UKTN Podcast: Karakuri founder on chip-frying robots and microchip geopolitics
Things have been hotting up in the microchip world of late. Joining the latest episode of the UKTN Podcast to discuss all things chips – from semiconductor geopolitics to a chip-frying robot (of the potato variety) – is Barney Wragg, CEO and founder of robotic kitchen automation startup Karakuri.
Wragg founded London-based Karakuri in 2018. Its two main products are a robotic arm that serves personalised sushi dishes, and an automated fryer for chips. It has raised more than £13.5m in funding.
Wragg talks to UKTN Podcast host Jane Wakefield about his time at chip giant Arm during the early 1990s and shares his solution for the Cambridge-headquartered company’s IPO dilemma.
Elsewhere on the show, Wragg discusses the danger of semiconductor trade barriers, explains why he thinks being a founder is “part of your DNA” and recounts finding a product-market fit for kitchen automation.
During the 35-minute episode, the latest in a series of conversations with founders of high-growth UK tech companies, Wragg shares Karakuri’s plans for 2023 and explains why he thinks AI is creating more – not less – jobs.
Listen to the full episode here, along with all previous episodes of the UKTN Podcast.
A full transcript of the episode, which has been lightly edited for brevity and clarity, is available below.
UKTN Podcast with Karakuri founder Barney Wragg
Jane Wakefield: Hello, Barney, thank you so much for joining me today. I’m going to start off by asking you that question, which some people find annoying, but I think it is true that if you become a founder you do go on a journey. So tell me a little bit about your journey to becoming a founder?
Barney Wragg: Gosh, I think that’s quite a difficult question to answer. I think if you’re a founder is part of your personality or part of your DNA. So although I’ve been fortunate enough to do lots of things in my career, I always was and was always destined to be a founder. I come from a family business background, by the time I was 15-16, I was running my own business. When I went to university, I took time out to go and join a startup and never went back to university, kind of full-time worked through that. So I worked for a succession of startups that I hadn’t founded directly after university, had other businesses that I was building on the side doing things with. So I think it’s it’s sort of part of your DNA. It’s something you either feel passionate and comfortable doing or you don’t, I think.
JW: You mentioned that you started your career at other startups, one of which was Arm, which is often lauded as one of the UK’s most successful tech businesses. It must have been an exciting time to be at a company like that. Can you tell me a little bit about it?
BW: My journey with Arm starts from being the geeky kid with glasses at school. So I remember the day when I was at junior school when the first BBC Micro arrived. Our head teacher proudly unboxed this thing, but then clearly have no idea what to do with it. This is a period of time before schools had VCRs. So the idea that you would connect this to the television – and there were wires and things and tape recorders to connect up – completely foxed him. And I was just fascinated by it, and sort of ended up playing with it. And that begat my love of computing, and I guess of electronics. And of course, at around that time, I would say the early 80s I guess, as my school years progressed, you started to see the BBC become prevalent in schools, but my secondary school started to get PCs and Archimedes.
And so I became familiar – and we’d have these huge debates at computer club about whether an Intel complex instruction set processor was better than what was going on inside the Acorn Archimedes. So it was a thing I was kind of fascinated in. I went off to university and by the time I got to university the debate about whether an Archimedes was going to be a mainstream computer was over because it clearly wasn’t. There were just PCs and Macs everywhere. And I sort of almost forgot about Arm until I’d been probably out of university two years, was in a startup where I was unhappy with the way things were going and saw an advert for Arm in I think Electronics Weekly. And at that time the company was – that would be like 1995-1996 – so the company was three or four years post spinning out of Acorn, just finding its feet and starting to get some success. It was pre the big announcement with Nokia, which you could say was the tipping point for the company going from a small, very specialist, quite interesting-looking company to a mainstream business.
So to join the business at that time, it was I think there were about 40 of us maybe, maybe heading towards 50 by the time I joined, it was a very small company. It still had the 12 tightly knitted-together founders that were there. It was a very, very exciting time. It was a time that predates the Dot Com boom, just. It was a time that predates everything we now think about startups. So it was a risky career choice for most people in the electronics industry, in the semiconductor industry, to go and work for an unproven company like Arm. I think they were struggling to recruit. And as a physicist who had been doing DSP design prior to that, I looked like a very unusual hire, but they needed project managers, they needed people to help them as the company was growing. And it was fascinating, it was great to go there. It was an unbelievably open culture, an unbelievably focused organisation. It was a real can-do [culture], you were empowered to get on to do what you thought was right. When I look back on it, I think I was in my mid to late 20s, and spending more time on a plane, visiting customers and partners in Japan, Korea, America, than I was at home.
And I was given the autonomy and the ability to be able to do that. I think that level, that culture, that level of trust, combined with the intellectual horsepower that was inside the business, both in terms of the technology we were developing, and the business model that we’d worked out, they were the seeds that it pre-IPO forced us to a fantastic IPO and the incredible growth that went from there. And so it’s a time of my [career] I think I was there four years. So a relatively short period of time in my career, but I managed to achieve a huge amount in that time. There’s a group of us that are still very, very close friends, all these years later. We still see each other socially, important life events are all celebrated together. And it’s a time in my life, my career, I look back on for a lot of learning and with a lot of happiness and enjoyment.
JW: And what do you think of what’s happening in the chip world now? We’ve been seeing these massive shortages, and we’re also seeing the chip world, and the wider tech world, being pulled into geopolitics. The divisions between China and the US. What’s your thinking about that?
BW: I think the supply cycle is always cyclic in semiconductors. It’s almost impossible – the amount of investment that is needed to build state-of-the-art fabs and production facilities, it’s always going to mean people are cautious. You’ve always going to have a take up for the design of chips to use those fabs to get the most efficiency, then people are going to see those benefits. And then there isn’t going to be enough capital in the fabs, there’s going to be a shortage of fabs as there are more chips that go through. So I think we’re in one of those waves, which has been made more difficult by Covid, more difficult by the macroeconomic trends. I think we are seeing – the political, geoeconomic stuff around where stuff’s manufactured, who has control of it, what’s happening with design rights – is a function of a change in the industry that happened probably about 15 years ago, when the cost of fabs became really, really high.
And the only people investing in manufacturing were the Chinese and Southeast Asian. So you’ve got a dearth of manufacturing capability in Europe and the US. And when I look at my time in the semiconductor industry, that was starting to happen, but it was still clear there were major wafer fabs in Europe, in North America, in other territories. And that’s very much centralised around Taiwan, China and, and some extent Korea now. So I think, yeah, when you get that much centralisation in any industry that much consolidation around certain geographies, it’s going to make people uncomfortable. How do you address that? Well, you’ve seen government intervention with subsidy and funding to redress that balance so that you could have US-based wafer fabs, maybe European based-fabs. I think it’s a natural readjustment. I think the small-mindedness around trade barriers and the political system is dangerous. I wouldn’t profess to be an expert on it. But one of the phrases we used to have an Arm was ‘think global, act local’, and we very much built a global business with international partners from different cultures, different parts of the planet. I spent my life travelling to see those people and work with them. And I think that was an enormously positive thing. I think it’s very sad that we’re in a political environment now that makes that much harder.
JW: Last question on Arm, it’s in the position where it’s looking to IPO later in the year. And there’s a bit of a battle going on between London and New York as to where this should happen. Any thoughts on that?
BW: I’m no expert on the Arm IPO situation, I wouldn’t profess to be. But it does make me smile that people forget we did an IPO before – this happened at Arm. And in those days, it was a bit like there weren’t really any IPOs for UK companies, it wasn’t what happened. Were we going to be LSE listed, or will we be NASDAQ listed? And we solved that problem by having a dual listing. So we simultaneously listed on LSE and NASDAQ. And I do kind of look at this sort of debate that’s going on and kind of going, we might have solved this problem in the past. I wouldn’t know all the details. But yeah, it does make me smile when I look at it.
JW: Similarly I guess to the fight between China and the US we’re perhaps not learning from the mistakes of the past and moving on. But let’s talk a little bit now about your move from the complicated world of chips to the some would say even more complicated world of robotics. What encouraged that move? What made you decide you wanted to get into robotics?
BW: I guess the point inbetween Arm and Karakuri I spent an awful lot of time in the entertainment industry. Which was really started by the work I was doing at Arm. A lot of the work I was doing at Arm was around early digital media, in particular MP3, and how that would work. And that kind of exploded and post-IPO at Arm I spent many years in major record companies and entertainment businesses dealing with digital distribution. And I think after 15 years of doing that, in large companies at a very senior level, I felt I was missing that startup thing that my career had really been all about. And back to that idea, do you biologically think you’re an entrepreneur and a founder, I sort of felt like I was and it was being sucked into a world of much more, sort of private equity, CEO managing existing companies, dealing with turnarounds, than creating things from scratch and growing things. And so I had an itch to go back to do that. But I didn’t know what so I spent some time after I last left my last role in the entertainment industry, working with some VCs just sitting in their deal meetings trying to help their portfolio companies, looking at some stuff. I also had a group of friends just because I eat too much in the restaurant business, who all had restaurants who were saying ‘could you come and help me, my business is failing, I don’t understand this’. So really out of interest I got involved in, on a pro-bono basis, looking at some restaurants.
And again, the same thing that happened at Arm, looking at the projects that we got involved with, the geek in me looked at restaurants and went ‘this is essentially a manufacturing plant’. We have raw materials coming in on one side, we have orders on the other side. And we have to assemble and put these things together. We have a kitchen, that is our factory, our manufacturing plant. And the number of errors of mistakes and problems that were happening in that process would mean the difference between a restaurant with a high turnover, losing money, because it wasn’t an effective manufacturing plant, and a restaurant with a low turnover, making money because it was being run very effectively. And it was really down to the quality of the staff and the people that you could have in there. So that made my brain start to think well couldn’t we start to bring in some of the sort of ERP and automation technologies that you’ve seen in other manufacturing industries to prove this. And that was the genesis of Karakuri. And I was just starting to really look at that. Obviously looking at IoT, distributed computing startups, a little bit of AI stuff that was kind of happening, looking at a bit of automation, mechanics, electronics just because they are interested. That was about the time Brent and Henry were starting Founders Factory. I think their first LPs into Founders Factory, one of the ideas they said is that we’d like to get involved with robots and particularly robots in hospitality. So that is really worth kind of two things came together, and reconnected with Brent and said, ‘look, I think there might be something in this’. And that was how the germ of the idea of a Karakuri was born.
JW: And Karakuri he name is an interesting one, isn’t it? It’s a Japanese word for a device that moves an automaton, which we’ve seen way back in history. What made you come up with that as a name?
BW: Well, originally, one of the hypotheses that we had for the business was that if you go to Japan you can get really great quality sushi for a not-very-high price. They call it ‘salary man sushi’. Go to a railway station, get great sushi. The sushi in the UK is not great, particularly sort of pre-packaged or supermarket sushi is not terribly great. Certainly not the quality you get in a railway station in Japan. So one of the reasons we looked at that was, well, maybe we could use automation to make sushi fresh on demand, and we will use robots to do that. So we started with sushi as our original hypothesis, we were looking for a name for the business, and this Japanese word karakuri, which is colloquial Japanese now, when they’ll say when something is amazing. Or there’s an automation or mechanical process, which is awe-inspiring. We thought, well that’s quite a good fit for the name of the company.
JW: And tell me a little bit about your plans for 2023. You’ve obviously raised quite a lot of money, where are you at the moment? And what do you see as being the next stage?
BW: I started Karakuri in 2018, when you were just starting to see stories about Flippy the burger flipper in California. I think a lot of people were looking at industrial automation. So PLC controller standard robot arms, and going ‘oh, well, maybe we’ll just whack these into a restaurant, see what happens’. There was a lot of excitement about that, but not a lot of practicality around it. We spent a long time inside Karakuri really looking at the product-solution-fit and the need in the market. Then the pandemic was a really interesting time for everybody in our industry. So a lot of the smaller startups doing things with standard automation failed, just didn’t have market fit, didn’t have a solution, couldn’t really have anything that was scalable. And obviously, it’s a very, very difficult time for the restaurant industry. So a lot of us had early-stage proof of concept products on the shelf ready to deploy into restaurants, even had deals with restaurants, that kind of imploded as 2020 saw the hospitality industry close. But what 2020 also did was as things reopened, particularly for takeaway, it highlighted what do you want to call it, the great resignation, or the shift in people’s working patterns.
I think the work-from-home thing, the number of options that people discovered that there were to make a living in the non-traditional spaces, has had a hugely important impact on the hospitality industry, particularly the quick service restaurants, or if you want the fast food chains, like Burger King, McDonald’s, KFC, Nando’s, any number that you might come to mention, of finding recruitment almost impossible post the pandemic. And so we’d identified this, we were working on some solutions for those industries, particularly on frying, taking me back into a different type of chips if you like. But the type of frying, open basket frying that they’re doing in QSR restaurants. It’s a horrible job. It’s the least popular job in a restaurant and the most difficult job to recruit and retain for. And it has the biggest impact on consumer satisfaction. So we’d started to do some work on that pre-pandemic. And as a result of the pandemic, we just saw huge interest in that. What do I think that says for Karakuri in 2023? Well, our product is just coming to maturity, we’re just going into restaurants now. We have a series of announcements coming across this year. I think this is the year that restaurant automation goes from blue-sky thinking and university lab prototypes, to serious pieces of commercial kitchen equipment that are installed and operated by significant numbers of restaurants that immediately see an impact on their P&L and the operation of their businesses having them in there. So a long way of saying I think 2023 is the year of maturity for Karakuri and also for the wider restaurant robotics industry.
JW: And excuse the food-based pun, but can give us a flavour of what those announcements might be? Can you give us any little tidbits?
BW: We have spent a long time making absolutely sure our product is right and stable and meets the requirements of some of the biggest QSR, fast-food restaurant chains in the world. And across 2023, we will be installing and working with them. So you should see some of those announcements happening this year.
JW: I once did a comparison, I had a cocktail made for me by a robot arm, and I had a cocktail made for me by a human. And the one made by a robot – it was a few years ago now – was a lot messier. And it just didn’t taste as good. At the time, it made me realise that actually, even with quite a simple job, which is measuring out spirits and adding to it, you do sort of need the human in that chain. Presumably wouldn’t agree when it comes to cooking? You think robots can do as good a job?
BW: No, I would agree. And I think this is where there’s a great deal of misunderstanding about what we do and where automation will work. And I think that is happening on several layers. So firstly, I would say I think I know where you will have had your cocktail and the particular robot that’s involved in that. And I would say that, while that’s a really interesting showpiece and has been used to generate footfall and a lot of conversation, robot arms are not best suited for working in kitchens. So traditional industrial robot arms of which will have been used to make your cocktail and are used in many of our competitors’ products, are usually designed for the automotive and electronics manufacturing industries. In those industries, they have a different set of requirements than you would find in a commercial kitchen. So typically, manufacturing plants are out of town centres, large areas, and floor space is relatively cheap. To get the speed and efficiency, the robots use a lot of force and power in their motors. That means that they’re dangerous to operate in close proximity to people – well, that doesn’t matter because you just put a big cage around them because your floor space is cheap.
If you’re running a McDonald’s or a Burger King or a KFC, you’re typically in a town centre, very expensive floor space, and your kitchens are absolutely tiny. And people have to work in very, very small proximities. So this is a long, again, way of saying I think you have to choose your automation for the application. And I think if you wanted to make the perfect cocktail automatically, I wouldn’t use a robot arm to do that. There are better automation solutions to achieve that output. And that’s very much what we do at Karakuri. If you look at our fryer product, it doesn’t use a robot arm, it doesn’t really look like a robot, it looks like a large piece of kitchen equipment. It’s actually a whole massive connected together, if you like miniature IoT-based robot that is coordinated to be able to achieve the perfect process inside a kitchen. Now, if you also look at our fryer robot, it has been designed to do one of the most boring repetitive tasks in a kitchen. We take frozen fries that you have to portion accurately into a basket, move quickly into oil for a very defined period of time, take them out, drain them, tip them up, and have them ready to serve. It’s a very linear, straightforward process. There’s not a lot of artisanal involvement in that process. It’s an almost mechanical process that a human being has to do. There are a lot of those tasks in a kitchen. Whether it’s peeling and chopping vegetables, or frying chips, there are a lot of very repetitive straightforward tasks. Those are the tasks that we think are most suited to automation. Those are the tasks where you aren’t really applying the value add of a human being in doing something.
We don’t think that robots will replace humans in restaurants, or in kitchens completely, because there is an enormous amount of value add that’s needed in certain cases. Humans have a very different role in a QSR kitchen than they would do in a Michelin-starred kitchen. And so, I think the way we see the world in Karakuri is to find the tasks that make sense to automate. Find the automation solution that will allow you to do that in the most effective way, i.e. don’t bring in a third-party solution that might be good for automotive manufacturing into a kitchen and expect it to work that way. So pick your problem and pick your solution in a way that adds the most value. And understand that in so doing you can actually free up people to do the things that are more complex, more artisanal, require human interaction require thought and consideration in a way that, for example, standing in front of a frying line doesn’t.
JW: And on that question, I’ve also seen a lot of robot demos, and I don’t think I’ve ever seen one that hasn’t gone wrong. And it is hard, isn’t it, even if you’re doing a very precise task and not asking a lot of your automated machine, it is hard to get it right. Is that something that puts your customers or potential customers off?
BW: There’s a lot of concern about that. If you go into a fast food kitchen, a QSR kitchen, it’s entirely optimised. And everybody in there wants to make sure the customer is happy. And if you get an unhappy customer because something’s gone wrong, and their meal isn’t there on time, it’s a horrible place for everybody to be. So what we’ve had to do and why we have been very careful with the development design of our fryer system is to build into it intelligence around what its fail-safes are. What could the failure modes be? And how do we design a system that moves away from that? And you have to do that in what’s going to be a very autonomous environment.
In a production line for a car manufacturing plant, if a robot goes wrong, the value of the product going down the line justifies a shutdown of that line, and an engineer will come out of their office, fix that problem on site and go. You can’t do that in a QSR kitchen. You have to have something that’s going to be autonomous and operate remotely and not fail. And if it does fail, you have to have a defined recovery mode to be able to deal with that. So a lot of what we’ve designed at Karakuri compared to what you’ve seen, with people making burger flippers, or cocktail makers with industrial robotic arms, taking factory technology into restaurants, you’ll see what we’ve done is completely different. And it’s about the speed you can operate in, the size you can operate in, the cost you have to deliver at, and the reliability and robustness of the end solution that you’ve got to provide.
JW: Now there are two big questions that people come up with when they think about a future that involves robots alongside humans. And the first, I guess, is more about automation. And it’s the simple question of whether automation, robots, whatever form they may take, is going to steal human jobs. Now, you’ve sort of touched on that already. But I’m just looking to get a little bit of an expanded answer on that.
BW: I think this is always a wonderful question to ask me and to answer for people. Because I always say, do you think you are more or less busy in your job today than you were 20 years ago? And I’ve yet to meet anybody, apart from those people who’ve retired, who say, well now I’m more busy than I was 20 years ago. And I go, how much automation do you carry with you today, compared to 20 years ago? 20 years ago, maybe you had a Blackberry, but you certainly didn’t have a portable web browser, you certainly didn’t have the ability to do all the things you can now do with an iPhone or an Android phone in your pocket, the interconnectivity that we have with PCs, cloud-based services.
The amount of automation we have in our lives has increased exponentially. It hasn’t taken away jobs, it’s actually making us all busy and creating jobs. And so you’ve almost got the argument that Luddites made around what was the invention of the plough going to do, are we going to take away all jobs, there’ll be no industry. I just don’t see that as being the case. As a species, we’re very good at finding higher-mode tasks and more important things that we want to use our time on. It’s our aspiration to do that. So if we can use automation to do the jobs we really don’t want to do, we will find more things to do further up the intelligence stack. And if you look right now, we have customers who are coming to us at Karakuri and saying, ‘we are not interested in reducing the number of people that we employ in our restaurants, our sole interest is increasing their terms of service’. In restaurant kitchens, it’s not unusual to have people who stay for between six and 12 weeks. You’ve got restaurants recruiting on average, sometimes six to eight times a year to replace those roles. That’s a very expensive process. And it has a hugely negative impact on morale and the end of product quality. And the reason for that is the job is frankly, boring, hot, smelly, a bit dangerous, not very interesting. You take that away, you can give people a more fulfilling job, they’ll stay longer, and it’s better. It’s cheaper for the restaurant group to have people in jobs for longer, more rewarding for the employee and ultimately, you get a better quality of product at the end of the day.
JW: And the other question that always comes up is whether we will see increasingly robotics combined with AI. We’re seeing companies like Boston Dynamics moving towards robots that really look and feel like people. Do you think that in the future, we will combine AI, which is also making great strides, with robotics?
BW: I mean, we are doing it today. But I think it’s important to be clear about the definition of AI. An AI that can chat with you or talk to you is very different from the types of AI that we use in our robotics today. So there clearly will be sort of humanoid robotics that are important for certain applications. You can see all the hype around chat at the moment and people going, can we take customer service jobs out? I saw Martha Lane Fox piece in The Times yesterday going, if I was a lawyer, I’d be very worried about what I’m going to be paid for, for advice, because this chatbot is so good, it can do these things. Clearly, there’s an important element for human interface around something that looks familiar or acts in a familiar way and speaks to us or provides text to us in a familiar way. I think the real win for us with AI inside robotics is more about how we get truly smart physical systems that are taking in lots of information about the environment around them, and making better decisions.
I’ll give you an example. A good friend of mine was involved in an AI project to create a smart fridge. And the theory behind this was they put cameras inside the fridge, it would look at the contents of the fridge, and it would be able to recommend things you might want to prepare and cook from what’s inside your fridge. All brilliant. Everybody who went down this route ever got very excited. Lots of sponsorship. Lots of endorsement, lots of investment. What they realised was it – sounds trivial – but they realised that a lot of the contents of your fridge are impossible to inspect visually. So if you’ve got a tub of margarine, or a pot of something, it’s not good enough to know that you have that pot, you need to know how much is left. So you need to have smarter sensors. You need to start to have ‘oh, yeah, I’ve got that margarine. But how much is in there? How much does it weigh? What does this need in this recipe? What are the other factors I might need to bring into play?’ Those things, I think, are where we will really start to see AI combined with things that are – once you take an AI and you connect it to its ability to measure, interpret, and understand the physical world around it, then you can make really smart automating systems. So systems that don’t crash, robots that don’t break because it expected to pick a glass up and the glass wasn’t there and it doesn’t know what to do and it spills vermouth all over itself. All those things go because it’s much more aware, it’s much more able to sense and it can therefore make much better decisions and judgments about how it should operate.
The UKTN Podcast is sponsored by Deazy, a tech build platform enabling cost-effective, flexible and scalable development services.