The Future of CFD: Five Key Trends to Watch

0:00 Hi, and welcome to the Neil Ashton Podcast. In each episode, we explained some of the fascinating ways that science and engineering are changing the world around us. We talked to leading engineers from elite level sports like cycling and Formula One to some of the world's top academics to understand how fluid dynamics, machine learning, supercomputing are bringing in a new era of discovery. We also hear some of their life stories, their career advice, the lessons they've learned on the way that I hope will be helpful to you too. So sit back and enjoy this episode. Hi, and welcome back to the Neil Ashton Podcast. So today's episode is just me

0:45 and I wanted to go through what I thought were five of the top important trends in CFD, let's say over the now and over the next 5 or so years. I'd love to know whether you agree with what I'm going to say. These are sort of more, I guess, high level trends rather than getting into the, you know, specific details. Very happy to do an episode diving a bit deeper into one of these, but I thought I would maybe stimulate some discussion and get you thinking about some of these. And I have to also give a, a bit of, I think that I'm coming from a certain angle here. I guess my experiences in everything I've done will bias me.

1:31 And I just want to admit that that bias. I come more from an external aerodynamics background in terms of CFD, you know, I currently work for a cloud computing company. I'm obviously going to see things through the conversations I have and the lens I see, but I'd like to think that I see enough of what's going on in different sort of verticals and and different areas that gives me a reasonable understanding of what's going on. So yeah, let's see if you agree. So yeah, five things that I think are important for CFD over the next five years or so, not necessarily in chronological or like in in important order, but number one for me is GPUs and based processes.

2:17 Why do I say this? It was the case when I let's say back in 2013 or something like that, I remember going to a path CFD conference. So this they still run parallel CFD. It's a great conference to go to sort of that blend between high performance computing and CFD. And I remember there like 10 years ago, this was at Barcelona Supercomputing Center, really great venue right on the Ramblers sort of strip. Anyone who's listening who went there will know it was great conference organised by the Barcelona people and I remember seeing presentations showing about GPUs and I can't believe that was like 10 years ago now. And there was a real sense in the room of I remember someone

2:59 said that you can only get good performance out of it if you write the code from scratch. I remember someone saying that and there being a bit of a debate and I remember there being sort of sensational claims, you know, that GPUs are this much faster and, and a little bit of a, oh, you know, ignore them sort of thing. And obviously at that time NVIDIA was doing a lot of work to invest in in CUDA and sort of giving funding to research groups. And I remember seeing this and yeah, it, it, it was, it registered on my radar. But a lot of people were, were, were debating whether it could be possible, whether there'd be

3:42 enough memory on the GPU's. Was it really possible to rewrite a code? You know, people said they had millions of lines of code. I'm never going to be able to do this. Fast forward to now and I would say for me it's incredibly clear that GPU's are going to be the the dominant computing platform for CFD. But as I'll say later, things move very slowly, especially the industrial CFD. And so the adoption is going to take time. But I have seen enough to make me believe that and I'll give for several reasons. One, there are published publications out there that show

4:32 it. I've certainly showed it in public settings and it's the Isvs and software companies have done it. And in fact just for auto CFD, this automotive workshop that that I created back in 2018 in 19 with Gary Page Loughborough and now it's in its fourth edition. I think the slides are now up and the video is on the website. So if if you go to autocfd.org and you Scroll down and you go to the HPC presentation, Herbert Owen from Barcelona Supercomputing Centre did a great job of taking everybody's submissions from CPU and GPU and then looking at what type of GPU, what type of CPU, and doing a really good, good job of looking at several different

5:14 metrics. So the speed, you know, per degree of freedom, per time step, looking at the cost. So we took a sort of a notional cost from AWS. It could have been any cloud provider. It's just, you know, it's the only way of getting an actual price for these things because obviously the cloud gives you the all in cost. Whereas if you were to find a website that would give you just the sort of the GPU or the CPU that wouldn't be the actual entire cost including the power and the the network etcetera. And what it showed is that for external aerodynamics of cars, GPUs were more cost effective, at least 23510. It depended on the code or even

6:03 more cheaper than an equivalent CPU at the scale that it was being one for a realistic problem, not sort of toy benchmark problem. And they were faster, but that's a slightly harder one because it depends what the parallel scaling of the code is and and you know, how many CPUs or GPUs do you throw at it. Whereas the cost, you know, obviously it's going to change a little bit on linear regime. You know, if you're really badly scaling, the cost is going to go up relative. But essentially what it showed was the cost. Now I have repeated this same thing myself personally with startup codes and ISV codes. And I have seen the same thing, which means that for a certain

6:49 classification of workloads and I can only speak what I've seen and personally, so I'll speak more, let's say in plain design, car design, but I've seen it and I believe it to be so in other areas. It is. You're ultimately going to run your simulation cheaper on a GPU than a CPU. So forget about speed, it's going to be cheaper. And why is it cheaper? Ultimately, these are things like power efficiency, energy efficiency. So even though the GPU looks more expensive to buy for one node, when you actually look at how long does that node need to run for, you know, it's easier like I said, to calculate with cloud because it paid, pays you

7:27 good pricing, it is cheaper, and as soon as something is cheaper economically it makes sense it is going to gain widespread industrial adoption. Now where's the new ones to this is what GPU there is in some ways a divergent, well, not divergent, but there is a, there is a market for machine learning GPUs with a particular focus on, you know, certain precision, lower precision ultimately than what most CFD codes would need. And so you're going, you know, from A1 hundreds to H1 hundreds, H2 hundreds, etcetera to Blackwell that are driving up, you know, more and more performance.

8:15 And CFD absolutely can take advantage of that performance. And it's great because there's a, we can jump on the bandwagon essentially of machine learning. But there is also another classification of cards that quite often are more designed for inference, but these sort of L40SR TX type cards that actually have a very, very good price performance. So that raw performance may not be as high as those sort of more top end GPUs, but for CFD, they actually are often when you do the maths, even better price performance. And that's what we showed of this, this auto CFD. Now the reason I'm saying that I think the GPUs will become the

9:02 main is because if you look at the trajectory of GPU development in terms of generation after generation improvement and you look at the same on the CPU front. I think it's hard to deny that again, because of the investment from machine learning, there is a much greater boost in performance from generation to generation on GPUs than there is on the X86 CPUs. And the fact that there is rising competition now between NVIDIA and AMD. And also maybe there's room for more, you know, startup or emergent technologies that's only going to drive more and more focus. Now, I say ARM 64 as well because that has emerged as a, at first maybe more of a niche

9:54 architecture that was harder to get people to convince. And I know speaking from being AWS, we have our Graviton ones and you know, we struggled for people to think, well, why should I bother to pull my code, you know, to, to harm. But now with more and more companies with, you know, the other hyperscalers also releasing obviously with video, with the Gray system sort of linking the two, there is far more and there's far more reason for a major code to port to it. And as soon as that happens, the ecosystem jumps behind it. And again, the the power efficiency and the performance numbers make it extremely attractive.

10:35 That's not to say that people aren't going to be running on, you know, AMD or Intel CPUs, of course. And as I said, the CFD market and engineering in general or any use of CFD tends to be slow. And so if from 2013 to now there's been that sort of movement, it's not going to happen overnight. But what I can see is every single one of the major Isvs has or is porting their codes. The issue is that not all the features in the codes are available. So yeah, some of the multi physics, some of the options are not yet done. So you may not be able to use the GPU version today because it may not have the feature that you need if you're doing some

11:21 more sophisticated modeling, but it's probably on the road map. And more importantly is the codes. And these are more the start-ups like luminary or Volcano or Flex compute. And there are others who are really who have really focused on a GPU native code. And in the case of, for example, Volcano taking advantage of GP US like designing it for GP us Cartesian methods and there's boundary, etcetera. And there are others as well doing this and, you know, hiring super talented people who have come out of universities and institutes with a deep, deep understanding of how GPUs work. Programming it from the base from the beginning with the latest, you know, programming

12:09 languages and and and knowledge of how to extract the maximum performance is really only going to accelerate it further. And I'll talk about this in the last point. This emergence of new start up codes is also invigorating because simply, if you look through the eyes of a commercial company, why change? If you are the main company, main ISV provider, why would you change? You're earning good money from your licenses. You are you've got customers and if your competitors are all basically doing the same thing, so none of them, let's say five years ago, six years ago, OK, none of them are moving to GPS. Well, why should we spend all the effort doing?

12:56 Is it really that important? Is it really going to make a difference? Are we going to lose customers from it? And of course the technical reasons, you know, was there enough memory? Is there enough performance to gain from it? The one thing that drives a company to change is competition. And so with the emergence of alternative providers, many of whom who are specifically marketing their code around being GPU performance, it forces companies to do it. And you've seen that with, you know, with ancestors sort of rewriting their their solver and you know, showing a very, very strong performance. And at the auto CFD workshop, we saw pretty much all of the Isvs

13:40 specifically focus on their GPU performance and their speed, which three years ago was not the case. So everything feels like the right momentum. Now the tricky thing is the whole ecosystem needs to to work. So is are the third party meshing tools going to work on GPS? And I mention this because we get to another point, which is around the sort of high performance computing, because obviously if you're going to procure a cluster, if one code works amazing on GPUs, but all the restaurant CPUs, you then need to figure out, well, what's my ratio of compute options to buy? And, and so this links it to one of the third points later that the other reason for GPUs is

14:29 national supercomputers. OK, what do I mean by that? Well, if you are, let's say, in Spain and you want to show to your government that you as a country are pushing the boundaries of science, of R&D, and you're attracting entrepreneurial entrepreneurialism, supercomputing has traditionally been a great way. It's a sort of, you know, a big lighthouse. And the US is a more obvious one for that. But I'll pick a country other than the US, your supercomputer obviously a lot of that is then based on its performance and the top 500 list is often you know the metric like what is the total performance of the system. You know the the sort of exaflop

15:18 race to try and get an ex scale computing. A lot of this has driven to GPU's because the total performance of the system is much higher for a given power output with GPU's, and arguably it is also newer technology. So more novelty, more differentiation can enable better science because there's more, you know, throughput can be produced. And secondly, even more importantly, secondly, is because of machine learning, because machine learning is now being integrated and is a key research topic as well. And so if you're a national super computing centre, you want to be able to do ML research and you want GPU's. What does that mean?

16:01 It means that if you're an academic and you want access to these supercomputers to do your research, like people do in Spain, let's say, you will have to write your code to fit on the system. And so sure enough, that motivated researchers in Spain. And now let's broaden it to the UK, to the US, to port their code to GPUs. Essentially, they can run on the systems that they need to. And so there's been a push and pull. There's been a push because they think it is the right architecture to deliver better performance. Or maybe that's the pull. But there's also a push because the system they want to run on has GPS. So I should say that has also been the drive towards GPUs.

16:47 There is a, once a decision is made that the big supercomputer are going to have GPUs, you almost are then having to, well, I'll just put my code, you know, to, to go to it. And so that's more the balance to the industrial side and the academic side, why both are now meeting. And again, remember, many of these startup codes emerged out of more academic or government work where they had moved to GPUs because one, they're just more aware of the the science. And I would say this is what companies like NVIDIA have also put a lot of groundwork in investing in helping people to understand CUDA and helping them to pull their codes.

17:26 And that has now moved through into the startup. And as I'll speak about later, almost inevitably many of these startups are going to be acquired by the larger companies. There's an interesting sort of ecosystem and flywheel effect. So the next one is on AIML And you know, again, you're, you're typically really for AIML or you're really against it. This tends to be the case of any new technologies. The reason I'm more bullish on AML and that's through my own personal experience actually, at least I would encourage anybody else on, you know, looking at code, writing code, doing experiments, publishing, organizing, really trying to immerse themselves.

18:13 And I would actually make it similar to the way that I saw GPUs back in, let's say the 2000 and 10s, where again, I saw a similar dynamic of some people really being against it and saying, oh, it's just all hype. You know, there's no way they're going to be able to do it. I'm not going to be able to pull. It's not, you know, the technology is not going to be there to others who are like, it's just a matter of when I see the AI in a similar way, but with some more nuances, which is what exactly can AI be used for within CFD? Well, the first point I would say is the commercial pull. And you know you may if you come

18:55 from a more pure background, you may hate this sort of speak, but this is just the way the world works. When the large ISV companies, many of them who are stakeholders or boards to please people invest in their companies in the current climate because of perception of how much are they embracing and leading when it comes to innovation, particularly around AI. And so a major ISV has to show to their board and stakeholders and shareholders that they are fully embracing and are doing something about it. And I would say in some ways, many of them have had to catch up. Now, I don't know that, you know, working to the companies, maybe they've had many proposals

19:37 beforehand. But certainly I would say the whole ChatGPT moment has, you know, caught some of them off guard and made them really have to double down and decide what to do and what and if they can release it. That's means that there's a lot of sort of R&D and focus within these major CFD and CE companies. But this is also fed a lot of startup market. So for a similar reason that many of the startups and there are many many to list, you know the Navistos, the Physics X beyond math. I'm probably going to insult people by not mentioning them. I can't think I'll probably list them more out newer concepts. Yeah, I can't remember.

20:25 There's, there's loads of them. So I, I, I'm sure I'm offending some of them. The point is that or keyword as well. And yeah, OK, I'll try and make sure I mention most of them so I don't offend people. They have got VC funding because a lot of the VCs are seeing. Could this be the next big thing? And so there are to my Alaska at least 10 start-ups looking at doing AIML specifically for CFD. And quite often for things like I spoke about card design or plate design. The fact that they have a lot of money does not necessarily mean that it will turn into success, but it does show that there's a lot of concerted effort. And when there are hundreds of

21:05 people or thousands of people collectively working on this, that often leads to advancement. And when I talk about advancement, I typically mean in terms of surrogate modelling. So what some people would call reduced order modelling, but obviously more sort of sophisticated version of that were essentially, and This is why it's particularly interesting from an ISV or commercial point where you may essentially not need the CFT solver during the solve step. So you train a model on some data and then once you have the model, you do inference and that's completely outside the CFT code, which obviously is worrying for an ISV that thinks

21:43 we could be losing some of our, you know, potential license or revenue or, you know, customers because they could be trying to use this other tool that we don't have. So and just speaking commercially now that means those IS VS really need to figure out is this an important technology that could transform? And if so, we either need to develop it ourselves or we need to go and buy a start up. And that's really the the key thing that's driving a lot of these companies to then, you know, have a look at these start-ups. Why economically does it work as well is because on the one hand, the end user just as GPUs. So, so what's the reason for the end user to be interested in

22:28 AIML for GPUs? The main reason was lower cost computing and potentially faster computing, you know, GPUs if if the pace continues, you know, we're going to see potential of the next five years 5 to 10X speed up in in stuff. So it not only is a lot faster and also talk a minute about this leads to more like high fidelity transit methods to become possible even in a day or less than a day for AIML. It's that the inference can be both cheap and fast and that is something that brings up a lot of potential for for companies. So for example, it means that real time CFD has long been spoken about and GPU's were sort of billed as potentially

23:17 enabling real time CFD if you go back 5-10 years ago. But I would argue that that's not really turned out to be the case because some problems are not. They can be parallel in space, but maybe not parallel in time. So doesn't matter how much compute you throw at it, sometimes it just takes a while, particularly for transient methods which you may need to do for accuracy purposes. OK you could do maybe a Rand or lower fidelity, but truly real time click of a button instantly get the answer. You have to make so many accommodations on accuracy that it ends up really not being worth the sort of the balance. The AIML is the interesting bit because the inference itself

24:01 could be real time, but you could spend a lot of money in computing time on training a model based on high fidelity data that could, if the ML model was accurate enough, give you quite a good answer in real time. And so this opens up a lot of potential, particularly in bringing CFD closer to the designer, closer to digital twins, more into, you know, like Nvidia's doing with Omniverse, into a sort of virtual world where you could be simulating things almost in real time. The class example is the car designer, you know who's not ACFD specialist who wants to draw a car, click a button, see it, get a sense of what the drag

24:48 is and then continue iterating. Now I should say this is not in the short term or even probably the medium term about replacing high fidelity CFD. But I do think it has a very potentially disruptive role in the lower fidelity conceptual design phases that could really be important. And so commercially, this could mean basically a key tool in the market and this is driving a lot of start-ups because there is a sense that there could be acquisitions or mergers from some of the big Isvs or they could simply become big codes. And the CA market is five, $10 billion at least and growing. And it's growing because there is a desire for digital

25:37 certification. There's a desire to go more into the virtual world because it's cheaper and faster than winter testing or physical testing. So now there is still, and I want to make this very clear in my mind, many challenges and many areas that will most likely change dramatically in this space. I do not believe that any of the current methods published today will be the method that we end up using in five years time. I could be wrong and maybe it'll only just be iterations of it, but the field is moving so quickly that I just find it hard to believe that what we have today is the best we can do. Essentially, and I say this

26:23 because if you look at the CFD world, actually a lot of the breakthroughs in terms of the physics or the modelling were sort of done 20 years ago. You know, most of the warmer LES or hybrid random LES or random modelling, it's actually 20 years old. And what's the novel bit is more the computer science, the use of GPU's, etcetera. But the actual breakthrough modelling, mathematical modelling was done 20-30 years ago. There's obviously some new developments, but most people, if you're using DDS, you know, what's that 2004 or something, depending on which version you're using of it. If you're using K Omega, SST RANS, it's 1994.

27:02 So what's that 30 years ago? So it's a bit like going back to 1994. The K Omega comes out and it gives good results. I'm thinking that's it. I think we're in similar time, the AML that yes, there are many startups and companies, but I don't think the actual modelling or the accuracy today is anything like we'll see in the future. And so that means there could be a lot of change, a lot of disruption. And that's why I think it's a really important area to get into because there's no doubt there has still been major issues improving scale. Can these AI methods, you know, work on hundreds of millions of cell meshes? Can they work in terms of the accuracy?

27:40 You know, have they really shown how accurate they are down to, you know, a drag count difference? There's been a lot of studies, but there's been limited real concrete evidence of it away from just sort of marketing. And the big question, of course, is around foundational models. So most of what the startups and most people who talk about it are doing is saying, we'll take the data you have as a company X train a model, you train a model essentially, and then you do imprints on it as opposed to ChatGPT like methods that take huge data from the Internet. And now by paying providers, they train a model for, you know, months and then all you're

28:26 doing is using it as inference. They've done the training and they're recouping the cost from that through, you know, subscription model or pay as you go through the inference. Whereas in CFD we're basically saying you have to do the training with your own data. The future is could there be a way of a company pre training model, then you just do inference. But and I'm, I, I will definitely do a more of a deep dive on this. And hopefully if you look in season 1, I've spoke to, you know, 3 or 4 top ML people that, you know, I think they largely agree that this is a super exciting space. But there's still many questions and uncertainties around

29:01 including physics, not including physics. And I go back to the statement that I just don't think the methods out today are the ones that we will use. But the breakthrough moment could be any time. And that's why I think it's such a super exciting space. So, so the second one is AML. And of course, why do I mention AML? Because it's linked. The first AIML needs GPUs and as a company if you're being asked to train ML models then GPU is the only way to do it and therefore it makes double the sense to procure a GPU type cluster or predominantly GPU cluster because it can be used for your ML and your CFD. Now the third one, which again

29:44 in I believe takes its evidence from these first 2 is cloud computing or SAS type products. And again, I have to be, you know, honest I, I'm clearly working for a cloud computing company, but my belief is independent of working for, for for that company. And I think, and I hopefully I can give some evidence to back that up. The reason I'm saying this is it is undeniably true that the hyperscalers have huge investments in compute and are able to give access to compute that most companies may not be able to get at the size that they want and the frequency and

30:36 the access and the variety. And now there may be times when there's a mixture of the two. And again, it's an evolving thing. If you already have, you know, a lot of on Prem work, then maybe the cloud for you is a sort of a burst option or you know, a backup option, or it could be a primary option. There's, you know, there's different, different models and this will evolve over time. But the fact that major, major companies around the world are moving to the cloud because the way they see it is less of a what's the cost of the compute, what's the cost of the hardware, but more what can people do with it? The point being is that there's a lot of the biggest

31:22 misconception of the cloud is that it's just service to rent. That's maybe the way it was 10 years ago. But now there's so many high level services added by all the cloud providers that actually offering is not only to access the compute, but many services that makes it easier to do things, whether it's machine learning, whether it's, you know, high performance computing or whether it's databases or virtual or end user computing, many, many, many things that makes it easier for your end users to do stuff or for your sort of IT teams to do things that they would have to have spent more time doing. So it's making them more

32:04 efficient. So often when people look at a sort of a cost thing, the cost only really works when you take the value of doing everything. And that's why many, many companies are driving this. And the reality of what I've seen at least is the top down is really pushing this. So the CI OS driving this. And so it's becoming, you know, a reality for many, many companies. But like one and two cloud for many companies is still a journey. And so they're not fully moving to it. But the reason it it's appealing is because if you want to get access to quite a different number of GPUs or you want to test out new GPUs, the reality

32:51 is it's probably the easiest thing. If you want to go and test GPU, it's probably easy just to, you know, get an account on a cloud provider and do some testing. Now, of course there are challenges sometimes with capacity and things like that, but that's more that I guess a transient effect. Given the still the things of COVID and and supply chains, it makes sense to procure it through a, through a cloud system. And you're seeing that, you know, with, with most companies. Now, the reason I'm sort of linking that to the others is there was also a time when many of the Isvs and companies start-ups said, oh, the cloud, you know, it's just fad.

33:28 But now many of them have turned around because they see it as an easier way of delivering their software to you. And this is where the SAS comes in, Software as a service. Now we're used to software as a service because many applications we use through an app, we don't, we just go to a website and we use it. That is software as a service. So, you know, one option used to be that you would download and install something on your computer or you can go on to something like, you know, over leaf I use for like writing papers. It's a web-based platform where you can write Latex documents and it's all stored there in the

34:04 cloud. You don't have to download anything locally. Now that makes it a much easier, nicer experience and it gives access to everything without you having to have it. The analogy would be for the CFD side is the current way of doing it is you would take a piece of software, a binary, let's say, or open source and you get the software from them, but you have to do all the the hardware. So you wouldn't store it on your local computer, on your hardware. But then what if you want a bigger computer? What if you need more memory? Well, you'd have to have to go and buy a bigger computer. What if instead you could go through some website, access

34:42 that same CFD software or some, you know, through website or application, and it sorted out the compute for you. So you just ran the simulation and in the back end it was provisioning a bigger GPU or a bigger CPU, depending on what you. You never needed to worry about that. That is the value of the SAS and it links to the cloud because they're all running on the cloud. So it gives the cloud gives those companies the ability that they don't need to go and buy data centres, They don't need to go and buy a compute. They can just pay a cloud provider. And so you are are getting that and they're using the cloud. And that's essentially how most the two ways that we see and I

35:25 see of using the cloud, you can either use the cloud and it's essentially just a mechanism to get compute in a in a remote fashion. And you don't really notice it. You just SSH one to machine. You don't know whether it's in your cloud account or where it's in some physical data centre that you own. And of course you can use high level services. And the advantage of that one is you keep everything, let's say in your account. So you have a native account, all your data's in there. You're a big enterprise, but you have to know how to use Atbus, You have to know how to do all the cloud. The alternative option is you just say, well, I want to use

36:01 ISV software. A that's just, I'm going to say, let's say, you know, star CCM has a SAS like, you know, a version, you can just go in and then use that or could be any other ISV. I'm just using it as an example. You can go in and pick the platform that you want, the compute that you want go in and they provision it for you. And so you've not had to know any ADDF knowledge to get access to that compute that is going to become more and more the mechanic you look at luminary cloud cloud is in the offering, right. They are a cloud first provider of CFD with the exact thing I just said that it makes it easier for you to run stuff without having to worry about

36:46 the compute. There are many of like SIM scale or flex compute or there's, there's more and more, whether it's startup or an ISV that are offering cloud based back end. And that's why it's can be, in my mind, a very positive thing because it's also, let's say, forget about the commercials. I want you an academic code. If you want someone to use your code or to give access to it, actually it might be great if you build your own sort of SAS platform, you can give people access to your code or collaborate. And it could be private. Remember, it doesn't have to be fully public. You could just do it through a username password and they log on and you can collaborate.

37:25 There are many sort of openings to this that haven't been fully realized. People can create codes and products and serve them to people without being a sort of delivering ACD to someone, you know, that they have to install on the computer, the cloud we use everyday. And it, it is a way of delivering things to you, software to you that CFD has to date not fully embraced. But that is something a bit like GPS and AML, which will only increase. And I personally think it's a good thing and it's something that you can explore yourself as again, not saying that people buying their own desktops or their own data centers is going to stop.

38:10 Of course it's going to continue, but I think the percentage and the data shows us out that the growth of HPC for the cloud is something like 15% year over year, whereas HPC on premise like 8%. And I think that's Hyperion third party analyst. The data shows it is growing. And so yeah, in this episode it's more making you aware of the trends. Whether you agree with them or not doesn't really matter. It's it's they are in my opinion anyway, the trends that that are emerging now. What do these three things mean? GPU technology, AIML cloud. In my mind, it means that companies can move to higher fidelity transient methods. Or if your use case doesn't

39:03 allow you to do, let's say you know, scale resolving type simulation, it could just mean bigger meshes. If it's still steady state it, it could mean running the simulation for longer conversion to even tighter tolerance. It could use more equations for your chemistry model. You know more physics in it. As soon as you have faster, more compute and access to more of it, let's say through a cloud like platform, you are going to see more use of it. And why does this matter? Because ultimately it's about accuracy. The biggest problem for CFD and trust for CFD by very senior peoples, people in companies or by academic research is, is accuracy compared to physical

39:45 testing, right. That's always going to be the put down. The CFD looks good, but how accurate is it? Well, as we all know, accuracy is heavily linked to compute. And that's why I think this really will set up a new wave of move to digital certification. Digital certification, or certification by analysis has long been a goal for aircraft and. Car companies in any industry, but I would argue that compute and the method you can use is a limiting factor. Well, once the compute starts to become relaxed, the methods you can use better methods, which gives you more accuracy and gives you more evidence to do less physical testing or even if

40:30 you do the same physical testing, just more virtual testing. And that will deliver, in my opinion, return on investment because companies can shorten that time to release a product or science can be done at an even greater scale. So, you know, DNS can be done at much better resolutions largely simulations can be done for more realistic problems. You can add more, you know, multi physics modelling inside case because you have more compute, you can do multi physics, so many more things. So I think whether it's more physics or just simply things like small time steps, laws and meshes, all of this will be more accuracy and I really do think that is going to drive greater

41:16 use of CFD and CE. So I think those markets are only going to increase because of 1-2 and three the adoption. This is not in my opinion static. This is a growing and will continue to grow because it's value to a business will be beyond just pretty pictures and actually meaningful improvements in, in, in design and making things safer, hopefully safe and faster better than the usual metrics. So what's my final fifth point to this is this means in my mind mergers and acquisitions, which is sort of interesting. And again, it's a slightly more commercial angle, but I think this is interesting for any people who are looking to find start-ups or you know, people

42:03 interesting inside of space. Whenever you have destructive technologies like this, whenever you have the option of new tech, new hardware, the emergence of AIML, the emergence of SAS providing and the the thing that has and high fidelity transit methods, it means it's perfect time for start-ups and we are seeing the start-ups. Why now in the past two years what why didn't we see you know, alumina cloud of volcano effects, compute, etcetera 5-7 years ago? Because I believe because of the GPU technology, because of the cloud it it's set the right environment. Why are we seeing 10 start-ups now with AIML methods? Well, part of it is the technology, but it's also the

42:43 environment because of the huge Gen. AI boom that's giving reason for these VCs to fund these sort of companies. Well, as soon as you have lots and lots of startups and lots of main players who own majority of the market, that means there's going to be mergers and acquisitions. For sure. The big companies will acquire the smaller companies to keep market, but there is always the potential for a disruptor that a company does not just get bought out but actually rises to really challenge these companies. I don't know which one it is because the to date you know the large companies like Cadence and Synopsis, the big EDA companies and people don't know EDA is

43:31 probably even bigger market than than the CAE. They bought in to those. So you know Cadence have acquired a number of companies, Synopsis have acquired Ansys and Ansys have acquired a number of companies, Siemens obviously Dasseau. So you have these players who see the bigger picture, which is the digital certification, which is the digital twins, which is the, so the way that it can retransform manufacturing and the engineering space. And for them, they see that simulation is a key bit of this. And this is really the, the sort of the big picture that there is A, it's a fantastic time in my opinion, to be in CFD and C it's

44:15 a fantastic time because there is real value that can be created. And the dream of someone sat there and, you know, virtually designing something and testing it and seeing it in like a, you know, through some sort of virtual glasses. It seems to be on the horizon. So I think the next 5-10 years we'll see two things. One, people doing the same they've always done. But I see it's true, it doesn't matter. There are some companies doing the same sort of modelling they did 10 years ago. And fine, if there is a listing technology that's nice, but I'm just going to keep doing it the way I'm going to do it. For some that will happen, but I

44:50 think for others there will be a dramatic change, an acceleration and, and I think this is only a good thing. So that was my five main trends. I would love to know whether you agree. It is totally just an opinion. I'm sure I've missed many things out and I'm happy to go into more details and other things and be corrected if I was wrong and some. But I hope you found this interested and interesting and it's got some of your brain, you know, thinking of does this apply to me and my thing is this just a more niche thing to certain segments. But yeah, I hope you enjoyed it. We've got some great guests coming up over the next couple of months, which I'm super

45:31 excited by. So I don't really like to say this too much, but it apparently it helps. So if you are watching this on YouTube, if you sort of like and subscribe, it helps the YouTube algorithms and also lets you know when there's something. But also if you're on Spotify, Apple, you can I think you can click one of the buttons to to follow the podcast so that when something new comes out, you'll get you'll get notified. All right, thanks very much. Hope you enjoyed it. See you soon.

46:23 The.