From Ironman to Industry: VC Insights on the AI Revolution in Robotics
What first comes to mind when you think of the word “robot?” Is it a humanoid machine coupled with artificial intelligence? Automated arms used in manufacturing, or any sort of unmanned vehicle? The possibilities to dream up what robots can do for us in the near future seem endless. Investments in the robotics space have been growing robustly as attention from global markets push towards growth and new innovation.
Towards the end of July, we hosted yet another fantastic event at Google Sunnyvale campus, this time a Robotics Founder & Funder Summit that highlighted the trends, investments and future of AI in the robotics space. The VC panel was moderated by Sahil Segal from Toyota Ventures, with esteemed speakers Xuhui Shao from Foothill Ventures, Sasha Ostojic from Playground Global, and Han Hua from (GV) Google Ventures. Please see their bios at the end of the blog.
Here are our top takeaways from the VC panel:
- Robotics Investment Trends: Robotics investments have fluctuated, primarily influenced by labor shortages and technological advancements, with expectations of a rebound driven by these factors.
- Vertical Integration in Robotics: Successful robotics companies often integrate both hardware and software, though there is potential for future specialization. The debate continues on the effectiveness of software-only solutions versus integrated systems.
- Challenges: Sectors like manufacturing, agriculture, and healthcare face high demands for efficiency and safety, along with regulatory hurdles, making them slow to adopt advanced automation and AI. However, advancements in LLMs are starting to bridge these gaps. There are significant data challenges, and owning a strong data strategy is essential for success in the robotics field.
- The Importance of Data: There are limitations of simulations in robotics, which places emphasis on the fact that real-world data is crucial for addressing edge cases and achieving high accuracy, which is essential for the development of robust robotic systems.
- VC Support: VC networks generally provide crucial support in areas like fundraising, recruiting, and customer/partner introductions. There is also opportunity for technical help, with the aid of engineering or creative design teams.
Below is the full discussion from our VC panel.
Sahil: Sasha, you’ve worked in Silicon Valley for the past 30 plus years, you’ve had a front row seat to innovation breakthrough technologies such as speech recognition, GPU computing and self driving cars. Which of these was the most effective?
Sasha: The most interesting one that came out of nowhere was speech recognition, which happened early in my career. It was one of those things where I got the chance to work on a truly transformative technology but without a clear idea for who would really use it, how it would be used, and the impact it would have.? This was in the late 80s and early 90s, before the internet, before streaming, and before fancy electronics in our homes. My parents would ask me what I was working on, I said we’re building these computers that recognize voices and they say, “why, who needs that?” Until one day a family member said “Hey, I called Fidelity and I just spoke my account number and they gave me my balance and I was done in 30 seconds. It was just mind blowing.” And I was like, “I built that!”, and finally people were like OK, now we know this will be huge. It was one of those typical Silicon Valley things where you invent a technology,it finds an amazing market, our lives are improved dramatically…until we all start hating it because you cannot talk to a live person anymore.
Sahil: Han, what has drawn you to AI and robotics? What excites you the most?
Han: I’m a big sci-fi fan — I love Ironman — Jarvis. Looking back at the history of human civilization, at the end of the day, human society wants more intelligence, more compute, and more energy over time. So we kind of know where the end state lands. The interesting part is we just don’t know the timeline and the step functions. Being in a VC role, I’m still learning as everybody else is, which is super exciting. To work with every entrepreneur and just witness and hopefully contribute a little bit to the journey is so exciting.
Sahil: Having a PhD in ML and having followed the past few AI waves, what do you see is different and what concerns you the most?
Xuhui: I never thought that what I did 25 years ago would still have some relevance today. Back then I was studying model complexity control and the VC dimension back. And I was surprised to learn that nowadays people really just use large language models without really being concerned about the statistical learning theory, the error bounds behind it. I think this AI wave is different in that it has so much money, so much attention, and such a broad impact that it is bringing to the entire world and investments, especially the Magnificent Seven companies. One of the most concerning things is that it raises the cost barrier so high for not only startups companies, but also for researchers in universities. They can’t afford to access 10,000 or 100,000 GPU cards. At the same time, I think it’s the most exciting time but it is also the most interesting time for us to reflect on what is investment, what is research, what is innovation.
Sahil: Funding in robotics has gone up and down — if you look at from 2019 to 2023, 2021 was the global peak of investments. Global investments in robotics was around $28 billion. However in 2023, robotics investments went down to around ⅓ of that, $10.6 billion globally. With robotics and investments expected to rebound in 2024, what factors are driving this growth? And do you think that this will be sustained?
Sasha: Investments in robotics, like most other novel technologies, go through the ups and downs of getting the timing right. The way we look at it, there’s the macro, which is the pull, and then the technology, which is the push. The macro is really driven by labor shortages and the dynamics of how different industries in the physical world handle scale. I happen to be on the board of a public company in the midwest that is really struggling finding machinists, electricians, and generally people to work on in factories, to build things. Unfortunately, potential workers have other options like working at an Amazon fulfillment center, driving Uber, or Doordashing, which aren’t really viable long term career options. The result is that you just cannot find enough labor to build goods, to reshore manufacturing, and all that will be really important for the US and in our strategic national interest. So that’s the pull. During the COVID-19 pandemic, many turned towards robots because no one else was able to do the manual labor. And that kick started that cycle of actually all of a sudden deploying these automation solutions, with manufacturers, warehouses and fulfillment centers showing active interest, and enabling the cycle of improvement. As you automate, you collect data, you incrementally improve things, and all of a sudden, you’re on par with humans or better. And now there are many proof points that robotics and automation do work, and do add value.
The push is all the technological breakthroughs. Everything is getting much better — the performance, the software,the capabilities, and the mindshare that’s going into this, with the intersection of LLMs and foundational models making their way into the hardware space. There’s a lot of optimism, followed by a lot of money.
Sahil: We’ve seen more and more verticals being introduced in the robotics space to be more specialized. What sectors are emerging that are better verticalized? What sectors are attracting investor attention?
Xuhui: Some traditional industries like manufacturing, agriculture, health care and services are primed to be disrupted and adopting robotics and AI technology in general. There is a reason why these industries are slow in adopting advanced automation, AI and robotics. One is that the demand for efficiency, safety and a success rate is really high. And many industries also have regulatory hurdles. Also, you need to really understand the domain really well to deploy effective solutions, and that’s what makes the current wave of innovation really interesting and enabling. Now with the advent of LLMs, we’re seeing the hope of giving robots additional capabilities, such as working and understanding the environment and other things such as accelerating the training in virtual spaces.
Sahil: Seed and Series A investments in robotics have declined about 27% from their 2021 peak, whereas later stage investments have declined about 60%. Does that mean that raising earlier rounds is easier then it becomes more difficult in later stages? How should we think through depressed valuations in larger rounds?
Sasha: It’s kind of true that for any area of investment that later stages are harder. Robotics is harder because there is hardware, there is the longer cycle, and there is the macro side again that has to catch up with the timing.
When the legacy companies are trying to raise, it’s harder because they’re burdened by legacy. They cannot talk about how they caught the wave of the latest technologies, because they can’t start from scratch.
For future founders, there is an opportunity to trailblaze with not just state of the art technology, but with the business model as well. Of course, the business model will vary also depending on the vertical and what you are doing, but the hard truth is that later stages are harder to fund.
Sahil: How would each of you three define robotics, as it can really vary person to person?
Han: It’s really difficult. It’s a very domain and scenario-dependent answer. If the robotics category is successful, you probably won’t hear the word “robotics ” again — as a general term, it’s pretty much just an intelligent machine with partial or full autonomy.
Xuhui: There’s a utilitarian definition of a robot which is just being able to automate tasks that were previously only performed by humans. Then there’s also the more romantic and more aspirational definition of a humanoid machine that will eventually replace us.
Sasha: My definition is somewhat simplistic — it has to have actuation sensors in some level of autonomy. And then it expands from there with a bunch of software and capabilities and grippers and limbs and what have you.
Sahil: We’re seeing more robotic companies right now, especially in humanoid robots that are developing their hardware capabilities and manufacturing hardware themselves. What is your opinion on software-only solutions versus software and hardware integrated solutions?
Xuhui: Alan Kay, Turing Award winner and a big inventor in computing, basically said that for people who love software, you build your own hardware to run it. So the takeaway is that a truly optimized system has to be jointly optimized across hardware and software. I think this is true in this case, for robots. Every company has to pick their own approach, but in general, I think there is a warming up of investors on hardware investments in recent years. Another reason, I think, is that there is a desire for all, if not most of the major countries to try to secure their own supply chain. And so I think there’s also a desire for the United States to also have a secure supply chain in critical areas like advanced manufacturing, robotics and AI. So there’s a desire to also invest in this area.
Han: I’m skeptical of the software-only approach as a starting point. First of all, there are a lot of software-only robotics companies and their hopes are to build the robot brains which can be licensed to hardware companies and expect to acquire data to train their models. However, hardware companies typically are reluctant to share data.
Secondly, if your hope is simulation, a lot of the real-world longtail edge cases cannot be fully simulated yet. The third reason is that vertical integration allows companies to move faster and innovate wherever the bottleneck is.
Sasha: I’m a firm believer in vertical integration wherever possible, but there is a timeline to these things. At some point it becomes possible to desegregate things and for software to add value, but we have to get there first. And I think we have to get there by doing vertical integration first, maturing the industry and then getting to a point where you can do a software-only solution. Before you can build up substantial business in any sub-component of the robotics stack, both hardware and software, there is a timeline to get there. I believe vertical solutions should be built in some generalized way in the hope that you can do other things along the way, but at this stage in the robotics industry development, we’re not there yet.
Sahil: What are some of the data challenges you’re seeing? Can synthetic data solve that or are there any other options that we have to solve these data challenges?
Xuhui: When we look at investments in this space, it breaks down to either a platform company, an infrastructure company, or an application company. or a platform play, your core competency is usually algorithms.
For infrastructure company the core competency is usually compute. For application companies your core competency is data. So you have to own your data strategy and how you can collect the highest quality, possibly privileged-access or exclusive data set that can be the moat of your company.
Sahil: How do you support your robotics portfolio companies?
Han: It’s generalizable to a question of how GV supports our portfolio. GV (Google Ventures) is an independent multi-strategy fund. In general, we respect founders’ autonomy but will actively engage whenever founders need us. In addition, GV’s operations team has worked across thousands of companies across different stages and sectors. We actively help our portfolio companies directly or with our partners on key milestones of companies’ journeys, e.g. communication, fundraise, strategy, products, engineering, etc.
Sasha: Playground Global is a little different compared to most VCs. We don’t look like a bank. We are a big warehouse with a bunch of robots, rockets and jet engines. We have an operating team, engineers, and designers in support of our portfolio companies, but also for our own experimentation — to educate ourselves before making an investment. For example, before we invested in a hydrogen powered aviation company, we bought a jet engine and put it in our lab. We converted it from jet fuel to ammonia, and measured the thrust and figured out how that all worked, just to understand what it took to do the conversion, how different fuel types worked, what were the trade offs, and then we made our bet. So that gives us credibility with our portfolio companies to go in and not impose anything, but to come in as a credible advisor-consultant.
Xuhui: We [at Foothill Ventures] can often pull a lot of expertise from our different partners, as they all have prior entrepreneurship experiences. We have had more than $2 billion in exits from our partners’ own companies before, so we share our experiences with our founders, especially first-time founders and early founders. The top three consistent asks from our startups are fundraising help, recruiting engineers, and introduction to customers and partners. We try to invest in these three areas the most and give them the help that they need, especially in the early years. And I think robotics is a very cross-disciplinary area. We are especially suited for that because of all the different domains that we have the expertise and support in, so we can introduce and facilitate networking and partnerships for our fellow entrepreneurs.
Speaker bios:
Sahil Segal is a Senior Portfolio Engagement Director at Toyota Ventures, responsible for working with startups across frontier tech and climate tech and developing and implementing programs that add value to portfolio companies. Prior to joining Toyota Ventures, Sahil led startup scouting for Deloitte Consulting’s mobility practice. He has held roles at The Bill & Melinda Gates Foundation and served on an advisory committee of the World Economic Forum, focusing on the role of cybersecurity within startups. Sahil also co-founded TrekConnect, a platform that connected travelers to locals for hyper-customized vacations. Sahil received a dual degree in finance and Spanish from the University of Texas at Austin and a master’s degree in business administration from the Fuqua School of Business at Duke University.
Han Hua is a Principal at Google Ventures, investing in AI, infra, deep tech and fintech/crypto. Coming from a computer science and robotics background, Han has been operating in engineering and product roles for years before GV. He loves deep science and engineering problems and enjoys working with founders from Day Zero. Before joining GV, Han was an engineer and later a product manager on various teams at Google, including Payments, Blockchain and Next Billion Users (Google’s then-frontier-market group). He joined and co-founded startups on fintech, marketplace and crypto before Google. Han was also a prolific angel investor. Han studied computer science at Fudan University (B.S.) and Carnegie Mellon University (M.S.).
Sasha Ostojić is a Venture Partner at Playground Global where he actively invests in the areas of AI, nextgen compute, and robotics, while also assisting and advising portfolio companies on defining and executing technology and business strategies. Sasha’s Silicon Valley career started in 1988 at IBM as an early pioneer in voice technology products, leading to the role of CEO/Founder at Ultimate Technology, Inc., followed by a series of high profile senior executive roles at Nvidia, Cruise (acquired by General Motors), and Zoox (acquired by Amazon). Sasha sits on the boards at Allison Transmission (NYSE:ALSN), d-Matrix, Anari AI (chairman), as well as advisory boards at Zoox, HTEC and the Serbian AI Institute.
Dr. Xuhui Shao is the Managing Partner of Foothill Ventures, a technology-focused venture fund investing in early stage startups in the US. His investment areas are AI applications, enterprise software and full-stack sensor/data/algo companies. Before that Xuhui has been Yahoo’s Vice President II Engineering for data and advertising platforms; Founding technologist of Ad Tech pioneer company Turn ($310M exit) and AI risk management startup ID Analytics ($170M exit). Xuhui earned his BS and MS degrees from Tsinghua University, and his Ph.D. in EECS from University of Minnesota. He is an early investor of Plus.AI, Weride.AI and Certik — all subsequent unicorn companies.
Acknowledgements:
A special thank you to our co-organizers and sponsors: CMU T&E, EchoHer, Google for Startups, Deel, GenAI Assembling, and the photographers and videographers: Vril Zhang, Yanhe Chen, Jackon Gong.
This blog is written and posted by Yuna Liang, Summer Intern 2024 at Foothill Ventures.
Follow us on LinkedIn: https://www.linkedin.com/company/foothillventures
Follow us on Twitter: @FoothillVenture
Foothill Ventures is a $250M technology-focused venture fund located in the Silicon Valley. We back technical founders across software, life sciences, and frontier technologies.
Questions, thoughts, reflections? Let us know in the comments below. We’re always looking for great entrepreneurs and early-stage ideas, and we’re always interested in having a discussion about venture, technology, and anything related. To see more about Foothill Ventures, please visit our website: foothill.ventures.
