Hardware is Hard

We have written a lot about the blind spot the US venture ecosystem has for semiconductors (and here) and hardware more generally. We recently helped a venture firm conduct due diligence on an electronics company, which put us in the awkward position of having to walk in someone else’s shoes. Sympathy for the devil, and all that. In our posts we have attempted to be fair, pointing out that US VCs are making sensible economic decisions albeit on a short-sighted viewpoint. Investing in hardware is much more capital intensive and thus risky than investing in software. In this post, we want to work through some of the math behind that.

Let’s say we have a company, Unicorn Electronics, that wants to build The SmartHorn TM, a device that magically solves customers’ problems perfectly. How do we get the company to “escape velocity” – positive cashflow sufficient to fund the company’s immediate growth needs?

First, Unicorn needs to design the Horn. Depending on how complex the device is this can cost anywhere from $100,000 to $1 million, from a third party design shop. Some people can do this design themselves, others are going to need outside help. For our purposes, let’s say the Horn design costs $500,000. This amount gets the company the “blueprints” and a few prototypes.

But most devices today are fairly complex. Even if the hard computation is done in the cloud or in the app, the device itself needs to be able to function, to be programmed and to communicate with the Internet. That means firmware. This is low-level software, and not only is their no standard “firmware” programming language or operating system, pretty much every chip company has its own instruction language. Hiring firmware engineers is not that expensive in dollars, but it can be very expensive in time. Here’s a typical scenario. The design shop hands the firmware designers some prototypes. The firmware team finds a bug that requires a new wire or capacitor. So the design firm has to build (by hand) a new prototype. The firmware team then finds that two of the chips are not speaking to each other – the Bluetooth and the temperature sensor for instance. Is the problem in one chip or the other or somewhere in the wiring in between? Often this will mean the design team needs to spin up yet another prototype. And since these teams are likely in different cities, if not different countries, every re-spin loses days in shipping and re-work.

So far, we have spent $500,000 and a month on the design, and $100,000 and four months on firmware.

Now we have to find a manufacturer. This company will need to create tools and molds as well as designing a manufacturing flow. They can do this fairly quickly, but as they are operating on expensive machine tools, they will have minimum quantity requirements. These can be as low as a few thousand units, but here is where the costs really start to add up.

Let’s say the SmartHorn costs $100 in parts and manufacturing, but the manufacturer requires a minimum of 5,000 units. Then the tricky part becomes sourcing parts. The global electronics supply chain is incredibly tight right now. We wrote about this back in October, and the New York Times was catching up last week. Lead times for many parts can be several months, and frustratingly it is hard to predict with often some of the lowest value parts in the shortest supply. So add $500,000 (5,000 units X $100 BOM) and four months.

Then we enter into a whole sea of minutiae. The device will need various certifications UL, FCC, etc. Add $50,000, but fortunately this can be done while the manufacturing is ramping up. But there are bigger costs as well such as insurance. Most distributors (i.e. retail) and enterprise customers will want the company to maintain a range of insurance policies, tack on another $50,000 or so. Then we have to factor in shipping and customs and taxes, so maybe $20 per device.

Next up, the company has to factor in warranties and returns. The accountants have requirements for setting reserves for these, but a rough rule of thumb (sufficient for our purposes here only) is 10% of the costs of the materials needs to be set aside to cover the cost of returns and replacements.

Finally, the company has to contend with marketing. This amount can range from nothing to A Lot, depending on whether they are selling to consumers or to a single large enterprise customer. A basic PR budget and some content marketing runs about $50,000, plus or minus.

Adding this up we have the following budget:

Design$500,0001 Month
Firmware$100,0004 Months
Manufacture$500,0004 Months
Shipping $100,0001 Month
Sub-total Cost of Goods Sold$1,350,000
Warranty Reserve$135,000
TOTAL$1,485,0009 Months

All of this has to take place before the company can collect any revenue, unless they have done some form of crowdsourcing. So just to break even the company has to charge $297/unit, and to get a 50% gross margin the retail price would need to be $594.

We also need to factor in software costs. For this example, we will assume Unicorn Electronics does their own app and back-end software. The company could probably build all of the hardware and software with a team of ten people. If the average salary is $150,000 year (assume some expensive onshore talent and some less expensive offshore talent), the total comes out to $1.5 million for a year.

So $1.485 million in hardware costs and $1.5 million for headcount, for a total of $2.985 million. The company then sells 5,000 units for $594 a piece or $2.97 million in revenue and yielding a net loss of ($15,000). Hardware can be dispiriting.

The good news is that there is a heavy fixed costs component to this model, so tweaking a few variables can have a big impact. For instance, if they raise prices by $6/unit to an even $600, the net profit goes to $15,000 (positive). Sell an additional 1,000 units and the net profits are close to half a million dollars. If they sell 100,000 units then the company has a path to becoming a true Unicorn. There is a lot of leverage in the model, which is why people keep trying to build hardware.

That being said, the realities of hardware can be daunting. Take the $3 million the company needs to get a product out the door, a similarly sized team could build a software application that can potentially grow much faster and they would still have a million dollars to do customer acquisition. Compounding this, the software would come with a recurring revenue stream. The trouble with hardware is that for the company to grow, they have to come out with the Smart Horn 2 and the Smart Horn 3, and go through all of the above again, and again.

Hardware has immense financial leverage, but the upfront costs and more crucially the time required are daunting.

Photo by Usman Yousaf on Unsplash

14 responses to “Hardware is Hard

    • Shenzhen has a few advantages. People assume it’s cost of labor, but Shenzhen is now fairly pricey on the front. The real advantage SZ has is the fact that you can go to a manufacturer and they can do the design and often the firmware for you, incredibly quickly. This comes with trade-offs but is often a much more efficient way to get a product out.

  1. Pingback: The maze like world of QAnon, where we're at on climate change and views on British class identity - Inspiral·

  2. Shenzhen gets dirt cheap if you are creating a derivative product. Those sub$100 (plus a bit of “marketing” money and designs from Intel) created a lot of cheap tablets very cheaply.

    • Shenzhen can get things done very cheap. They have an incredible ecosystem there, but all that comes with its own overhead in time managing the relationship.

  3. Pingback: The SPAC Hangover | Digits to Dollars·

  4. Ha! This time schedule is completely unrealistic. You can easily use the “pi” factor. Multiply those 9 months by “pi” and those 28.27433 months are much more realistic. Just check all the kickstarters etc. to see how many of them can deliver within a year.

  5. Pingback: Opening Bell 28.01.2021 - StockViz·

  6. What a great analysis! Depending on whether the product is a better mousetrap serving a market overrun with mice, or an entirely new idea, intended to serve a need that has not even been validated, in markets and channels that may be unknown, a good deal of money and time can be spent on all of those activities. Prototypes are also often required to test need and demand, because people lack imagination and market research methods are often unreliable.

    • You make a good point. Wish I’d thought to frame it this way in the post – but it boils down to “finding product-market fit” is much harder in hardware than in software. You can go through a lot of money before you find it in hardware. Software is not easy, but it is possible to iterate much faster.

  7. Pingback: Read, Learn, Improve – 30-Jan-21 – Random Thoughts of Analyst·

  8. Pingback: Hardware is Hard | Snippet Finance·

  9. Pingback: How To Build IoT Project From Scratch? [2022]·

Leave a Reply