Is Hardware Really That Hard?
Here's a beaten-to-death horse for you: "hardware is hard". i'm sure you've all heard it before and if you haven't, I'm not sure you're on the right website? It's become such a commonplace saying that it's been adopted as a universal truth. And while even I am galivanting about with it on my homepage, I do want to offer some nuance to it.
Why is it hard?
In very simple terms: time & money. You probably all know the iteration cycle from the design thinking world as "design-build-test". For hardware two new contestants enter that loop, namely "pay" and "wait".
I hate those guys
If you design something in hardware, you then have to pay to order the (custom) components and wait for them to arrive. Then you have to assemble it, hope it fits and hope it works. Spoiler alert: it won't. But you'll learn, adjust your design and create a better version, after which you'll pay and wait again and so on.
Yes, I'm aware 3D-printers and Raspberry Pi's exist, thank you. But in reality those are only useful for very early-days proof-of-concepts or low-volume design for users to play with in a limited beta MVP launch that in all fairness should not be classified as a "product" and is barely more than a science fair experiment. However, extremely useful to gather early user insights! And even then, they only reduce the time part and limit the cost only so much depending on what you're building.
I'm mostly talking about the real deal here when you’re looking to develop something that scales. 3D printers — while useful — only get you so far and are far from representative of a final product. If you're using commercial services to expand your material and process options beyond your desktop FDM printer, you’re still waiting at least a week and costs might rack up easily. In other cases printing doesn't cut it and you need to look at machining or sheet metal for prototyping. And while some services exist today that can deliver parts within a week, that's often limited to pretty simple and straightforward manufacturing operations. Once you need something more specific or precise, you need to start looking at a machineshop and turnarounds can be anything from 3-6 weeks and prices start rising.
The same once you’re getting your hands dirty into building an actual PCBa. A lot of services exist today that can do it quite rapid, but it limits you in the type of components you can use and the complexity that the board can have.
Another point is of course risk as pointed out in “Always De-Risk Your Market First”. And especially the cost of change once you're getting further in the design cycle.
Consider injection molding. If you look at classical hard tooling — which can range anywhere from 10k USD to 100k USD for a single mold — going from a finished design to a finished part can take anywhere from 3 to 9 months. Main reason is it takes long to build a tool and it often needs to be shipped around the world on a boat to save cost, but also because of changes. If you change the design once the tool is made, your options are very limited, there can be some mild to aggressive retooling with their respectively ascending costs associated or you can even run at risk of having to chuck the tool entirely and having to start form scratch! Ouch! Both in time and money
Point being, shit's expensive and takes time.
What else? That cost of change really bites you in the ass once you've launched. We've all heard about the exploding Samsung phones: you can't fix that with a simple over-the-air update. Electro-mechanics are what they are once in the field and if you want to chance it, you have to do a recall. Which sucks in terms of brand damage, but is also insanely expensive. In the Samsung example they had to recall 2.5 million phones to the tune of 5.3 billon USD in losses as a result of the incident.
Hard as compared to what?
So yes, that makes hardware risky, expensive and time-consuming, but does that mean it's hard? And if so, as compared to what? Software is the typical victim of this comparison and the subject of a lot of engineering banter such as "if software were difficult it would be called hardware".
During COVID I picked up the hobby of video game programming and I can tell you it pissed me right off. I was so friggin' jealous of how easy it was to run an iteration cycle! If something bugged out, I adjusted some things and tried again, and again, and again, … all-in the span of a couple of hours, it was so liberating! But on the other hand I once spent three days looking for a misplaced semicolon, so there's that.
And while software startups can be much more agile, scale quicker, test for willingness-to-pay more easily, … They face different challenges that can be just as hard as those of hardware. They often say that "hardware engineers design for core functionality, software engineers for edge cases". And that is so true. The more layered and complex your software becomes, the more unintended bugs can pop up, the more possibilities users have to perform a wrong sequence of events.
And you have to consider the hardware randomness as well. Your hardware systems might unintentionally flip a bit due to any number of reasons, including cosmic rays from outer space! You really need to cover all those edge cases because they can cause huge problems, security breaches, crashing systems,… while if one person misuses your one hardware device in one specific way, you might end up with a brick, but that's that, replace it and be done with it. Although, even that depends on the type product risk it carries. If a coffee maker craps out, that’s annoying, sure. But if your pacemaker dies on you, well…
And it's not like software doesn't incur any costs, again depending on what you're building, but server costs are real and in the age of AI tokens the COGS of software are becoming more and more important when considering your unit economics. That's becoming increasingly challenging for consumer software companies since most of their scaling tactics are based on offering free services until virality kicks in. Also leading to large delays before being able to achieve profitability.
There of course examples of blitz-scaling (as what Uber did) where you scale relentlessly despite of negative unit economics by burning huge amounts of cash until those numbers flip. But as of this writing, capital is becoming more expensive and those are some huge risks that fewer and fewer investors are willing to back.
And then we haven't talked about operations-heavy businesses. Think grocery delivery services and the likes, everything that involves orchestrating a lot of people and goods as your core business is a nightmare to run and scale.
I had the privilege of working with a smart steamer startup that combine the holy trifecta of hard things: a hardware consumer device with a companion app and a frozen food delivery service. I can tell you that those guys had it really hard due to the logistics of it alone.
One of the reasons hardware is inherently called hard is due to the heavy operations involved when dealing with consumer devices. If you look at low-volume B2B machines it's a whole different ballgame. If you’re only building one each year and you don't have the burden of scale, a lot of things become easier. It's still hardware, but it's completely different in terms of operations.
Let's talk numbers
In general, all new ventures have a failure rate of about 70%, main reason often being something's being built that no-one asked for. For consumer hardware startups, that number is even higher with 97% according to some sources! As stated, that's the killer combo of hardware and operations-heavy businesses at work.
On the flip side, I've seen numbers of accelerators that show that — while they are in the minority — the hardware startups tend to outperform their software brethren by a factor of almost two in terms of survival rate. Now, there's a lot to be said about what you define as a "failure" or “survival” and how far in the future you're willing to look. And while no research was done on why it was the case, I feel it's most likely that because software has a lower barrier to entry, it attracts a wide profile of people which can result in higher chance of ill-fitted personalities to run a startup.
Hardware founders are almost always technical and with an engineering background and are naturally risk-averse. They already have to be damn sure about their tech and their own abilities before even attempting to start a new venture. But that's just conjecture from my end.
Conclusion
I think it's fair to say that the "hardness" of you startups, be it hardware or software, really depends on how challenging your tech is and how heavy your operations are. You could be building a really simple low-volume B2B device that's two clamshells and a basic PCB, or on the other hand you might be trying to build a commercialized decarbonization unit for home use.
The really hard business usually springs forth when trying to combine high product risk with high technology risk. The same is true for software, you might be a lookalike B2B SaaS application or attempting to build the next Google. Those things are not the same.
Mainly because of time-to-ROI and increased cost, hardware startups tend to get treated like hot garbage by some circles of investors, but it seems unfair as all startup categories have their unique challenges.
This unfortunately results in a much more absent ecosystem for hardware startups not getting the focus they deserve. There are of course exceptions such as the wonderful work funds like HAX/SOSV and BOLT VC are doing, but I always feel like our hardware buddies are left out in the cold, receiving generic startup advice and having to deal with investors who lack the tactical knowledge of what it means to do hardware right.
Which is a shame. Because I don't think hardware has to be hard. If you accept the cost and time frame and seek out the necessary help to work according to a good process (an there are plenty, as pioneered by the likes of Philips), have patience and can restrain from wanting to move too fast, hardware can be really fun and satisfying to work on. I sure love it, it feels great to work on something tangible that you can see affecting the real world.
In the end, I feel it's time to move past this unnecessary comparison and that VC's and investors grow some balls so a hardware ecosystem can flourish!
