AN-042 Build guides · pillar
How to Start a Hardware Company: From Idea to First Production Run
How to start a hardware company: the seven stages from first customer to first production run, with timelines, cost ranges and founders who shipped.
by the HardwareMap editors8 min read
Contents
- 1How to start a hardware company in seven stages
- 2Stage 1 to 3: customer, proof of concept, prototypes
- 3Stage 4: design for manufacturing and the bill of materials
- 4Stage 5 to 7: builds, factories and certification
- 5What it costs to start a hardware company
- 6What the index shows about founders who shipped
- 7Frequently asked questions
How to start a hardware company comes down to seven stages: find a customer who will pay, prove the core works, design it for a factory, run engineering and validation builds, certify, ship a small first run, then scale. For a connected consumer device that takes 18 to 30 months. This is the hub of the BUILD notes, linked from the hardware startups index.
HardwareMap (hardwaremap.com) is a hand-sorted directory of hardware startups building robots, rockets, reactors, chips and the tools to make them, catalogued like parts on a datasheet. This guide is written for the founder about to spend money: what each stage produces, what it costs, and where the index shows founders getting it right. Every price below is a dated range. Quote your own.
How to start a hardware company in seven stages
The stages overlap, but each one has a gate. Do not pay for the next stage until the current gate is passed.
Durations are for a connected consumer device with one or two custom enclosures. A robot arm or a medical device stretches stages 5 and 6. A dev board with no custom mechanics compresses them. The detailed build sequence lives in prototype to production.
Stage 1 to 3: customer, proof of concept, prototypes
Start with the buyer, not the board. Hardware punishes late discovery. A software team can ship the wrong feature and roll it back in a day. A hardware team that tooled the wrong enclosure has paid for steel. Before any spend beyond a bench prototype, write down who buys, at what price, through which channel, and how many units a year. If you cannot name the channel, you cannot price the product, because channel fees decide your margin.
Prove the hard part first. The proof of concept answers one question: does the core physics or chemistry work? A sensor that reads through skin, a gripper that holds an egg, a battery that survives 1,000 cycles. Off-the-shelf dev boards, breadboards and printed brackets are correct here. Spending on industrial design before the core works is the most common way to burn a seed round.
Split looks-like from works-like. The looks-like prototype is a model the size, weight and finish of the product, often printed and painted. It goes in front of customers and investors. The works-like prototype is the electronics and mechanism, often in a bigger box. Merging them too early forces compromises in both. The design for additive manufacturing note covers which print process fits each prototype.
The founding team
A hardware company needs four skills before it needs a factory: electrical, mechanical, firmware and someone who owns the supply chain. One founder can cover two. Nobody covers four. The gap most teams leave open is supply chain and manufacturing, because it does not show up in a demo. It shows up at EVT, when the first production-intent parts arrive out of tolerance and nobody on the team has read a first article inspection report. Hire or contract that skill before you cut tooling. The hardware engineer salary note has current bands for each role.
What to prototype with
Move one column right only when the left column has answered its question. A custom board before the circuit is proven is two weeks and a board run thrown away. A machined housing before the layout is fixed is worse.
Founders in the index show the pattern. Flipper Devices launched Flipper Zero through Kickstarter in 2020, so the first customers paid before production. Panic, a software company, co-designed its Playdate handheld with Teenage Engineering rather than building an industrial design team from zero.
Stage 4: design for manufacturing and the bill of materials
Design for manufacturing (DFM) means changing the design so the factory can make it repeatably at your target cost. It is where a prototype becomes a product. The rules are concrete: injection-molded walls between roughly 1 and 3.5 mm depending on resin, at least 0.5° of draft on vertical faces and 1° to 2° on most, ribs half the wall thickness, no undercuts you do not need. Protolabs publishes its injection molding design guidelines with wall thickness by material. The full rule set, with examples, is in design for manufacturing.
The bill of materials (BOM) is built in parallel. Every part gets a manufacturer part number, a second source where possible, a lead time and a price at your production volume, not at quantity one. The bill of materials example shows the columns. The manufacturing bill of materials adds assembly order, consumables and packaging. Once the BOM passes a few hundred lines, bill of materials software earns its seat cost.
Two decisions in this stage set your cost for years:
- Process per part. Machining, molding, sheet metal, casting or printing. Molding has the highest tooling cost and the lowest unit cost; printing the reverse. The break-even quantity decides it. The injection molding cost and injection molding tooling cost notes have the numbers.
- Board architecture. Fewer layers, standard component packages and parts in stock at the big distributors. The PCB assembly guide covers the line, PCB assembly cost the price, and prototype PCB assembly the fast-turn builds before DFM is done.
Framework is the clearest index example of a DFM decision that became a business model. Its 13-inch laptop has taken several generations of new mainboards that drop into older chassis, so the mechanical interface was designed to outlive any single board.
Stage 5 to 7: builds, factories and certification
Pick the factory before EVT, not after. A contract manufacturer (CM) quotes from your drawings, BOM and test plan, then builds engineering units on its own line. Choosing a CM is choosing a partner for years. Electronics contract manufacturer explains what a CM does; choosing a contract manufacturer is the scoring sheet. For boards alone, turnkey PCB assembly has the assembler buy the parts, and PCB assembly in the USA covers domestic options. In the index, Seeed Studio offers prototype PCB assembly out of Shenzhen, and Re:Build Manufacturing acquires American engineering and manufacturing firms and links them for design-to-volume work.
Run the build sequence. EVT (engineering validation test) proves the design works with production-intent parts. DVT (design validation test) proves it survives drop, thermal, humidity, life-cycle and certification pre-scans, using parts from production tooling. PVT (production validation test) proves the line can build it at rate with target yield. Each build is a few dozen to a few hundred units. Skipping one moves its failures into the field.
Certify early. Anything with a radio or a clock signal sold in the US falls under FCC Part 15. Products sold in the EU need CE marking under the directives that apply. Mains-powered and battery products often need a safety listing from a lab such as UL. Book pre-scans during DVT so a failed emissions test costs a board spin, not a launch date. Requirements depend on the product and market: confirm them with a test lab and counsel.
Ship small. The first production run should be big enough to reach your BOM price breaks and small enough that a design flaw does not sink the company. A few hundred to a few thousand units is typical for a consumer device. Measure returns and failure modes before ordering the second run.
Three things to have in writing before the first production purchase order:
- A golden sample. A signed-off unit the factory keeps on the line and builds to. Disputes about quality get settled against it.
- Acceptance criteria. The test fixture, the pass and fail limits, and the inspection sampling plan. Without them, "defective" means whatever the factory says it means.
- Ownership of tooling and test fixtures. You paid for the molds, so the contract should say you own them and can move them.
What it costs to start a hardware company
The money question has two halves: what it costs to reach the first production run, and what each unit costs after that. Rough budget lines for a connected consumer device with one molded enclosure:
Salaries for the team over 18 to 30 months sit on top and usually dominate. The line-by-line version, with what moves each number, is in hardware startup costs. The per-unit side, including freight, duty and channel fees, is in hardware unit economics. If those two notes do not agree, the business does not work.
Funding follows the stages. Preorders and crowdfunding can pay for tooling on a simple product. Seed rounds usually fund stages 2 to 5. Larger rounds fund inventory and scale. See hardware startup funding, how hardware startups raise seed, and the hardware accelerator programs that subsidize the prototype stages.
What the index shows about founders who shipped
The HardwareMap index lists 24 consumer hardware companies and 9 fabrication companies as of REV 26.09. A few patterns repeat across the ones that reached volume.
- Sell to people who can tolerate version one. Prusa Research was started by RepRap developer Josef Prusa and sold to makers who would assemble a kit. It now manufactures its printers in-house and runs its own filament production. The index records no disclosed outside rounds.
- Hire people who have shipped. Bambu Lab was founded in Shenzhen in 2020 by former DJI engineers. Production experience on the founding team shortens every stage from DFM to PVT.
- Finance the process, not just the product. Formlabs, an MIT Media Lab spinout from 2011, has $254 million in disclosed funding in the index. Printers, resin chemistry and powder systems each needed their own development and validation cycle.
- Make the first customer pay early. Kickstarter, preorders and paid pilots all move cash ahead of inventory. That is the cheapest capital a hardware company will ever raise.
Nothing in this list is a guarantee. Many funded companies never ship. The index records the ones with a product, a public source and an operating team, and marks status changes each REV.
Frequently asked questions
How long does it take to start a hardware company and ship a product?
For a connected consumer device, plan on 18 to 30 months from first working prototype to units in customers' hands. Engineering and validation builds take most of that time. Tooling lead times, certification lab queues and component allocation each add weeks you cannot compress with money. Industrial and medical products usually take longer because qualification and regulatory review sit on the critical path.
How much money do you need to start a hardware company?
A simple electronic product with one molded enclosure can reach a first run of a thousand units on a budget in the high six figures. A connected device with custom mechanics and radio certification usually needs one to three million dollars. Robots, vehicles and medical devices need far more. The hardware startup costs note breaks the budget down line by line.
Can you start a hardware company without venture capital?
Yes. Prusa Research and Adafruit Industries have no disclosed outside rounds in the HardwareMap index, and both sell to makers who pay before or soon after shipment. Crowdfunding and preorders can fund tooling for a simple product. The trade is speed: without outside capital you build fewer units per batch and grow from margin, which works best when gross margin is high.
Should a hardware startup manufacture in-house or use a contract manufacturer?
Most first-time founders should use a contract manufacturer for electronics assembly and a molder or machine shop for mechanical parts. In-house production makes sense when the process is the product, as with additive manufacturing companies, or when volumes are low and assembly is skilled. Prusa Research builds in-house; most consumer device startups do not. Decide on cost per unit, quality control and cash.
What is the biggest mistake first-time hardware founders make?
Tooling a design that is not finished. Changing a hardened steel mold after cutting costs weeks and thousands of dollars, and every change ripples through certification and packaging. The second most common mistake is pricing from bill of materials cost instead of landed cost, which hides freight, duty, returns and channel fees until the first production invoice arrives.
HardwareMap catalogues the companies that made it through these stages. Building one? Submit it and get a part number.
Everything in build guides17 notes
- AN-043Prototype to Production: The Stages Nobody Draws Correctly
- AN-044Design for Manufacturing (DFM): The Working Guide
- AN-045Design for Additive Manufacturing (DfAM)
- AN-047Bill of Materials Example (With a Free Template)
- AN-048Manufacturing Bill of Materials (MBOM) vs Engineering BOM
- AN-049Bill of Materials Software, Compared (From Spreadsheets to PLM)
- AN-052PCB Assembly: The Complete Guide for Hardware Startups
- AN-053PCB Assembly Cost: What Drives the Quote
- AN-054Prototype and Low-Volume PCB Assembly: Where to Go, What to Expect
- AN-055Turnkey vs Consigned PCB Assembly (and Quick-Turn Options)
- AN-056PCB Assembly in the USA: When Domestic Makes Sense
- AN-060Electronics Contract Manufacturers: How to Pick One
- AN-064Injection Molding Cost: Parts, Tooling, and the Volume Curve
- AN-065Injection Molding Tooling Cost: Why the Mold Is the Real Bill
- AN-067Hardware Startup Costs: What the First 18 Months Really Cost
- AN-068Hardware Unit Economics: COGS, Margin, and the Landed-Cost Trap
- AN-069How to Choose a Contract Manufacturer: The 15-Question Audit
Building one?
Submit it to the index and get a part number. Every entry is reviewed by hand.