Most branding advice was written for software companies selling to other software companies. Ship fast, iterate on messaging, A/B test the tagline. None of that maps cleanly onto a company building fusion reactors, photonic chips, or autonomous industrial robots.
Deeptech branding is the work of building brand systems for companies commercializing breakthrough science, where the brand must earn credibility with scientists, industrial buyers and investors at once.
Deeptech and hardware startups operate under a different set of constraints: R&D cycles measured in years, claims that sound impossible until they're demonstrated, and three audiences (scientific talent, industrial customers, and investors) who each read your brand through a completely different lens. A brand built for one of them often actively repels the other two.
This article covers what changes when the product is a genuine technical breakthrough rather than a workflow tool, and how to build a brand that holds up in front of a PhD, a procurement director, and a Series B partner in the same week. It's the thinking behind our deeptech and AI design practice, condensed.
The credibility paradox: your best claims are your biggest liability
Here is the core problem. If your technology is genuinely 10x better than the state of the art, saying so plainly makes you sound like every failed vaporware company that came before you. The more impressive the claim, the more skepticism it triggers.
Software startups rarely face this. Nobody doubts that a new CRM exists. But "we've achieved net energy gain," "our chip runs inference at a tenth of the power," or "our robot navigates unstructured environments autonomously", these are claims the audience has heard before, from companies that never delivered. Theranos poisoned the well for every hard-science startup that followed.
The instinctive response is to pile on superlatives and futuristic visuals: glowing particles, dark gradients, orbital lines. This backfires. Sophisticated audiences (and every audience that matters to a deeptech company is sophisticated) pattern-match sci-fi aesthetics to companies compensating for weak fundamentals.
The counterintuitive move is restraint. Look at how the credible players do it:
- Commonwealth Fusion Systems talks about fusion (arguably the most over-promised technology in history) with the visual language of an engineering firm, not a sci-fi film. Photography of actual magnets, actual facilities, actual people in hard hats. Milestones stated as engineering facts with dates attached.
- Boston Dynamics barely writes marketing copy at all. Their brand is built almost entirely on unedited-looking demonstration footage. The robot doing the thing is the brand. No claim is made that a video doesn't back up.
- Anduril made a deliberate aesthetic bet: defense technology presented with the confidence and production quality of a consumer brand, but grounded in named, specified products. The polish signals seriousness of intent; the specificity signals substance.
- Mistral AI entered a market saturated with AI hype and differentiated by publishing weights, benchmarks, and technical documentation as primary brand assets. In a field where everyone claims state-of-the-art, showing the receipts became the positioning.
The pattern across all four: evidence carries the claim, and the brand's job is to frame the evidence, not replace it. That's a fundamentally different brief than most agencies are used to, and it's why generic startup branding often fails deeptech companies. We wrote more broadly about building trust in skeptical markets in our piece on how fintech brands earn credibility; the mechanics for deeptech are stricter still. If you want the worked examples rather than the method, ten fintech brands and the trust problem each one solved show the same evidence-first logic applied to money instead of physics.
Three audiences, one brand, zero room for error
A B2C brand optimizes for one buyer. A deeptech brand has to work simultaneously for:
1. Scientific and engineering talent. Your hiring pool consists of people who can evaluate your technical claims directly, and will. A brand that oversimplifies or overhypes reads as a red flag to the exact PhD you're trying to recruit away from a national lab. This audience wants to see the hard problem stated honestly, the team's credentials, and evidence that the company respects the science.
2. Industrial customers. Procurement at an aerospace prime, a utility, or an automotive OEM doesn't buy vision. They buy risk reduction. The committee dynamics here are close to those of software sold to an enterprise buying committee, with one difference that changes everything: the technical evaluator can verify your claims themselves. Their unspoken question is: will this startup still exist in five years, and will this component pass qualification? For them, the brand needs to signal operational maturity (certifications, standards compliance, manufacturing partnerships) even when the company is thirty people in a lab.
3. Investors. Deeptech investors underwrite scientific risk, but they still need a narrative they can defend to their LPs and repeat in their own words. For them, the brand must compress years of research into a thesis: what becomes possible if this works, and why this team wins.
The tension is real. The talent audience punishes hype; the investor audience needs ambition; the industrial customer needs boring reliability. Most deeptech brands fail by picking one audience and alienating the others, usually by writing everything for investors and ending up with a site their own engineers are embarrassed to share.
The resolution isn't three separate messages. It's a layered architecture: a plain-language claim at the top (for investors and generalists), the mechanism one level down (for customers evaluating fit), and the technical depth (papers, benchmarks, specs) one level below that (for scientists and engineers). Each audience self-selects its depth. The brand's tone stays constant across all three layers; only the resolution changes.
Making the invisible visible
Software startups screenshot their product. What do you show when your product is a catalyst, an algorithm, a novel material, or a machine that lives inside a factory?
This is where deeptech branding becomes a genuine design problem rather than a messaging problem.
Diagrams as first-class brand assets. In deeptech, the system diagram is often the single most-viewed piece of communication. It appears in the deck, on the website, in sales conversations, in press coverage. Yet most companies treat it as an engineering afterthought exported from a whiteboard tool. Investing in a rigorous, owned diagrammatic language (consistent line weights, a defined color logic, real information hierarchy) pays back across every channel. When journalists and analysts reuse your diagram to explain your category, you've won the framing war.
Hardware photography is not product photography. E-commerce conventions (white background, three-quarter view) make industrial hardware look like a catalog part, interchangeable and cheap. The companies that get this right shoot hardware the way architecture is shot: context, scale, texture, evidence of precision. A macro shot of a machined surface communicates manufacturing quality no paragraph can. Boston Dynamics' entire communication strategy is arguably a photography and videography decision.
Visualizing the invisible layer. For companies whose core IP is a process or a material property, abstraction is unavoidable, but it should be specific abstraction, derived from the actual physics or data, not generic tech-glow. A visual system built from your real waveforms, lattice structures, or sensor outputs is both more distinctive and more defensible than stock 3D renders. This is the approach we take in brand identity work for technical companies: the science supplies the raw visual material; design gives it discipline.
Naming in a jargon-saturated space
Deeptech naming pulls in two bad directions. One is the descriptive-technical trap: names built from morphemes like quant-, syn-, -onics, -genix that are legally clearable precisely because they're forgettable. The other is the aggressively abstract consumer-style name that gives an industrial buyer nothing to hold onto.
A few practical observations from the field:
- Names borrowed from adjacent intellectual traditions age well. Anduril (Tolkien), Palantir (Tolkien again), Atlas and Spot (Boston Dynamics' plain-English robot names) all avoid tech-morpheme soup while carrying meaning for the people meant to notice.
- Product names matter more in hardware than in SaaS. A SaaS company is one product; a hardware company ships generations and variants. You need a naming system: how do versions, models, and configurations relate?, before you need a clever name. CFS's "SPARC" and "ARC" reactor names do double duty: memorable, and encoding a roadmap.
- Check the acronym landscape in your subfield first. Every technical domain has overloaded terms. A name that collides with a standard acronym in your customers' industry will lose the search battle inside their own documentation, which is where purchase decisions actually get written down.
The investor angle: branding for a decade-long story
Deeptech fundraising has a structural problem: the milestones that matter (first plasma, tape-out, flight test, FDA clearance) are years apart, but you're raising every 18–24 months. Between technical milestones, the narrative has to carry the round.
This is why investor materials for deeptech companies are a different discipline from SaaS decks. A SaaS deck argues from traction, an up-and-to-the-right chart does the work, and the standard twelve-slide pitch deck structure, taken slide by slide assumes that argument is available to you. A deeptech deck argues from inevitability: the physics works, the team is the best in the world at this specific problem, the milestones are de-risking a defined sequence, and the market at the end is enormous. Design's job is to make that logical chain feel as concrete as a revenue chart.
Concretely, the strongest deeptech decks we see share three traits: a technology-readiness narrative (where you are on the path from lab to product, stated honestly), evidence hierarchy (data and third-party validation given more visual weight than vision statements), and continuity, the Series B deck should visibly descend from the seed deck, because investors track whether you did what you said. Knowing what a Series A deck has to prove that a seed deck did not is what makes that descent legible rather than repetitive. That continuity is itself a brand asset, and it's a core part of how we approach investor deck design.
For a sense of how public perception of deeptech capital has shifted, Boston Consulting Group's research on deep tech investing and Dealroom's annual European Deep Tech Report are useful reference points, both document the growing expectation that deeptech companies communicate commercial maturity earlier than they used to.
What this means in practice
If you're a technical founder reading this, the summary is uncomfortable but useful: your instinct to under-invest in brand ("the technology speaks for itself") and the generic agency instinct to over-style it ("make it look like the future") are both wrong, and wrong in opposite directions.
The technology does not speak for itself. It speaks through demonstrations, diagrams, photography, documentation, and a naming system, all of which are design decisions someone will make either deliberately or by default. And the future-glow aesthetic actively costs you credibility with all three audiences that decide your outcome.
What works is a brand built the way you built the product: from evidence, with discipline, layer by layer. We've applied this thinking in practice, including our work with Reecall on an AI technical product. Not every studio briefs this way, so it is worth reading how the leading startup branding agencies differ in approach before you hand a technical brief to a generalist. If you're heading into a raise or a commercial launch and the brand hasn't kept pace with the science, start a project with us.
