VCs and Robotics…a complicated relationship
If you present a SaaS company at $1m ARR growing 3x year-on-year to a VC there’s a 99% chance they take a look. If you present a robotics company at $1m in sales growing 3x YoY, they most likely won’t. We’ve spent the last few months meeting many founders, researchers and investors in the space to explore why that is, why it’s changing and narrowing down the exciting areas we’re bullish on! We look forward to investing in robotics, if you’re a founder — feel free to reach out to me: ssukumar@balderton.com
Introduction
It’s worth exploring the history of robotics first (with help from Jeremy Wyatt at Amazon). Actuators and motors made it technologically feasible to automate precise and repetitive control problems, leading to the first wave of robotics. These robots solved point problems and were built from the ground up, predominantly for industrial use cases. Wave two was driven by advances in mobility and machine vision increasing the complexity of problems robots could solve. Crucially, this wave only saw success in controlled environments such as warehouses. Deep learning is the main driver behind Wave three, which can be labelled as “unstructured manipulation”. This broadly means tackling unbound / generalised problems — imagine a robotic hand that can pick up any object with no prior training.
The success stories from Wave one are still the biggest success stories in robotics, with many companies founded in the 70s still generating $bns in revenue today. They require a lot of CapEx and balance sheet strength, making it difficult to disrupt. Interestingly, Europe is home to the majority of incumbents (e.g. Kuka and ABB) with Japan coming in second.
Wave two success stories are scarce (there are some examples — the majority in warehouse automation). It’s evident that more capital was required for smaller exits with long timelines (vs SaaS). A few highlights below:
However, AutoStore is a great example of why VCs shouldn’t ignore robotics. It hit $328m in sales in 2021, up 80% on 2020, and its order intake grew 160%. Below we can also see that it even trades in the top quartile of SaaS companies.
I won’t waste time discussing the tailwinds for robotics as these have been consistent for the last decade and there are many McKinsey reports you can refer to. However, the aftershocks of the pandemic have amplified many of these, such as labour shortages and margin pressure.
Investing in Robotics
First and foremost, investing in robotics is far from easy. Balderton did a deep dive into robotics back in 2014, highlighting 22 early-stage companies to track across Europe — 18 of these companies no longer exist and the other four certainly aren’t thriving. A SaaS company can be built in a few months and you’ll quickly be able to judge PMF. Robotics companies have to push pilots meaning it often takes years before they have their first production contract. Many companies get stuck in “pilot purgatory”. They also require a lot of follow-on capital to scale and this is all before we question the technological feasibility of many of these companies (although it is worth noting 50% of jobs are feasible to automate).
As seen below, this leads to expectations driving valuations rather than sales.
Over the past couple of years, a large amount of venture capital has flooded into the robotics space with Tiger Global participating in 19 deals. GV, Lux Capital, Khosla Ventures and Founders Fund also lead the way in this space.
One trend that’s making robotics more appealing to VCs is Robotics as a Service. This means providing hardware, software and maintenance in one package for a monthly fee. This solves the CapEx problem for customers and provides more predictable revenue plus sticky relationships. There are many asset-backed lenders moving into the space and enabling this business model, such as Silicon Valley Bank and Kineo Finance. SVB produced a great report summarising some key metrics they look for when underwriting “Hardware as a Service” businesses — many parallels can be seen with SaaS, making VCs more comfortable. However, RaaS models do come at a price, as shown by revenue multiples below.
Exciting Trends
Autonomous Inspection
One of the most exciting trends is automated data collection i.e. using robots to harvest regular, reliable data which is vital for operations. This seems to be where most capital is flowing into with ~15% of deals in the last two years being in the space. In many industries it’s compulsory to periodically gather data, for example, warehouses need to consistently do a stock check and oil rigs need to check a range of sensors to ensure no anomalies. Many of these tasks are incredibly tedious and often dangerous for humans. Current options are either hiring more operators or installing thousands of IoT sensors that struggle with power, data networks, cloud storage and maintenance costs — neither of which are scalable.
There’s also an increased emphasis on predictive maintenance, meaning operators will not only gather data to check if something has gone wrong, but they can also leverage this data to forecast anomalies and do many other clever things.
The once fabled Boston Dynamics Spot is now fairly commoditized, with quadruped technology springing up in startups across the globe. This means success will be driven by deep software layers, which can turn vanilla data collection into useful insights. It’s interesting to see that Boston Dynamics itself has shifted most of its focus to the autonomous inspection space
Abstraction Layers
Demand for most technology is outpacing talent but this is particularly pronounced in robotics.
Many startups are trying to abstract away the unnecessary complexities associated with robotics to enable more operators to leverage the technology. This also opens up success for software in robotics.
The most notable is Wandelbots, which enables operators to program the path of a robot by simply pointing with a magic wand (something which would have previously required a robotics engineer and several hours). The problem with abstraction in robotics is that the hardware and use cases can be very fragmented and scattered making scalability difficult. This means focusing on specific sectors or use cases is very important.
One founder I spoke to stated that:
Automation is a change management problem rather than a tech problem, there needs to be a focus on customer success so you can’t cover the tail end
Collaborative Robotics
Collaborative Robotics (a.k.a. cobots) is one of the fastest growing but also most controversial areas of robotics. Industrial robots weren’t designed to work with humans — in fact, they’re very dangerous, meaning there’s strict legislation on what robots can do with humans in proximity (e.g. limits on speed and torque). Cobots are designed to work with humans so that the robot can cover the repetitive task in the workflow and the human can tackle and bespoke adjustments required (e.g. in electronics). There’s also evidence to show that humans + robots are actually more productive than robots independently.
Universal Robots paved the way for cobots and has seen immense success — it hit $312 in sales, growing 40% year on year. Many industrial incumbents are pivoting to cobots, justifying that this could be the future but Universal Robotics still dominates.
However, there’s no avoiding the fact that there are limitations on what these robots can do due to legislation — cobots have been described as “automation in slow motion”. Investing in cobots would be a bet that this legislation will be loosened going forward.
Warehouse Automation
It’s hard to ignore warehouse automation when discussing robotics. Amazon wouldn’t be where it is today if it hadn’t shifted its focus to automation after the acquisition of Kiva Systems. It’s no longer a nice to have, with the growth in e-commerce and demand for shorter delivery times it’s become imperative for sustainable growth. It’s worth noting that Exotec, one of the most exciting robotics companies in Europe (if not the world), operates in warehouse automation.
It’s a mature sector but there are still opportunities, 75% of warehouses still lack any form of automation. The majority of these are in the sub-1000m² segment — showing the main opportunity is for “micro-fulfilment” / SMEs. It’s important to note the goal for all companies is e2e automation with “lights out” operation. This means there’s enormous value in integrated solutions, which can make it harder for smaller players who focus on point problems.
Building on AutoStore’s success — there are numerous startups trying to deploy the Rubik’s cube approach on a smaller scale. The biggest hurdle to cross here is proving ROI; in an Amazon-scale warehouse the economics of automation make clear sense due to hundreds of operators needing to be present. In a smaller warehouse, with only three or four operators, it’s harder to justify.
Conclusion
In many cases, automation is no longer a nice to have — it’s now a necessity for sustainable growth. Incumbents have dominated historically but there are many new tools for startups to disrupt, such as Robotics as a Service. There are now also case studies (e.g. AutoStore) to show investors there is scope for immense exits.
Balderton Capital has backed many of Europe’s most ambitious businesses and there are strong signals to show robotics will play a key part in the next wave. If you’re a founder in the space we’d love to hear from you, feel free to reach out to me: ssukumar@balderton.com
