Jacomo Corbo, CEO & Co-Founder at PhysicsX: Helping companies design & manufacture industrial products faster

0:00Jacomo Corbo:Hello and welcome back to the Scaling Europe show. I'm Seb Jolson. Today on the show we have an amazing guest, Jacomo, one of the co-founders of PhysixX, one of the most exciting companies coming out of the UK and Europe. Jacomo, thank you for joining me today. How are you doing? I'm very well and delighted to be here with you, Seb. So for those who don't know PhysixX, maybe give a quick 30 second, one minute pitch on what you've built. So we describe what we're doing is building a new engineering simulation software stack, one that's purpose-built to bring AI to how engineering, manufacturing, production get done in industrials.

0:34And that starts really from changing what the common substrate for all of engineering manufacturing is, which is simulation, so physics and circumchemistry simulation. We're moving that from legacy numerical simulation to inference. And so now it's AI models that are able to do this, able to effectively predict the physics in ways that give us a few things. Like one, a massive acceleration over numerical simulations. We're talking about 10 ,000 to close to a million times faster. we're also talking about um being able to you know um train not just on a corpus of numerical simulation data but also on real world measurements so that's you know data coming from a wet lab from a wind tunnel from some other you know test dino um from a you know jet engine flying in the sky all of this is allowing us to have these models um you know supersede the accuracy of numerical simulation alone.

1:39So, you know, our ability to get to, you know, a fundamental and rigorous, you know, prediction of the phenomena that we care about is now tied to data generation rather than running through scientific discovery. And maybe the last thing is we're able to bring now, you know, high fidelity simulation to the whole product development life cycle. And so, you know, not just where it's used today, which is in detailed design, very much on the engineering side of things, but, you know, downstream of that into manufacturing, testing, verification, assembly processes into operations. And so that's a whole new frontier for the use of simulation and engineering and manufacturing.

2:23Jacomo Corbo:And you're working with some amazing companies. Perhaps could you give maybe one or two examples to bring it to life. And so what's a real use case that a company is using the technology for? So we work across a number of different industries. So we work in aerospace and defense, in semiconductors, where most of that work is centered on the equipment manufacturers, but we're also starting to get into the fabs. We're working in automotive and in processing intensive manufacturing industries. So think of commodities like lithium and copper, steel, aluminum. And so within that, you've got a range of different applications given the sector and given where across the product development or engineering lifecycle the technology is ultimately being used.

3:21And so let's say in aerospace, we're using technology to change how large casted parts are done. So this is really, you know, deployments into a foundry. And we're talking about controlling, you know, the casting processes, so the process parameters for how an individual cast is done in order to, you know, mitigate defects. This can be used upstream to change the actual, you know, to go towards even more aggressive geometries because now we have a better way of having those geometries be produced defect-free. We're working in aluminium smelting. What we're doing is controlling the actual smelting process, so trying to maintain as homogeneous a mixture in a cell bath as possible in order to reduce the amount of electricity ultimately required to produce some tonnage of aluminium, but also to try to improve the quality of that mixture and reduce defects again or improve yield.

4:32Jacomo Corbo:And things seem to be going really well. You very recently announced a$300 million Series C at I think a$2.4 billion valuation. This was, I think, less than 12 months in the Series B and over double the valuation of your Series B. So I wanted to ask, what's changed since you closed your B to closing the C? What's changed at the company that has kind of enabled you to raise such a big round at such a large valuation? um maybe like you know for a whole lot of other businesses as well that are raising large rounds and relatively quickly um some of it is tied to two things like one is um you know the um uh demand side of the equation um you know uh improving so um one of the things that we've seen is that our customers better and better understand why they need this technology.

5:27And so we're very much getting into a place, and this isn't just happening in one sector, this is happening across the different industrial sectors that we serve, including high-value industrial machinery. So think of a lot of the data center supply chain. So everything from you know what's happening inside the data center to what's happening outside of it including on the energy production so you know think gas turbines think you know chilling units you think step-down transformers um we announced earlier this year you know work that we're doing around data centers together with nvidia and and siemens um on on the whole what we're seeing is you know an inflection in demand which means that we're you know very supply side limited um we're in a position now where our customers have to wait close to six months or two quarters for a next deployment.

6:20So we need to right-size the organization in order to be able to meet that demand that we're seeing. This is global demand as well. So our footprint right now is very much centered on Europe on the one hand, where we have about 190 or so of our colleagues, about 80 people We're sitting in New York, in North America, and we're, you know, we're growing to become a more global organization. So, you know, in Singapore, in the Bay Area. One of the other big things is that over the course of the last 18 months, we've really established scaling laws for the pre-trained models that we're building. And so like in a lot of other domains, like for language models, for a lot of the models that you see in robotics, so that are enabling things like embodiment free control, you know, there are, you know, there are compute scaling laws, there are data scaling laws, and there are model size scaling laws.

7:20And so we've established data and model size scaling laws. The upshot is that larger models trying on more data are better at generalization, right? Their zero shot performance is better. And that's giving us the conviction to spend more on building larger and therefore more capable, more performant large physics models as well.

7:44Jacomo Corbo:And so going to your point around having this huge demand, you can't actually service yet. What do you need to do to be able to unlock or increase the supply? I mean, it's a great point. I think there are a number of things. In the first place, we have a relatively horizontal platform that's setting out to meet the requirements of a lot of engineering, manufacturing, production across a number of different industries. But there are things that we can continue to do to accelerate the time to value for our customers. Some of that hinges on more vertical modules, which ultimately makes it easier for our customers to onboard on the platform, to deploy it into specific applications that are going to be relevant to their own sector.

8:35You know, just think of use cases that might be, you know, different in an aerospace context than they might be, let's say, in a mining in materials or energy context. There's also, you know, a lot more that we can do to, you know, grow the organization to have a more global footprint. Every one of the organizations that we serve, or almost all of them, are global organizations. and, you know, they don't just have multiple business units. They are also scattered in very different parts of the world. We have to become a more global organization for us to be more effective partners to these organizations.

9:18And so, you know, I think it's a question of being right-sized. It's a question of building, you know, building out into what's ultimately a pretty large surface area of product that we have. And it's continuing to build scaling laws, which are giving our customers more and more of a warm start and ultimately, you know, eating into the complexity of individual deployments or at least individual applications.

9:47Jacomo Corbo:And when you talk about the customers and the demand that you're seeing, you are both pitching to and working with a range of companies, both from very large, older, established manufacturers and companies to more newer, agile companies that are trying to compete with those incumbents. Are you seeing a difference in both procurement and adoption across those two kind of customer profiles? That is a great question. So, you know, to be perfectly candid, you know, when we started, you know, and I would say this is more so let's say since the series a um you know which we we raised led by general catalyst in the fall of 2023 um we were very you know we've been very focused since on large um large enterprise customers uh but within that segment um you know especially um i would say you know sort of uh uh you know traditional manufacturers these are large engineering and manufacturing organizations like the uh they are the the primes or the incumbents uh within their uh different industry verticals and um you know we have worked with let's say new manufacturers so smaller organizations that you know uh are you know one smaller if only by market cap from these larger much larger uh organizations um but uh are you know are a little bit nimbler, are, let's say, you know, faster to adopt new technologies, might have, you know, slightly simpler procurement and vendor onboarding processes.

11:33You know, the reason that we've shied away from those in the last couple of years has been really on the back of, you know, understanding that some of the most high value engineering manufacturing is done by the incumbents in these sectors. What we wanted to do is to prove to ourselves, but also to investors that we could touch high value engineering, which ultimately is very high complexity engineering. We wanted to also prove that we, you know, could be relevant to organizations that were already very sophisticated with respect to their use of high fidelity simulation or CAE simulation. These are also the ones that, you know, I think best understand that the bottlenecks that we are addressing with the technology that we are building can't be addressed with numerical simulation.

12:26Like it's not just a few more, let's say, CAE simulation licenses that will somehow into the complexity of running 500-hour-long simulations and running complex design of experiments around that, or that somehow you'll be able to take some complex numerical simulation that takes, say, 500 or so hours and distill it in some way or reduce it to some form where it is relevant to a control loop. These are all the things that we are setting out to address, And it's the larger, you know, companies that have been, you know, best able to understand that value proposition. All of that said, we're seeing two things happen right now.

13:14One is that those incumbents and leaders within the respective industries are really starting to get it or really starting to understand that they need to invest more in innovation themselves and into the software that's powering the engineering manufacturing that they do. This is not a question of just, you know, leaner and better processes. Like there's been a, there has just been fundamental underinvestment in the software powering the highest complexity engineering and manufacturing. And I think, you know, across verticals, we're seeing, you know, those enterprises, you know, double down and invest more there because they know that they need, you know, they just need better software to power the hardware innovation that they're doing.

14:10The other thing is that I see neo manufacturers in, you know, this is true in a few different verticals, really starting to get to scale. And so we're not talking about, let's say, the small, you know, the market caps. And this is true in a semi-cap industry. It's true in quantum. It's true in fusion. It's true in aerospace and defense. You know, it's true in automotive as well, that there's there are a number of these new manufacturers that are very quickly getting to larger production volumes that are getting into manufacturing that are, you know, becoming, you know, major, you know, engineering and manufacturing organizations in their own right.

14:50And so we're very much looking at growing and partnering closely with those as well in this next chapter for us.

14:57Jacomo Corbo:And when you talk about the kind of incumbents, these large enterprise companies adopting this new level of technology and being more open or receptive to the need to adopt it, have you seen any difference geographically? I know that you've spoken to some extent about the risk aversion of European industrial and defense players. Have you noticed a real difference in the established European players versus kind of maybe the US as you start to expand over there? It's a great point. I think we're increasingly seeing a greater disposition and leaning into new technology and working with, let's say, smaller companies like ourselves.

15:39So the rhetoric is very positive, both from industry as well as from policy circles and government. And, you know, whether that's translating into, you know, greater momentum commercially, that right now is limited. We're seeing a little bit of an uptick, but we're still not seeing the rate of adoption and ultimately of spend that we see, one in the US and two in Asia.

16:10Jacomo Corbo:Yeah, that's really interesting. You know, you're building this country from London, from Europe. What's that experience been like? Do you think that's given you maybe an advantage in the sense that you've been thinking global from very early or has it been a disadvantage being able to like, you know, dealing with European enterprises? What's been the take so far? I mean, Seb, you just alluded to it in the question. So one, I think it's been a tremendous advantage to be building out of Europe, you know, in the first place for the talent pools that we're able to tap into. So, you know, we're, you know, we, 190 or so of our colleagues are in London altogether, including the European continent.

16:50That's about 250 or so. And so, you know, our roots and where so much of our human capital sits is in Europe. You know, Europe has tremendous, you know, depths in engineering, in physics, you know, producing the kinds of technical profiles that we're ultimately looking for. um it you know europe is also for all of its disadvantages with respect to you know you know some of the incumbent players being you know uh slow to work with smaller smaller you know companies like ourselves um it's also very much from inception forced us to think globally and And that, I think, is a huge advantage. We've always been reasoning about the opportunity, one, relatively horizontally, and that we knew that we were committed to bringing in the capital required to scale the company, not just to go after one individual vertical, individual industry at a time, getting to critical mass, moving on to the next sector.

18:03we very much think there's a land grab opportunity. That one cuts across a number of different industrial verticals, but both because of the different industries, including semiconductors. When we think about whether it's the fab side or advanced packaging, the equipment manufacturing space, so much of that has a global footprint to it that we've had to reason about the industries very much from the vantage point of how do we build a company that's building technology that's relevant to these sectors, but how do we serve the major players in these industries, which almost by definition, most of that TAM is going to be sitting outside of Europe.

18:56Jacomo Corbo:Got it. I want to touch on, I think you alluded to this earlier, the partnership that you've got with Deutsche Telekom and NVIDIA on the kind of industrial AI cloud. Can you talk a bit about what that is and what it means? Yes. So, you know, we've been working with Deutsche Telekom and NVIDIA for some time. NVIDIA joined our Series B in Extension late last year. the industrial AI cloud is, we're working with both on the industrial AI cloud in two ways. In the first place, we're using that infrastructure to be able to train some of our largest and most capable pre-trained models. So that's really training compute.

19:46But the other part of that is very much using that cloud in order to be able to mobilize effective training compute at our customers. And so all of our customers are very heavy users of the hyperscalers of Azure and AWS. And those are great places and effective clouds on which to be able to run production workloads, have large deployments. But it really doesn't address, today, how do you mobilize training compute for the kinds of models that we're building, so large physics models. And so we're working together to be able to bring that compute as software and compute to those customers.

20:45Jacomo Corbo:And of course, it's running out of time. I want to ask one final question, which is that we've seen this sort of like fundamental change in the way that software engineering is done. It's a very different job to how it was five years ago. In five years time, what do you think kind of industrial engineering will look like? What will the job or how different and in what ways will the job of an industrial engineer be? It's a great question. And I think working in a large engineering manufacturing organization is going to change substantially. Irrespective of whether we're talking about a designer, a simulation engineer, a systems engineer, a process engineer, a manufacturing operator, across the board, I think we'll get to a place where people are going to move up the abstraction ladder.

21:36You know, a simulation engineer that today is spending so much of their time, you know, setting up numerical simulation runs, manipulating meshes, doing the pre-processing and post-processing of data coming from these simulations. um you know being guided by these simulations to you know use what's effectively a you know a world model that they have um a mental model for for the you know the these these phenomena whether it's you know to do with one physics like electromagnetics or another like computational fluid dynamics um all these things run through very specialist expertise and i think we'll get to a place where, you know, these engineers or, you know, whatever technical profiles working with these companies are enabled to be able to operate end to end much more.

22:36They'll be able to, you know, really think about designing something and then all the downstream tasks with respect to, you know, simulation, with respect to understanding manufacturability, a lot of these things will be done less by handoffs and more by people working with the tools that they have, you know, to be able to get at the answers that they need so that they can, you know, ultimately make decisions, the decisions that they need to make, whether that's with respect to, you know, designing some geometry, whether that's looking at, you know, investigating the choices of different materials and trying to understand their suitability for different applications, whether that's, you know, controlling a castings or additive manufacturing or injection molding process like all i think the the jobs will continue to exist the world needs more hardware but you know i i think these a lot of these jobs will start bleeding into each other i think talent will become more fungible i think

23:38Jacomo Corbo:people will operate end-to-end much more so the 10x engineer but in but in yeah industrials well look jackman thank you so much for joining me it's been great chatting congratulations on the recent rose. I'm sure there's so much more work to be done, but it's great to see the continued success of the business. Thanks so much, Tom.