A hardware product costs between roughly $75,000 and $300,000 to develop and reach a first production run through an outside engineering firm, on the figures published across the industry and in our own experience quoting these engagements, with simple non-electronic products landing below that and anything cellular, medical or first-of-a-kind landing well above it. The spread is that wide because four independent variables (whether it has a radio, whether it has a battery, how complex the enclosure is, and whether it enters a regulated market) each move the number by tens of thousands of dollars.
Thank you for reading this post, don't forget to subscribe!Most published estimates answer a different question than the one you are asking. They budget a founder building it themselves, buying their own dev boards and doing their own CAD. This guide budgets an outsourced engagement, and structures every number the way it appears on a real quote.
What a hardware product actually costs, end to end
The distinction that makes a budget legible is the one between money you spend once and money you spend on every unit forever. Get that wrong and your spreadsheet will be out by a factor of ten in one direction or the other.
The four cost buckets
| Bucket | What it covers | Paid |
|---|---|---|
| NRE (non-recurring engineering) | Design labor across every discipline, plus supplier setup: engineering review, CAM setup, fixtures, stencils, test programming | Once |
| Tooling | Injection molds, jigs, assembly fixtures, test rigs | Once per tool, plus modifications |
| Certification | Lab testing, filings, and in some schemes an annual fee that never stops | Once, then again after design changes |
| Unit cost | Bill of materials, assembly, test, packaging, freight and duties | Every unit, forever |
One trap inside NRE is worth knowing before you sign anything. Paying a supplier’s NRE line does not automatically give you transferable ownership of the tooling, fixtures or CAM data it produced. Founders routinely assume that a paid setup fee means they own the stencil and the fixtures and can move them to another factory. That is a contract term, not a default.
Why published ranges are so wide
Two real examples of the same job priced very differently, both from compliance firms selling the service:
- One puts a wireless CE marking job at roughly €1,500 to €3,000
- Another puts the equivalent Radio Equipment Directive work at €10,000 to €60,000
Neither is wrong. The first prices a simple consumer device built on a pre-certified module using harmonized standards and self-declaration. The second prices a complex multi-radio product that needs third-party involvement. Same regulation, same paperwork, twenty times the cost, because the products are not comparable.
Read every published range as a statement about a product you have not seen.
Cost by stage, from idea to first production run
Proof of concept and feasibility
The cheapest stage and the one that saves the most money. Feasibility work runs from about $1,500 for a narrow question to $15,000 or more for a full assessment covering technical, regulatory and cost viability. What you are buying is the answer to whether the physics, the certification path and the target price can coexist. Our guide on product feasibility analysis covers what a proper one includes.
Skipping this stage is how projects discover in month nine that the battery cannot fit, and that discovery costs six figures rather than four.
Prototype and engineering builds
Prototypes split into two kinds that get confused constantly, and the difference between them is the difference between a prototype and an MVP.
- Looks-like models prove form, size and feel. Foam mock-ups start around $100. A finished appearance model runs from roughly $2,000 to $150,000 depending on finish quality.
- Works-like prototypes prove function, and carry your first custom circuit board. Published ranges run from about $2,000 to $250,000, and the spread is almost entirely about how much custom electronics is involved.
- 3D printed parts cost roughly $100 to $1,000 per iteration. Urethane casting costs $250 to $1,500 for the mold and $20 to $50 per part after that, which is why it beats printing past about twenty units. See our guide on how 3D printing is transforming product development for where each process stops making sense.
EVT, DVT and PVT
Nobody publishes credible dollar figures per validation gate, which is itself worth knowing when you read a guide that offers them. What is well documented is the shape of each stage, and you can construct your own number from unit counts and build cost.
| Gate | What it proves | Units | Duration |
|---|---|---|---|
| EVT (engineering validation) | The design works | Tens | 6 to 10 weeks |
| DVT (design validation) | It survives the real world and passes certification | Dozens to hundreds | 8 to 14 weeks |
| PVT (production validation) | The factory line can build it repeatably at yield | A pilot run | 4 to 8 weeks |
Build size is set by the risks each gate has to uncover, not by a fixed count. The way to build your own estimate is unit cost at that volume, multiplied by units, plus the engineering time to analyze the results and respin. The third term is the one founders leave out, and it is usually the largest. Our prototype testing checklist sets out what actually gets tested at each gate.
Tooling and pilot production
Injection mold tooling is the single largest step change in a hardware budget, and the number depends heavily on where the tool is cut. Published figures from a Chinese molder:
| Tool type | Cost | Expected life |
|---|---|---|
| Prototype or soft mold, single cavity | $1,000 to $3,000 | 10,000 to 50,000 shots |
| Bridge mold, 1 to 2 cavities | $3,000 to $5,000 | 50,000 to 200,000 shots |
| Production, simple, 2 to 4 cavities | $5,000 to $10,000 | 300,000 to 500,000 shots |
| Production, complex, 4 to 8 cavities | $10,000 to $30,000 | 500,000 to 1,000,000+ shots |
| High precision or medical | $30,000 to $100,000 | 1,000,000+ shots |
Every figure in this section is a supplier’s own price list rather than a market survey, so treat them as the shape of the number and get your own quotes. The same molder reports its pricing runs 40% to 60% below Western equivalents, which is a self-interested claim and not an implausible one. Two cost drivers move these numbers more than anything else:
- Cavity count. Each added cavity adds somewhere between 15% and 50% to the tool, and the right count is set by your annual volume, not your ambition.
- Side actions. Every lifter or slider needed to form an undercut adds roughly $1,500 to $8,000, or 10% to 40% of the mold cost. A part with several can cost more than half again what a simple one costs.
Lead times are the part that wrecks schedules. Standard production tooling runs 45 to 65 days from order to first sample shots. A small four-cavity ABS part takes about five weeks. A large single-cavity part takes eight to nine. Expedite fees add 20% to 35% and buy back only a couple of weeks.
You can see how tooling and enclosure decisions play out on a real product in our smart brush cabinet project.
Cost by engineering discipline
Start from what the work is worth, because it explains the rates you will be quoted. US Bureau of Labor Statistics data for May 2025 puts the median wage for electrical and electronics engineers at $125,040 a year, about $60.12 an hour, and mechanical engineers at $104,110.
Design and engineering firms bill between roughly $100 and $450 an hour. That runs from a little under twice the underlying hourly wage to more than seven times it, and the multiple is not margin alone. It covers benefits, equipment, software licenses, unbilled time, project management and the fact that a firm carries several disciplines you would otherwise have to hire and coordinate yourself.
| Discipline | Typical engagement | What moves it |
|---|---|---|
| Electronics and PCB | $25,000 to $75,000, up to $200,000+ | Custom radio, high speed, layer count, safety isolation |
| Firmware and embedded | $10,000 to $50,000 | Connectivity stack, security requirements, cloud integration |
| Mechanical engineering | $25,000 to $60,000, up to $120,000+ | Sealing, moving parts, thermal, part count |
| Industrial design | $15,000 to $30,000, up to $60,000+ | Finish quality, number of concepts, user research |
Offshore engineering runs about $20 to $25 an hour against the $100 to $450 band above. The rate gap is real and so is the coordination cost, and the projects that go badly wrong offshore are usually the ones where nobody owned integration. Firmware and electronics work is covered in more depth on our PCB and embedded system design page.
Certification, itemized
Certification is where founders find the largest gap between what they budgeted and what arrives, mostly because they budget government fees and pay private labs.
The FCC charges almost nothing. Under 47 CFR 1.1103, the FCC’s only material equipment-authorization charge is a one-time $35 fee for a grantee code, plus $50 if you file a confidentiality request. The testing is performed by private accredited labs and Telecommunication Certification Bodies, and that is where the money goes. Any guide quoting FCC certification fees in the hundreds or thousands is citing a schedule the FCC removed in its 2020 application-fee order. Certification moved to accredited third-party bodies in stages between 1999 and 2014.
| Item | Typical cost |
|---|---|
| FCC Part 15B, unintentional radiator, self-declaration | $1,500 to $3,500 |
| EN 55032 and EN 55035 for CE marking | $3,000 to $6,000 |
| Connected device using a pre-certified radio module | $3,000 to $10,000 |
| Custom single-band radio, chip-down | $8,000 to $20,000 |
| Medical EMC, IEC 60601-1-2 | $8,000 to $18,000 |
| Cellular, including carrier approval | $50,000 to $200,000+ |
| UN 38.3 lithium battery transport testing | From about $1,700 per model at Asian labs |
| ISTA 3A packaging and drop testing | $850 to $1,250 per test cycle |
Three things that get missed:
- Bluetooth is a separate bill, and a module does not cover it. The Bluetooth SIG restructured dues and fees effective 1 March 2026. Product qualification at Adopter level (free membership) is $12,000. A Contributing Adopter tier at $3,500 a year discounts a first qualification to $8,000, which beats the Adopter route on your very first product, and Associate membership discounts every qualification to $6,000, which pays off across a portfolio. Later products reusing the same qualified design carry no further fee. Using a pre-certified radio module cuts your FCC testing cost, and your end product still needs its own Bluetooth qualification under your own membership. Check the current Bluetooth qualification fees before budgeting, because this schedule has moved twice in three years.
- UL is an annual cost, not a one-off. UL Solutions does not publish prices. Its published fee structure includes an annual data maintenance fee, a quarterly-billed certification fee for ongoing use of the mark, and follow-up services with onsite factory inspections. Your budget needs a recurring line.
- FDA filing fees are public and large. For fiscal year 2026, a 510(k) submission costs $26,067, or $6,517 for a qualifying small business. A De Novo is $173,782. Annual establishment registration is $11,423 regardless of size. Those are filing fees only, before any testing or quality system work.
Budget a retest. EMC test labs commonly report that around half the products they see fail on the first attempt. No authoritative dataset sits behind that figure, and a failure still adds roughly $2,000 to $10,000 and four to twelve weeks whenever it happens. Pre-compliance scanning at $3,000 to $8,000 is the cheapest insurance in the whole program.
What drives your number up or down
- A radio. The difference between a self-declared unintentional radiator and a custom chip-down radio is roughly $1,500 to $3,500 against $8,000 to $20,000, and Bluetooth adds its own qualification fee on top. Using a pre-certified module is the single largest cost lever available to you.
- A battery. Lithium cells trigger UN 38.3 transport testing, which consumes samples destructively. Mains power triggers safety certification and its annual fees.
- Enclosure complexity. Undercuts drive side actions, and side actions drive tooling. A design review before the tool is cut typically pays for itself several times over.
- A regulated market. Medical, children’s products and anything touching skin or measuring a vital sign change the cost class entirely.
- Volume. NRE amortization is brutal at low volume. A $600 assembly setup fee adds $120 per board across 5 units and $1.20 across 500.
- First-of-a-kind versus reference design. Building on an existing reference design can halve the electronics engineering. Inventing something genuinely new cannot be shortcut.
- Whether it works first time. This is the variable nobody budgets, and the next section is about it.
Costs founders routinely leave out of the budget
The forced redesign. A change after the design is frozen is the most expensive kind, because it lands on work already paid for. It arrives as engineering rework, then re-verification and compliance retesting on top, then whatever the schedule slip costs commercially. A single component going end-of-life late in a program can trigger all three at once.
Component obsolescence. Component sourcing firms report semiconductor lifecycles compressing from over ten years before 2020 to under five for consumer-grade parts, and put a six-layer board respin at $12,000 to $45,000 in engineering alone. Those are trade estimates rather than measured data, and the direction is not in dispute. A product designed today can face a forced redesign before it turns five, and a meaningful share of end-of-life notices never reach every affected customer.
Everything else that lands after the quote:
- Tooling modification after the first shots: $300 to $8,000 per change
- Certification retest following a design change, plus the schedule it costs
- Sustaining engineering and firmware maintenance, commonly 15% to 20% of build cost per year
- Cloud and connectivity for a connected product, as an ongoing per-device cost
- Packaging design, retail packaging, and the drop testing that validates it
- Safety stock, and the working capital tied up in it
- Freight, customs and duties. Rates on electronics have moved repeatedly, so check the current position with USTR and CBP at the time you import rather than trusting any figure published earlier
Building it yourself versus budgeting an outsourced engagement
Most cost guides online budget a DIY build. That is a legitimate path and a completely different number, because the largest line in an outsourced quote is labor you are proposing to supply for free.
| Building it yourself | Outsourced engagement | |
|---|---|---|
| Engineering labor | Your time, uncosted | The largest line on the quote |
| Integration risk | Yours | Contracted, with a named owner |
| Certification strategy | Learned during the process | Designed in from the schematic review |
| Design for manufacture | Discovered at the factory | Applied before the tool is cut |
| Timeline | Elastic | Contracted, with a schedule attached |
| What goes wrong | Rework you pay for twice | Scope disputes, and cost when requirements move |
The honest comparison is not price against price. It is a lower cash number with your time and your risk absorbed invisibly, against a higher cash number with the risk transferred and the schedule enforceable. Founders who compare only the cash numbers consistently underestimate the first column by the value of a year of their own work.
How to turn this into a number you can raise against
Investors do not fund a range. They fund a plan with named line items and a stated point of proof.
- Fix the specification first. Every unresolved requirement is an unpriced risk, and a firm quoting an unfixed spec is quoting a guess.
- Price the stage, not the product. Raise to a defined gate (working prototype, or EVT complete) rather than to “launch.” Gates are estimable and launches are not.
- Separate one-time from per-unit. Show NRE, tooling and certification as capital, and unit cost as the thing that determines whether the business works.
- Carry contingency explicitly. A 20% to 30% contingency is a realistic plan rather than a padded one, and you can justify it from named lines: an EMC retest at $2,000 to $10,000, tooling modification at $300 to $8,000 per change, and the engineering time to analyze a failure and respin. Showing it makes a budget more credible, not less.
- Name what you have not priced. Duties, sustaining engineering and obsolescence belong in the plan even before you can size them.
Frequently asked questions
How much does it cost to develop a hardware product?
Roughly $75,000 to $300,000 through an outside firm for a connected consumer product taken to a first production run. Simple non-electronic products can come in at $20,000 to $60,000, and cellular, medical or first-of-a-kind products commonly exceed $500,000.
What is NRE and why is it a separate line?
Non-recurring engineering is one-time design and setup cost that is not repeated per unit: engineering labor, CAM setup, fixtures, stencils and test programming. It is separated because it amortizes across volume, and at low volume it dominates your unit cost.
Why is certification so much more expensive for a wireless product?
An unintentional radiator can often self-declare after $1,500 to $3,500 of testing. A custom radio needs full certification through an accredited body at $8,000 to $20,000. A pre-certified module avoids most of that difference on the FCC side, though a Bluetooth product still needs its own qualification with the Bluetooth SIG.
How much should I budget for injection mold tooling?
A simple two-to-four cavity production mold runs about $5,000 to $10,000 in Asia and considerably more domestically. Complexity drives it: every side action adds $1,500 to $8,000, and each added cavity adds 15% to 50%.
What do most founders forget to budget?
The second attempt. Retesting after a failed EMC scan, tooling modifications after first shots, and the engineering time to analyze test results and respin. EMC labs commonly report around half the products they see failing first time, and that is a schedule cost as well as a cash one.
Is it cheaper to manufacture offshore?
Tooling and assembly usually are. Asian molders commonly claim 40% to 60% below Western pricing, which is a self-interested figure and not an implausible one. Freight, duties, travel, communication overhead and the cost of a problem discovered at distance narrow the gap. Compare landed cost against domestic, not factory-gate price.
Where Inventornest fits
Inventornest quotes hardware engagements as line items rather than a single number, because a budget you cannot interrogate is a budget you cannot defend to an investor. That covers feasibility, industrial design, mechanical, electronics and firmware, through certification support and design transfer to manufacturing. Our OEM services page sets out the full scope, and product design and prototyping covers the earlier stages on their own.
If you need a real number for your own product rather than a range for a category, get a quote and we will break it down by stage.