This is part of the Startup Insider series, articles with the goal of helping aspiring founders, entrepreneurs and investors understand the ins and outs of the world of startups. You can sign up to stay up to date with this series here.
Growing up in the Bay Area and studying mechanical engineering in university kicked off Adam Kell’s interest in hardware. Subsequent experiences working on side projects, working as StartX’s Hardware Lead, trips to China’s Silicon Valley of Hardware and his current role as a partner at Comet Labs, an incubator fund focused on robotics startups, have taught him a lot of lessons about what it takes to build a hardware startup. Adam took time to share some of these lessons he’s learned throughout his hardware journey, along with trends he’s excited about specifically when it comes to the data flywheel effect of hardware, the rapid growth of artificial intelligence and the various use cases for robotics in solving hard problems.
First Hardware Startup While At Stanford
Adam would work on his first hardware startup as a material science master’s student at Stanford University. Adam would meet his co-founder through the material science community and their first product would be a way to charge mobile phones.
He shared, “I learned a lot through that process. A lot of the lessons I learned were from making mistakes along the way; a lot of it was from product testing. I moved to China for 6 months and spun up a bunch of manufacturing lines trying to produce our things then went back to the us to try to get our product sold in retail stores. It was like a 2.5-year path for the first product then a year in, we started to build the second product. That was kind of the genesis for the first company.”
However, Adam would realize that these consumer hardware products were a tough way to make money because if you build a product where you charge $100 and make $30 per product, it sounds good on paper but you have to factor in having upfront cost far in advanced then you only get the product a month or two later. He shared, “There’s a huge time gap and it really takes a long time to scale something like that. That’s why software companies are good and everyone likes investing in software since the marginal cost for shipping another product is low but hardware is not like that.”
Data Economy For Hardware
These realizations and early challenges would change the way Adam viewed hardware technology companies, as he started looking into this new economy of data. He shared, “Specifically data is what would make business models around hardware a lot more exciting. Companies we invest in at Comet Labs have this data flywheel effect where basically the company does 50% of what I want it to do and there’s this learning component that improves over time.”
He gave an example of a Comet Labs portfolio company called AMP Robotics that uses computer vision to sort recycling in a big assembly line. In most places you put all the recycling in one bin, the glass, cardboard, aluminum and only at the sorting facility does it get separated. He added, “What this company does is they use a computer vision system and robotic arm to basically pick out whatever thing it is programmed to pick out. It’s then able to figure out where the best place to grab it for the highest chance of success is. That’s interesting because the first installation is the worst one for this company. But once they have 100% units deployed, they’re generating a lot of data and the learning is a lot faster so the system gets better overtime.”
This type of system is what Adam’s first company lacked, a feedback loop and way to make their products learn overtime, which was one of the pieces that Adam wanted to learn when he decided become part of the founding team for Comet Labs.
Early Days of Comet Labs
Adam shares how during the early days of Comet, there weren’t really many investors looking at hardware companies with data flywheel effects which allowed Comet Labs to write some of the rules of the sector and position themselves as a thought leader. He shared, “We focus on enterprise companies so we only look at companies that sell to a retail store, farm or manufacturing facility—those are the types of companies we focus on. The reason why we focus on them is because that’s where we think this type of technology will take root in first.”
Adam believes that it’s these large corporations that understand the value propositions and are able to make these decisions to purchase expensive technology since the operations are tighter and purchase decisions are more economic as compared to the consumer side. Additionally, the Comet Labs team believes that for a lot of these opaque industries, the pool of people who know how to use these cutting edge tools in artificial intelligence and robotics are really small so they want to engage corporate partners to help this pool of early stage founders think of applications of these technologies so they won’t end up building the wrong things.
To get this conversation started, Comet Labs would host events inviting large corporations and startups to meet each other. Additionally, Comet Labs also built relationships with generalist VCs who would pass along early stage robotics and artificial intelligence companies to the team at Comet Labs to take a look.
Hard Things About Hardware
One of the big challenges Adam has sense when it comes to hardware startups is knowing what to build yourself and what parts to outsource. He shared how a lot of these very technically minded founders feel like they have to build everything themselves. Additionally, Adam shared, “If I want to build a mobile phone, the supplier is very clear. This isn’t the case for people who want to build robots versus cell phones. There’s still a gap when it comes to the prototype platform, who the suppliers are and what can be shipped.”
The other core area that Adam thinks that hardware startups have a hard time with is figuring out what their data flywheel is and collecting enough data at the beginning to make sure that the MVP is functional. He shared, “So I think making sure that you can achieve this baseline level of success is important because in most cases once the bridge is burned, you probably will never get another shot with that customer. Hence, I think the ability to have a proving ground where you can say let’s actually exercise our system and develop 1k hours and realistically do pilots in prime time environments.”
Advice To Founders Working On Hardware Startups
Finally, when asked what advice he had for hardware startups, Adam emphasized the importance of the customer development process. He shared, “Before you cut a piece of metal or screw a single screw, you talk to 20 customers and it feels like that’s what you’re supposed to be building but that’s actually when you have to talk to 50 more. The customer development process is a short loop but building a prototype and testing it is a long loop and that’s the place you spend all your money and waste all your time.”
He added, “I think this is why customer and product centric founders are sort of my favorite because they obsessively think about this question. They think about is the problem they’re trying to solve a real problem. You need to make sure you’re building the right thing.”
You can learn more about Comet Labs here.
About the Author: David is a senior at Penn studying Cognitive Science and Computer Science, originally from the Philippines. At Penn, he’s heavily involved in the startup scene as an investment partner at Dorm Room Fund. Currently, he’s working on SkillStackers, the easiest way to scale work using a virtual workforce. Previously, he started a nonprofit organization called YouthHack which has gone on to scale to do programs in over 8 countries in the last 3 years. David enjoys meeting new people and sharing the stories of exceptional entrepreneurs!