By sector
Deep tech & hardware diligence
Technical risk you cannot assess alone, capital intensity that dwarfs software, and the gap between a working prototype and a manufacturable product.
Realistic time Twelve to twenty-five hours, including an expert call
Deep tech and hardware break the assumptions that make angel investing manageable. The capital required is larger, the time to revenue is longer, the technical risk is genuinely beyond most investors' competence, and the failure mode is specific: the science works and the company still cannot make the thing at a cost anyone will pay.
The honest starting position is that you cannot assess the technology yourself. This is not a failure of diligence — it is a fact about specialist fields, and the appropriate response is to find someone who can. An hour with a genuine expert in the field is worth more than twenty hours of your own reading, and experts are more willing to take such calls than most angels expect.
The second thing to internalise is the capital path. A hardware company that needs $60m to reach production is a company where an early angel will be diluted enormously and where the exit must be very large to matter. That arithmetic should be done before the technical assessment, because it often settles the question.
01
Technical risk and how to assess it
Establish what is genuinely unproven. In most deep tech companies some parts are settled science and one or two are not, and the company's value depends entirely on the unproven parts working at scale. Founders are usually clear about this when asked directly.
Then find an expert. Academic researchers in adjacent fields, engineers who have worked in the industry, and technical people in your own network who know somebody are all reachable. The question to ask them is narrow: is this approach plausible, what is the hard part, and what would you want to see demonstrated.
Distinguish between a demonstration and a product. A laboratory result under controlled conditions, a prototype that works when an engineer is present, a pilot unit running at a customer site, and a manufactured product are four very different stages, and the gaps between them are where most hardware companies fail.
Check
- Which specific technical claims are unproven at scale?
- Find and speak to one independent expert in the field.
- What stage is the technology — lab, prototype, pilot, production?
- What has been demonstrated under real conditions rather than controlled ones?
- What is the single hardest remaining technical problem?
02
Intellectual property
Patents matter far more here than in software. Establish what is granted rather than filed, in which jurisdictions, and whether the company owns it outright or licenses it from a university.
University licences deserve particular attention. Terms vary enormously, and a licence with high royalties, field-of-use restrictions or diligence obligations can constrain the company permanently. Ask to see the licence rather than a description of it.
Freedom to operate is the other half. A company can hold patents and still infringe someone else's, and in crowded technical fields that risk is real. Ask whether a freedom-to-operate analysis has been done and by whom.
Check
- Granted patents versus applications, and in which jurisdictions.
- Is IP owned outright or licensed from a university or institution?
- Read the licence terms: royalties, field of use, diligence obligations.
- Has a freedom-to-operate analysis been performed?
- Were any inventors employed elsewhere when the work was done?
03
Manufacturing and cost
The characteristic hardware failure is a product that works and cannot be made economically. The bill of materials at prototype volume is always far above the target, and the path from one to the other depends on volume, redesign and supplier negotiation — each of which takes time and capital.
Ask for the current bill of materials, the target, and the specific steps between them. Founders with manufacturing experience will have this as a document. Founders without it will describe economies of scale in general terms, which is the answer that should worry you.
Supply chain concentration is the other structural risk. A component with one qualified supplier, a long lead time, or an export-control status is a dependency that can stop production entirely.
Check
- Current bill of materials, target cost, and the plan between them.
- What volume is required to reach the target cost?
- Which components have a single qualified supplier?
- Lead times on the longest-lead components.
- Any export controls or regulatory approvals on components or the product?
- Who on the team has manufactured at volume before?
04
Capital path and dilution
Before anything else, model the capital required to reach revenue and what it does to an early position. A company needing three more rounds totalling $80m will dilute a seed investor to a small fraction, and the exit must be correspondingly large.
Non-dilutive funding changes this arithmetic materially. Grants, government programmes and research funding are far more available in deep tech than in software, and a company that has won competitive grants has both money and third-party validation.
Check
- Total capital required to reach revenue, and to reach profitability.
- Model your position after the expected rounds.
- What non-dilutive funding has been secured or applied for?
- What is the realistic exit universe, and who are the buyers?
- How long until first revenue, honestly?
Stop and think
Red flags
- A technical claim that no independent expert will confirm as plausible.
- University licence terms that have not been shown to investors.
- A bill of materials far above target with no specific plan to close the gap.
- Nobody on the team who has taken a physical product to volume production.
- A capital path requiring far more money than the founders acknowledge.
- Patents described as filed and presented as though granted.
- A single-source component with a long lead time and no qualified alternative.
Take these into the room
Questions to ask the founders
- Which part of this is genuinely unproven?
- Who is the most credible sceptic of this approach, and what do they say?
- What is the bill of materials today, what is the target, and how do you get there?
- How much capital does this need before first revenue?
- Who on the team has manufactured at volume?
- What does your university licence require of you?
- What would make you abandon this approach?
From the decoder