four questions get most of the way there without it.
What has been demonstrated, and who verified it? Deep tech claims separate cleanly into simulated, demonstrated in a lab, demonstrated in the field, and certified by an outside body. A company that has cleared a regulator, passed an independent benchmark, or shipped to a paying customer is in a different category from one with a promising model, however good the model is.
What has to be true for this to work at scale? Almost every deep tech company depends on something outside its own control: a manufacturing process, an input cost curve, a regulatory decision, or a supply chain that does not exist yet. Founders who name those dependencies precisely tend to be further along than those who describe only the upside.
Who is the first customer, specifically? Naming one buyer with a budget line is a stronger signal than a large addressable market. Deep tech reaches revenue through narrow beachheads, so a vague first customer usually means the commercial work has not started.
What happens to the company if the timeline doubles? Long development cycles are normal in the category. What matters is whether funding, partnerships, and staffing survive a delay, because delays are close to certain.
None of these require understanding the underlying science. They test whether the company has translated the science into a commercial plan, which is the more common failure point.
FAQ
Is AI considered deep tech?
Only when the advance itself is scientific: novel model architectures, new training methods, specialised inference hardware, quantum computing, or AI validated against physical reality such as in robotics or scientific discovery. Applying an existing model through an API to a business workflow is AI-enabled software rather than deep tech, because there is no fundamental technical risk. The distinction matters because the two go to market in completely different ways.
What is the meaning of deep tech?
Deep tech describes companies whose product depends on a genuine scientific or engineering advance rather than on assembling existing software components. The defining characteristic is technical risk: the thing might not work, and proving that it does takes years, significant capital, and specialist expertise. Long timelines, high capital intensity, defensible advantages, and usually a regulatory path all follow from that starting point.
What is an example of deep tech?
Pillo Health built an FDA-cleared in-home robot that dispensed medication and managed adherence, combining medical regulation with consumer usability, and was acquired by Stanley Black & Decker. Other examples include fusion and grid-scale storage companies, low-carbon cement and novel battery chemistries, sequencing and lab-automation platforms, satellite and launch systems, and companies building new AI model architectures or inference silicon.
What is the difference between tech and deep tech?
Ordinary technology companies assemble known components into products, so their main risk is commercial: whether anyone buys it. Deep tech companies carry scientific or engineering risk on top of that, meaning the core technology might not work at all. That difference produces longer development timelines, heavier capital requirements, regulatory paths, defensibility based on difficulty rather than speed, and a go-to-market approach built on proof rather than persuasion.
How long does it take a deep tech company to reach revenue?
Five to fifteen years from research to meaningful revenue is typical, depending on the field and its regulatory path. Biotech and energy sit at the longer end because trials and pilot plants cannot be compressed. Robotics and advanced computing can be faster. The consequence for marketing is that most deep tech companies spend years needing credibility for recruiting, fundraising, and design-partner acquisition before they need demand generation.
Do deep tech startups need marketing before they have a product?
Usually yes, but for recruiting, fundraising, and finding design partners rather than for sales. Those three are genuine constraints years before revenue, and all of them respond to a credible public narrative. What pre-product marketing should not do is make claims that cannot yet be verified, because in a field where buyers validate everything technically, an unsupported claim spends credibility that is expensive to rebuild.
