| Episodes drawn on | 283 |
| Cited here | 69 |
| Claims extracted | 119 |
| Gathering | concept union |
| Other themes | |
A hardware startup is a small company built around a physical product, and it differs from a software company chiefly in being cash-heavy rather than cash-light: parts must be bought before anything can be sold, and the software startup model most hardware founders inherit does not account for that.[509] A first team of fifteen people frequently spans eight to ten distinct engineering disciplines that have to be coordinated effectively, on top of product, platform, distribution, supply chain, logistics and regulatory compliance.[327][495] Depending on the survey, between sixty and eighty percent of conventional software startups fail, and hardware startups carry additional risk on top of that baseline.[495] The internal experience of such a company is therefore dominated by three things: how much money is in the bank, how many disciplines each person is covering, and how much of the company the founders still control.[327][509]
Founders and the transition into the role
Hardware founders are rarely straight out of school; most have prior industry experience, in contrast to the young-founder pattern common in software.[166] Studies of founders find that successful ones skew older than the popular image, with roughly twice as many over 50 as under 25, and twice as many over 60 as under 20.[83]
The characteristic barrier between engineering and entrepreneurship is an engineer’s unwillingness to sell a product he judges unfinished; a business-minded founder ships through that barrier and raises money against it.[325] The move out of employment takes several forms. One founder went part-time and then quit before the product generated any income, intending to fall back on consulting if it failed — a staged exit that limits risk without waiting for revenue.[237] Another began involuntarily, asking for a raise and being let go on the spot, then finding that a prospective employer preferred to hire her on contract; she quit, bought a house and started the business at the same time.[424] Founding a business specifically to eliminate a task one dislikes is self-defeating, because the founder ends up doing that task continuously as the core of the business.[131]
A technical founding team is generally advised to recruit a capable business person early to incorporate the company, handle the legal work, identify the target market and take on fundraising.[407] The underlying trade is a direct one between roles: someone who wants to keep doing engineering does not also get to be the person raising money, and someone who wants to run the company takes fundraising on as the job.[267] Raising money is not a side activity — founders who have secured funding typically describe roughly twelve months of full-time effort spent on it.[267]
Legal structure is decided at the outset and is hard to undo. One hardware founder identified incorporating as a corporation with a board and shareholders, rather than forming an LLC, as her single largest early mistake.[475]
The decision to take outside money
Bootstrapping means starting with a small amount of one’s own money, selling a few units, and reinvesting the profit to build more, growing with no outside investment capital.[24] The loop is concrete: buy parts for a hundred kits from savings, sell the hundred, use the few thousand dollars of proceeds to buy more parts, and repeat.[264] Its failure rate is very low, because the bootstrapper only risks what has already been earned, against a very high failure rate for a single concentrated venture-funded bet.[24] Unless the product is unusually capital-intensive, starting a hardware business costs thousands rather than hundreds of thousands of dollars, which makes it feasible to work three or four years after graduating, save on the order of fifty thousand dollars, and fund a year of full-time work from savings.[94] Wayne and Layne began in autumn 2009 with each founder contributing about $300 to build a hundred units of a single kit, an outlay small enough that failure would merely leave a lifetime supply of parts; the company was later selling tens of thousands of its Blinkies units to retailers.[167]
What bootstrapping cannot buy is a step change in scale. MakerBot’s move from a 5,000 to a 50,000 square foot facility was funded by venture investment rather than retained profit, and incremental reinvestment does not deliver an order-of-magnitude jump in production capacity.[151] Entering medium-volume consumer hardware typically requires between two and ten million dollars, which in practice comes from venture investors who expect defensible intellectual property — a direct tension for companies that want to be open source.[113] Scaling to production requires capital on two distinct fronts, cash to buy the bill of materials and the continuing cost of engineering salaries, so a raise sized only against parts underestimates the requirement.[340] A software-for-hardware company that ran three years at an average of seven people with effectively zero sales accumulated roughly a million dollars of payroll before any revenue, a sum no ordinary founding team can fund personally.[163]
Between the two extremes sit intermediate sources. Cree’s founders raised their initial capital entirely from friends and family, in individual cheques of around $5,000 written informally rather than through any institutional investor.[71] Bank debt for a design-and-build business is secured personally, with the founder signing over their house and giving personal guarantees, so the entrepreneur rather than the bank carries the downside.[109] That is the underlying choice: fund the company personally and retain ownership while carrying personal risk, or risk an investor’s money and limit personal exposure at the cost of signing away part of the company.[109] Delaying a round while building on the founders’ own money improves the valuation, so the same sum buys the investor a much smaller share — on the order of ten percent instead of fifty.[95] Energy Micro started with about eight or nine employees in January 2008 funded entirely by the founders’ own money, much of it proceeds from an earlier acquisition, and operated for several years before taking venture capital.[95]
An explicit decision point is whether the company is meant to be venture-backable at all. The overwhelming majority of products sold in mainstream retail were never venture-backed, and a business that is merely good is well served by a modest crowdfunding campaign followed by successive products.[327] One EDA chief executive argued that the sensible target is a $10 million business — about the size of a dry-cleaning franchise — managed to throw off twenty percent profit, rather than a company built for acquisition; such a company can consist of little more than two people with laptops writing code.[99] The countervailing doctrine treats a lifestyle business — comfortable, interesting and self-sustaining — as anti-entrepreneurial, because prevailing advice pushes founders toward a growth business that can be built and then walked away from.[417]
What outside money changes
Venture-funded companies generally operate under an implicit exit strategy rather than a plan to become an independent long-term vendor, because the investors who supplied the money need a return and press for a sale as soon as possible.[2] Investors are not aiming at a twenty- or fifty-million-dollar outcome but at a five-hundred-million or billion-dollar one, and that target drives decisions such as closing a previously open design.[114] Returns require outliers — typically around $100 million of annual revenue within five to seven years — so a declined pitch usually means the company does not fit that revenue-and-timeline model rather than that the product is poor.[402] Returns come from a handful of large winners against thousands of investments that failed completely, so the visible successes are unrepresentative of the distribution a founder is entering.[491]
Those constraints reach inside the company in specific ways. A fund answers to limited partners who want returns quickly, so a moderately profitable but non-explosive company is the outcome it least wants in a portfolio, and it will encourage portfolio companies toward high-risk strategies that either succeed or fail fast.[394] Venture capital works as accelerant rather than ignition: it multiplies a business already burning, and pouring it on a venture too small to absorb it does not create growth.[327] Investors optimise for growth speed, but scaling a physical widget from one unit to a million is a machine that has to be built and is inherently slow, which is the fundamental mismatch between venture expectations and hardware.[491]
Control is the other thing that changes. Holding out for smaller investments preserves control, because the majority shareholder decides company policy; once a founder’s stake is diluted to a few percent the shareholders rather than the founder set direction.[24] Once a company has taken a first round and seated a board with a vested interest in its success, further rounds are attainable as long as the team keeps delivering and pivots sensibly, which is what makes moving from startup to startup a more stable engineering career than it appears.[407] A rescue can also cost control: Adapteva would have gone bankrupt after trouble with its crowdfunding campaign had a large telecommunications equipment maker and a venture investor not stepped in.[254]
Investor behaviour shapes engineering decisions directly. Investors routinely open by asking how many patents a company holds, which pushes founders to file patents they would not otherwise pursue simply to make the company look attractive.[173] A patent costs on the order of 100,000 to defend the first time, so for a startup the emphasis on intellectual property as a way of keeping competitors out is usually misplaced.[232] For a company raising 80 million with a steep burn, spending on the order of a hundred thousand dollars on a patent portfolio is nevertheless rational, since patents function as value signalling to investors and acquirers and the money was lost anyway if the company fails.[420] Some investors take the opposite view and treat a founder who leads with a patent portfolio as a warning sign, on the grounds that a patent position is not a business.[327]
The instruments themselves differ. Angel investors typically expect a return across a portfolio and invest partly for non-financial reasons, whereas venture funds invest solely to make money; the distinction is between someone whose job is investing and someone whose hobby is.[163] A fund is a trust relationship in which limited partners supply nearly all the capital and hold very little control while general partners supply little capital and hold most of it, with the limited partners taking most of the financial upside in exchange.[402] Because of how management fees and fund structures work, a first-time fund is often better raised from individuals — including partners at other funds writing personal cheques — than from institutional funds.[402] Under long-standing United States securities rules an angel investor had to hold a net worth of about 1 million in $10,000 increments from investors who did not meet it, and by raising the investor count triggering registration from 500 to 1,000.[87] Securities law dating from 1933 made public solicitation of investment illegal, so for the entire modern history of startups fundraising has had to happen privately and by introduction, which is a direct cause of the closed, tightly networked character of investor communities.[260]
Even a light investor relationship can be worth more for the discipline it imposes than for the capital: a quarterly request for financials obliges the founder to maintain an accountant, a bookkeeper, and real profit and cash-flow statements.[369]
Runway, burn and the size of a raise
A million dollars of funding buys roughly five people for one year in a company that is not yet manufacturing, which is the basic rule for judging how much runway a given sum represents.[172] Upverter left Y Combinator with 650,000 raised over three years across the whole first three years of the company before closing an institutional seed round.[163] A seed cheque of around 500,000 is large in absolute terms but small against the cost of building a hardware business, and companies that raised similar sums through crowdfunding have gone bust for want of enough capital.<sup><a href="#ref-327" title="Ep 327: An Interview with Avidan Ross (December 14, 2016)">[327]</a></sup> One hardware-focused seed investor's cheques ranged from roughly 100,000 or 500,000 or $750,000, sized to the company’s progress.[402]
The funding cycle runs at roughly annual intervals: a round is sized to be exhausted in about a year, so a company still holding a large cash balance past that point is read by its investors as behind schedule rather than as prudent.[394] Raising more than the plan requires is itself a hazard. One company was making rapid progress on a 15 million, grew to about twenty people, and found the investor expectation had shifted to burning cash quickly to accelerate.[394] Raising too much is more damaging than raising too little, because the excess sets expectations and spending patterns the business cannot support.[402] A company can reach a hundred million dollars of annual revenue and still fail, because taking too much capital sets a required outcome the revenue cannot satisfy; when investors stop believing that outcome is reachable, funding ends and the company closes almost immediately.[441] A media business with 125,000 magazine subscribers failed while smaller competitors ran profitably, in part because it had taken around ten million dollars of venture capital early on against an expected tenfold return.[447] Arduino took around fifty million dollars of venture capital, which sets a floor on any acceptable acquisition price far above the company’s earnings: it sells about two million boards a year with roughly eighty percent of income from hardware, and that margin does not support a valuation in the hundreds of millions.[707]
A startup that finds it has roughly a third of the money required to reach its target — a bank balance covering four or five months against an eight-month plan, before the usual schedule slip — faces a discrete set of options: pivot to something cheaper, go back to existing investors, or fold and return the remaining capital.[163]
Accelerators and intermediaries
Accelerators change the investment decision from a one- or two-week judgement on a pitch into a three- to six-month observation period, during which investors watch how a founding team functions under stress, handles obstacles and pivots before committing funds.[113] A batch culminates in a demo day at which each company gives a short pitch, with the actual fundraising happening in the weeks afterwards rather than on the day.[131] Y Combinator’s standard offer at one point was approximately $20,000 for about seven percent of the company, with founders expected to be resident for the three-month batch.[268] The application itself functions as a diagnostic — how the company will make money, how equity is split between founders, and what it understands about its market that competitors do not are questions founders should be answering regularly whether or not they apply.[268] An accelerator is not a source of hardware engineering expertise, and founders who join expecting to be taught hardware have misread the offering, which is funding, focus, peer group and help scaling.[531] Before accelerator funding, one solo founder sustained the company with outside writing and contract coding, and the principal value of the money was that it allowed full-time focus rather than the amount itself.[531]
Venture firms sometimes recruit a founder through an entrepreneur-in-residence arrangement, paying the individual and providing an office and access to the firm’s contacts in the expectation that they will start a company the firm finds interesting, though the firm is under no obligation to fund it.[129] When the sponsoring firm declines, the refusal itself becomes a black mark with other investors, who ask why the firm passed, and the founders face the ordinary pitch grind regardless.[129]
Outside firms fill capability gaps rather than replacing in-house capability. Engineering consultancies serve as surge capacity, supplying a mechanical engineer or a materials specialist alongside an electronics-heavy core team.[399] A product-engineering firm working with hardware startups described its typical client as having reached an eighty percent point: a cobbled-together prototype that may run only two or three minutes before breaking, venture or private-equity backing, capital available to buy inventory, and a working understanding of its market.[113] The typical client volume band for such a consultancy is between 5,000 and a million units, where the process knowledge is concentrated, and companies building a hundred units sit outside it.[451] Pebble worked with an outside manufacturing consultancy for roughly six months before its crowdfunding campaign, having already had false starts and collapsed financing, and at that point nothing but the team and product distinguished it from a hundred other startups with the same profile.[451] Foundries are similarly selective about direct startup business, expecting to lose money on any startup and treating the engagement as an investment of their time, so a fabless startup’s access depends heavily on personal relationships held by its investors and board.[129]
Equity, salary and employee risk
Stock granted to early employees has no realisable value unless the company floats or is acquired, since absent a liquidity event there is no market in which to sell it.[72] Founders normally hold common stock while investors hold preferred shares, which rank ahead in a liquidation, so a founder can retain a nominal fifty percent of a company and still receive nothing when it is wound up.[301] Paying a contractor in equity signals that the company lacks the funding to proceed and changes the relationship from consultant to consultant-and-investor; because founders systematically overvalue their own company, the shares are also near-impossible to price absent a valuation already set by an investor.[409]
The employment risk is concrete and arrives without warning. An engineer joining a startup is advised to hold three to five months of living expenses in savings, because the downside outcomes — sudden layoff, or a pay cut to minimum wage pending a funding round — are not visible in the recruiting pitch.[509] A stated runway also overstates the job security it provides: management will not spend down to zero, so a two-year runway typically produces a headcount cut at the one-year mark and another at six months to stretch the remaining cash.[509] One engineer who left a stable instrument company for a hardware startup found it out of money within four months, after which he continued working while accruing shares in place of salary — a common terminal state for an underfunded company.[305]
Working full-time at a small startup with wide responsibility also tends to produce heavy emotional investment in decisions made above one’s level, where contracting deliberately restores distance, since an hourly contractor can decline to attend meetings.[422]
The work itself
Large-company engineering work tends to confine an engineer to a small, long-established corner of a system, whereas an early-stage startup building something from scratch forces exposure to motor control, sensor interfaces, flash storage and everything else in sequence, which is where breadth is learned.[479] The corresponding cost is that a small company often leaves a specialist as the only person in their discipline with no senior colleague to learn from; deepening a specialism such as analogue generally needs a team large enough to contain two or three practitioners of it.[296]
Breadth in practice means absorbing whole functions. In a small hardware startup one engineer typically takes on supply-chain, test-engineering and fixture work alongside design, including pressing leadership to commit to fourteen-week lead-time parts long before the build — a warning that does not land because startups run reactively.[509] Planning beyond the immediate customer problem is difficult precisely because the inherited software model is cash-light and hardware is cash-heavy.[509]
The staffing sequence runs generalists first — broad engineers who can get something working end-to-end — and specialist experts only once funding is secured or clearly forthcoming, because experts are expensive and each requires supporting staff around them.[407] A one-person business can be grown well beyond the usual point before hiring, by contracting help as needed and farming out logistics and manufacturing while keeping functional testing in-house rather than trusting the manufacturer’s results; administration is generally the load that eventually forces a hire.[355]
Execution rather than novelty is where the value sits: the value of a product is the idea multiplied by the execution, and in products shipping to customers most of the execution effort belongs in the last ten percent of the project, which is where failure occurs.[232] Making one, ten, a hundred and a thousand units are materially different problems, and separate again from establishing that anyone will buy them; the step most often skipped is selling ten units first.[293] At the earliest prototype stage, before manufacturing and per-unit cost matter, a hardware startup should not be materially slower than a software one, and obtaining populated boards or a 3D-printed enclosure should add at most about a week to any iteration.[268]
Two structural asymmetries against software persist. Hardware has no built-in distribution channel — no equivalent of publishing to an app store and instantly reaching millions of users — and no recompile step, so a mistake cannot be fixed by rebuilding and redeploying.[268] Against that, customers expect to pay for a physical product, so revenue does not depend on converting a free user base.[268]
Pricing, margin and product scope
Pricing by taking the bill-of-materials cost and adding a fixed markup of around twenty percent does not produce a viable business; margins have to cover the whole cost structure, not just parts.[123] A business selling a physical widget at twenty to thirty points of gross margin is intrinsically hard to sustain, which is the underlying reason hardware was historically underfunded, and the category became fundable when hardware was reframed as a delivery mechanism for software.[402]
Crowdfunding pricing is set by the minimum order quantity a founder can afford: a 100 retail price, whereas an investor funding an order quantity ten times larger can drop the parts cost enough to sell the same product at 75 and reach a wider audience.<sup><a href="#ref-327" title="Ep 327: An Interview with Avidan Ross (December 14, 2016)">[327]</a></sup> Reward tiers can be structured to control which customers arrive: one campaign offered bare boards at a token 20 level, assembled boards for backers who wanted to save money and spare the team labour, and a fully hand-assembled final unit with cut acrylic, aluminium washers and a getting-started kit.[87] One augmented-reality company planned from the outset to use a strong crowdfunding result as leverage for a later venture round, having first pitched investors and been offered money only in exchange for a large share of the company.[173]
Product scope is a business decision, not only an engineering one. A development board sold without the daughter cards and application solutions that make it useful scores zero commercially, and the third-party accessory ecosystem that grew around Arduino or Raspberry Pi only appears once a large customer base already exists.[254] Running a kit business at maximum profit by packing and shipping every unit personally converts the venture into a permanent twelve-hour-a-day job, where revenue can be reasonable but margin per unit is thin and the labour never ends.[142]
How hardware startups fail
Death is usually abrupt rather than gradual: the company operates normally until the remaining runway is measured in days and no further investment arrives, at which point it stops.[351] Failing companies rarely disclose their position, because admitting to being near bankruptcy destroys any chance of a decent sale price, which is why outside observers see an apparently healthy company vanish overnight.[368]
Several distinct mechanisms recur. A connected-device startup that raised roughly a quarter of a million dollars from family and friends, then failed to close a $2 million seed round, ran out of cash after about a year of development and shut down.[249] A company that goes out to Series B investors with no market proof point and non-working prototypes burns those relationships permanently, since each investor is effectively a single pitch opportunity.[394] Crowdfunding stretch goals are a recognised failure mechanism: in one headphone project, adding mobile support expanded the software scope from two operating systems to five, added a 32-core on-board processor and required business development with handset makers, ultimately doubling the size of the development.[396] A campaign that raises three or four hundred thousand dollars is the beginning of a company’s risk exposure rather than the end of it, although founders commonly read it as the moment they are clear.[327]
Crowdfunded efforts fall into three bands: the barely funded one-off, the outlier that raises millions and becomes a full-time company, and a middle band described as the pit of despair, where the money is too little to hire a team but the obligation is large enough to consume the founder’s life.[104] A consulting firm formed around a group of researchers who intended to keep an unstructured side operation running alongside the billable business found the arrangement collapsed within a month or two, once the venture investors asked what the non-billable staff were producing and what the exit strategy was.[60]
Investor appetite is itself a variable rather than a constant. Around 2011 and 2012 capital was concentrated in software and social media and little was available for physical products.[402] The hardware boom of roughly 2012 — Fitbit, Pebble, Nest before its acquisition, Google’s first hardware products and heavy crowdfunding activity — created a market for prototyping services that has since cooled markedly, with fewer people starting hardware companies.[550] Investors in consumer electronics frequently will not engage below roughly ten million dollars of scale, leaving smaller open-source hardware companies in a gap between being too small for investment and too small to act like a large company.[127] The failure-to-success ratio in consumer hardware has been skewed far enough toward failure to move venture investors away from the consumer category in favour of industrial and business-facing hardware.[495] Faced with a software and a hardware investment at the same projected return, most investors take the software one because the risk is lower.[495]
Startups nonetheless exist because large companies have committed roadmaps and resources and structurally cannot pursue risky unproven directions; startups take that technical risk cheaply, and the few that succeed are acquired while most simply fail and disappear.[302]
Outcomes for founders
Selling a company does not necessarily leave the founder with money. In one case the structure of the sale produced a seven-figure tax bill and a net negative position despite a headline sale worth millions.[349] Returning to the job market afterwards can be difficult in its own right: a founder with decades of broad self-directed work may fail to qualify even for entry-level engineering postings, because those listings screen on narrow specific requirements such as five years of a single language rather than on breadth of shipped work.[349]
Founders who start again often change the model rather than repeat it. Before restarting a consumer hardware company, one founder listed everything he liked and disliked about running a hardware startup, found the dislike list substantially longer, and used that to decide which parts of the model — including venture funding — to avoid the second time.[715] Technical seed investors, for their part, select for founders who bring half-disassembled prototypes and want to discuss implementation detail rather than for polished pitch decks.[437]