prith banerjee, chief technology officer, ansys, inc.
Click to listen on Apple Podcasts or Overcast.
For a menu of other apps, click “Subscribe” above. You won't have to subscribe to listen. Or just listen via the player above.
An unedited transcript is below.
Welcome, welcome, welcome to this episode of Tech. Cars. Machines. This episode was recorded in the era of our unwelcome guest, SARS CoV-2. You might notice from the audio that unlike every other episode, this one was recorded remotely rather than in person. What you’ll also notice is that it’s one of our great episodes in terms of the fundamentals discussed, the groundbreaking progress being made and the enthusiasm of our guest, Prith Banerjee.
Prith is the Chief Technology officer of ANSYS, a simulation company. I first met Prith right before the first of two occasions where he kindly served as a panelist at our private conferences. Prith’s bio is star-studded: Director or HP Labs, a partner at Accenture, CTO of European conglomerates ABB and Schneider Electric. You’ll notice that Prith is particularly excitable and has a particularly natural way of explaining his subject, and this is no surprise because he was also a professor at the University of Illinois at Urbana/Champaign, one of the finest engineering universities.
As you our kind listeners know, TCM explores the the impact of sensing, connectivity and analytics on the world of cars and machines. We’ve talked quite a bit about the type of analytics that gets performed on data that’s already been collected. For decades, there's been another type of analysis that is prospective: rather than collect data, the system use software and the laws of nature to generate data which it then analyses to model the behavior of a physical system under operation conditions. Sometimes these models are “top-down”, in other words simulating the performance of a large, high level system (let’s say an aircraft wing), and sometimes they work at the foundational level, for example modeling the behavior of an electronic circuit. ANSYS is closer to the foundational level.
ANSYS is a public company with the kind of valuation metrics that would put many a venture capital backed unicorn to shame: It’s $24Bn enterprise value means it’s trading at 16x last twelve month’s revenues of $1.5 bn. At that substantial scale, the company is growing 10-18% in the last few years. Net Income margins are an extraordinary 24-32% range. The stock price, no surprise, has tripled in the last 3 years.
Without further ado, let’s get to it!
Ali Tabibian:
Our guest today is Prith Banerjee, Chief Technology Officer of ANSYS, which I gave you a little bit of a description of in our introduction. I first met Prith about four years ago. We had the pleasure of having Prith present at one of our conferences, which are the private conferences that you can access on our website.
And Prith comes from a very, very accomplished background. And there's a lot to go through, we'll have his LinkedIn profile in the show notes, but I have known Prith in the context of three of his jobs. One is a CTO of a very large Swiss industrial conglomerate, ABB. The other one is at the helm CTO of the large, I would say electrical products, conglomerate Schneider Electric and now at ANSYS, which is a simulation software company. And Prith, thank you for joining us. And I have to say when preparing for this podcast episode, which we're very thankful you've contributed your time to, I don't think I've seen you as excited in the media collateral that I was going through, as you are now, in the last three or four years that I've seen you.
Prith Banerjee:
Oh, thank you very much Ali. Its absolute pleasure to speak to you and your team. And you said it's been a pleasure working with you in the past. And as you mentioned, I have joined ANSYS about two years ago as CTO and a fantastic role. It's a very exciting company and of course I'll have chance to talk about it, but thank you for inviting me.
Ali Tabibian:
It's our pleasure. And Prith in the introduction I gave our viewers a little bit of a really financial profile of ANSYS more than anything else, and brilliant extraordinary story on its own. Twenty five billion dollar market cap, one and a half billion of revenues, extraordinarily high margins, stock price has tripled over the last three years and a real testament to both the value that the company has always been adding and really the dramatically accelerating relevance of ANSYS to all the things that are favorites of our Tech. Cars. Machines, podcasts listeners here. Why don't you give us a little bit of a background Prith? What does ANSYS do?
Prith Banerjee:
Thank you. Yes. So at a high level, ANSYS helps companies around the world design and develop the most amazing products from computer chips to rocketships. Our customers use our products to engineer and test products completely in the Digital Domain, without the need for costly and eloquent physical prototypes and experimentation. So we call this approach simulation based product development or software based product prototyping. So in the past or 20, 30, 40 years ago people, when they build their product, they used to do hardware prototyping and these things required a lot of time and cost and so on. But over the years, ANSYS has developed this amazing, very detailed physics simulation capabilities, where the world around us is governed by the laws of physics. And if you can accurately model the physics, you really don't need to do a hardware prototyping.
So in the past, you have to actually take a model of a plane, a model of a wing, put it in their wind tunnel and pass wind at 300 miles per hour to see if that wing will lift or not. But today with ANSYS simulation tools and fluid dynamic tools, if we say that we will lift, it will lift. If we say, there's a dragon, there will be a dragon. If you say, there's a stall, there will be a stall. So we have completely eliminated the need for physical prototyping. And that is the value we provide. As a result, we enable our customers to drive top-line revenue by designing much better products, faster and with higher quality and also bottom-line cost savings by reducing the cost of R&D. So our customers are the most innovative companies in the world. In industries, as diverse, as high tech and semiconductor, aerospace and defense, automotive, industrial, energy.
So these companies are to innovate and solve incredibly complex challenges in areas like 5G Academy, electrification and industrial internet of things. That in a nutshell is what ANSYS does. And ANSYS today, as you said, is a one point five billion dollar company, market cap about twenty five billion. We have 4,200 amazing employees, always working to solve customer problems. And interesting financial thing that since you talked of a financial is, something that our CEO talks about. You look at all these public software companies in the world, there's about 2000. Then of them, if you take of those, how many are more than a billion? There's about 200. Of them, how many are more than 30% growth margin? There's only 12. Of them how many are growing at 10% or more every year? It's only four companies.
That's an amazing statistic. Public software company, more than a billion in revenue, more than 30% gross margin and growing at 10% a year, every year, there's four companies at the top, there's Microsoft, there's Adobe, there's ANSYS. It is an amazing, amazing story. So, we are very, very excited to be part of that group.
Ali Tabibian:
That's amazing statistics. Thanks you for adding to that. When I was a young puppy in the former century, I remember an engineering school, finite element analysis, at least the computer modeling of finite element analysis was a big deal. Seems that's where ANSYS started. Is that right?
Prith Banerjee:
That is correct. So ANSYS started exactly 50 years ago. In fact, last Friday, when we celebrated our 50th anniversary. It was started by a founder, John Swanson, who essentially created this technology called ANSYS mechanical, which is finite element analysis. And the core, what is doing is you look at the structures around the world and those structures are governed by pure physics. There are second order, partial differential equations called the Euler equations. And you have to solve them numerically, and the numerical method is called finite element analysis. And that's where John Swanson got started. But over the years we have acquired more capabilities. So we can now solve Navier-Stokes equation, in fluid mechanics, which is again, a second order set of equations. Then we acquired the capability to go for electromagnetic problems, which is Maxwell's equations. Then we acquired the capability to have semiconductor chip equations.
So Kirchhoff's equations, voltage equations and so on. Then optical solutions and most recently we just acquired technology called numerical which allows us do photonic and IC simulations. So over the years, our core value proposition has always been take the physics around us, which is modeled by pure partial second order, a partial differential equation, and solve them numerically using finite element analysis, finite volume methods, boundary element methods, finite domain, finite different time to think methods, different simulation methods, but they are all the value is the same. Try to solve these problems, most accurate manner, as fast as possible and as robustly as possible. So you don't need hardware prototyping.
Ali Tabibian:
Outstanding. And Prith, for our listeners, and please correct me if I'm wrong. It's important to note that one of the important fluids in any system is essentially heat as it travels through the air inside that system. So that the finite element analysis combined with fluid dynamics, if I'm not mistaken, gives you a real sense for the thermal behavior of a particular system as it's being modeled. Is that correct?
Prith Banerjee:
That is exactly correct. So actually, when we started ANSYS 50 years ago, we were solving single physics equations, only the structural equations or only the fluids equations. Navier–Stokes or or Maxwell's. Increasingly, the problems that our customers want us to solve is the multiphysics interactions, the fluid structure interactions, as you talked about. For the thermal, we're in a method called the conjugant heat transfer. You have a hot material inside and that hot material may be generated because of the computer chip. So imagine a computer chip sitting on a printed circuit board as that computer chip goes at two gigahertz, it generates maybe 1000 watts of power and that generates the heat. That heat you can actually model and that because of that heat, it sits on a printed circuit board that printed circuit board expands, so we can model the thermal expansion of the printed circuit board through ANSYS.
And then in order to cool it, you need to put cool air or cold water, and that's the fluid. So how do you actually take heat away from the hot chip, through thermal conduction via convection or conduction and so on. So we can model all of it to the fluid structure electronic interactions. And those are the multiphysics interactions that our customers are always looking for. And we provide all of these capabilities inside of high performance workstation or low costs clusters on-prem on a private cloud or the public cloud. So all our software now runs on Microsoft Azure cloud.
Ali Tabibian:
Very impressive. Thank you for helping bring that all alive for our audience.
Prith, I would like to go a little bit into your personal background and how you wound up at ANSYS, but maybe an interesting way of doing that would actually be to first talk about some of the roles you had in the IoT space at ABB and Schneider, APM, predictive analytics, and how that predictive analytics would be depends on physical simulation. And maybe that's we go through that path first, if you will to, then bring you to a why you're at ANSYS and what's your targeting. Is that okay?
Prith Banerjee:
Actually, that is exactly how I landed up at ANSYS. So I have been working on the IoT area for more than 10 years and as you know Ali, I started my career way back in academia. I used to be a professor at University of Illinois at Urbana-Champaign. I was at Northwestern and so on but then through those days, I landed up in my role as director of HP Labs. And at HP Labs more than 10 years ago is when I started my first work on IoT. We had this incredibly exciting project called CNS, central nervous system for the R&D and division was, this was the year was 2008 where we said the world around us will have all the IoT components. And what if we could have sense everything around the world. The central nervous system for R&D. And you collect all the data and you analyze all the data, what could you do? And what could we predict?
So we were doing IoT work way back when, a couple of years ago. And on that time, as you know, GE produced this very exciting work paper on the industrial internet of things in 2012, and that literally changed transform the whole industry. They said, "Hey, here is a fourteen trillion dollar market revenue opportunity," and everybody, well, became excited. And so ABB recruited me as their CTO in 2012 and essentially said, "Prith, help us build towards this IoT journey." So I help ABB building their IoT platform called liability, and that allowed ABB to take all their large assets, they had transformers, they're robots, they switch gears and so on all those large asset intensive industries. How do you connect them through sensors and so on and connect it to their IoT platform? And the reason they did that was to enable them to do remote services when a robot in the field fails.
If it is not IoT connected, the customer would have to call up ABB and say, "Hey, my robot failed and send a replacement part." Well, when that it is IoT connected, ABB found out as soon as the robot failed and they could ship a service technician. Not only that they could remotely diagnose where that product was failing. And so when the service technician arrived, he could take the replacement part with him. Oftentimes, remotely, if there are enough redundancies, they could fix that part rebuild robot, just remotely through a service engineer sitting in Bangalore. Now, so the IoT was essentially meant for solving the simple problem of remote services, is a low hanging fruit. For the next step was, well, can we use it to predict when that robot will fail.
And so that generated this whole area of predictive analytics which I will talk about. And that is a journey and I will... So I did that for a couple of years, and then I moved to Schneider Electric and essentially Schneider Electric was in the same business as ABB. And I helped them build their EcoStruxure IoT platform to connect up all their UPS devices and their breakers and panels and so on to this EcoStruxure platform. Same problem they are trying to solve remote services, predictive analytics and so on. The problem that I found in working with these companies is that, the quality of the predictive analytics is based on the amount of data that you see. If you... An AI is only as good as the data you train it with. So they say, "Show a picture of a cat."
You show 10 pictures of a cat, 100 pictures of a cat. Then the AI algorithm knows what a cat looks like. But if you show it a dog, it gets confused because it has not seen that before. So data analytics based on training data is only as accurate as the data you train it with. If you have not seen that before, you can never predict that. An example, when space shuttle challenger exploded, no amount of predictive analytics could have predicted that that space shuttle would expand because the number of times it exploded was only one. There was no historical data of our space shuttle exploding. On the other hand, if you do physics based simulation, you can actually model the fact that space shuttle is coming down to earth at this speed of 5 Mach, 10 Mach, 20 Mach, tremendous fluid interactions happening with the tiles and that because of it there's heat that is being generated, the tiles expand and they explode, you can actually model the space shuttle challenger exploding if you had a simulation based model.
This was the aha moment I actually faced I say, "Hey, these database analytics things that we are doing at ABB and Schneider is not working, we need to have a physics-based model." I used to be a customer at ANSYS when I was at ABB and Schneider. I started doing a little bit of a study myself say, "Hey, it would be good to have a simulation-based method to digital twins." That is what attracted me to ANSYS. So literally I knew you asked me that question, why ANSYS? It is because I was personally very excited about IoT. I wanted to solve the problem and the way to solve the problem is through simulation-based methods that ANSYS not very known for, with their digital twin product. But that is not the endgame. I will tell you where the future is. It's not pure simulation is not pure data analytics. It is in the combination. And I will share with your readers and participants what the future is later on.
Ali Tabibian:
Excellent. Thank you very much. And that's such a great point Prith, in the sense that sometimes we get ahead of ourselves in terms of what a computer can infer from data without a formal understanding of what drives the inputs to the outputs. Sometimes that formal understanding can allow you to land a spaceship on earth and land a spaceship on Mars, because you understand how the laws of gravity work. And there's no amount of modeling of landing on earth that would allow you to understand how to land a spaceship on Mars. And that's just really something I think for our listeners to grasp, because obviously in this tech series, we're big proponents of analytics, ML, all the fun and cool stuff, but in the end you do need to have a formal understanding of at least some parts of the problem, to either make progress or make that progress a lot more efficient from a computing perspective. The-
Prith Banerjee:
Absolutely. And in fact, let me just, for your listeners, because you mentioned this thing about analytics and AI, I just wanted to share a very interesting point with you. So you mentioned gravity. So 500 years ago, when Isaac Newton observed this apple falling and he said, "Oh, Hey, there's this thing called gravity." And he essentially drive this whole four equals times acceleration and so on. How did he do it? I mean, he's an amazing guy, a brilliant scientist, a genius. And in a way we call it, almost like he had super power instincts right along. I mean, he just saw that apple fall and he said, "Oh, my God, there's this fundamental physics of gravity and gravitational pull that is defining why an apple goes down at a speed.
You and I maybe we watch an apple go down, a ball go down, a airplane go down and we never made that connection how is the physics work? But what you can do with analytics is essentially, you take this observation that, here is an apple sitting 10 feet above ground and you let go of the apple and it drops down and in exactly two seconds, it drops on and at a certain velocity of, I don't know, 10 meters per second it is hitting the earth ground. Then you take the same apple and you take it at 2,000 feet and it dropped it and it hits the ground at 4,000 meters per second. So essentially what is going on is Isaac Newton was observing the laws of physics. Here's its input, here's its outward. Here's its input, here's output. Based on it, he created the physics of gravitational pull that the physics of neutrons law, force equals mass times acceleration, which is a D two squared over the X squared over DT. And now you have this.
So Isaac Newton created those equations and partial different equation. And here ANSYS can go in and actually model that ball going down at whatever speed and whether it will crash or not. So here was an observation that Isaiah Newton did. And he found this hypothesis this equation for the force mass acceleration gravity and so on. He introduced the concept like equations, and then, well, you can solve it analytically or you can solve it through simulation with ANSYS. This is what ANSYS simulation does. It can take an average human like me and you and Tom, into a superpower. We can all become the Isaac Newton's of the world. This is the part of simulation.
Ali Tabibian:
Great. Thank you. And Prith thank you for assuming that I would be in the same category as you. I'll take that and hurry on before you change your mind. So Prith, at ANSYS what are the business outcomes you're targeting? And if you prefer, you can tell us in the general corporate context or specifically with respect to autonomy 5G and electrification, which are particularly germane to our listeners, whichever approach you like.
Prith Banerjee:
So at the fundamental level, what ANSYS is trying to do, as I said, is to help our customers build amazing products using simulation and value of simulation is if you think, we promote, we allow much more rapid innovation. Why is it rapid? Well if you had to do a hardware prototype based stuff. I want to build a car or an engine. I have to melt the actual hardware prototype of that engine. It would take six months to build the first prototype and it will test it or it doesn't work. "Oh, my God, I have to do it another time, another six month." So instead of, if you could do that, innovative product design and simulation on a computer, you can do much more rapid innovation. You can do 30 different designs, 100 different designs and all within a matter of days, so its rapid innovation.
That allows you to lower cycle times that allows you to reduce risk. So once you test their product and you have, in a physical product, we can only test it maybe three different ways. We have a... On a simulation, you can test it many, many ways. So you can reduce that risk. You essentially increase quality of your products and you manage the complexity. So at a high level, these are the impacts of simulations. From an end customer's business perspective, we enable our customers to essentially have top-line revenue growth. So our customer such as ABB that I used to work for, we allow ABB to make more amazing robots, faster, more rapidly, et cetera. So it gets to that top-line revenue growth, as well as cost savings. Improved our efficiency, so the cost to build a robot with hardware prototypes and so on is maybe a million dollars.
The cost to build a robot is in simulation is only two hundred million dollars. That allows them to build fewer physical prototypes and lower warranty costs. So essentially, the business outcome we are driving is top-line revenue growth and bottom-line cost savings. So therefore the CEOs of their customers, they absolutely love us because it's helping go top-line and bottom-line. Now with respect to autonomy and so on. So in the past, we were focused on building these individual solvers, like mechanical structural solver or fluid solver or liquidity solver and so on. Increasingly, we are helping our customers with end solutions. Solutions like autonomy, as BMW is trying to build an autonomous car. What is the difference between a normal car and autonomous car? Well, in normal car URI are the driver. In an autonomous car, that car drives by itself. How? Because it senses things around it.
So what kind of sensors does the car have? Well, it has a LIDAR sensor, it has a camera sensor, it is a thermal sensor, it has a radar sensor. We have the ability to simulate all of those sensors in the most accurate manner. And then we combine that with scenario planning, and then we do the full, what is called closed-loop simulation to simulate the car for those eight billion miles that you would like to test it. I mean, it would be hard to certify a car, unless... I mean, people say that you need to have a car driven on the road for about eight or 10 billion miles. When it turns out all these autonomous car companies that are building this car, they have given about 25 million miles so far. It will take them 1000 years to actually do road testing of all those cars, hence simulation.
So through simulation. You can take that entire autonomous car with all the sensors, with all the scenarios and simulate all the possible edge cases to say, "Yes, this car is going to run." For example, if you have an autonomous car, we both live in San Francisco. And you say, "I'm going to test this out on driving on Golden Gate Bridge at 9:00 AM in the morning when it is sunny." And at that point, there's one child that is crossing the street, my car actually senses it, it stops. Check box. Okay. The car is safe. Now I would like to say, Hey, what if it, instead of 9:00 AM, it is 9:00 PM in the evening and it stopped being sunny, it is raining. And oh, by the way, it is snowing also. And it's not one child, there's 20 children and a pedestrian and a cat and a dog. Well, I would have to wait 10 billion years to get to those combination of scenarios.
In a simulation environment, I could just create that scenario, Golden Gate Bridge, 9:00 PM, snowing, frozen, 32 pedestrians, going at 90 miles per hour. Does my car stop? Yes. Done. Check box. So that's the part of simulation and that's what we are doing with autonomy. We are doing similar things with electrification. We are doing similar things for 5G and IoT. So that's what we are doing in terms of business outcomes for our customers.
Ali Tabibian:
That's very impressive, so Prith, does this mean that I could give you an environmental model, meaning the Golden Gate Bridge with this particular features and then I could tell you what type of LIDAR and what type of sensor packages my car has, and you'd be able to run that simulation dynamically and interchangeably. I could take a different vendor's LIDAR and put it in and you'd be able to simulate that just as easily?
Prith Banerjee:
That is exactly the way. So we actually are a platform solution. We provide this solution for our OEMs. OEMs such as BMW. But also for the people who are the Tier 1 one suppliers to the OEMs, like Aptiv or the people who are making these LIDARS themselves or the cameras, like AI. So essentially, our simulation works with the component suppliers. We can simulate those components, they supply with the actual LIDAR things. Like the Aptivs and so on and with the whole cars. And we provide that entire multistage simulation, and exactly we can take any environment and we work with the company, with our partner company called Edge Case Research. It's a startup company based in Pittsburgh and from Carnegie Mellon University. So basically when a Waymo is going around and driving and taking pictures with the cameras, that's a real video that they're taking. We can take a 20 second part of a actual camera in any location that you have. Golden Gate Bridge or Eiffel Tower in Paris or whatever.
And in that there maybe 20, 30 chips. We have an automated way to replace those pedestrians with twice as many pedestrians or 10 times as many pedestrians, more cars, less cars, skidding conditions and so on, create the scenario. This is what it's called scenario generation, and then do a full vehicle dynamic simulation with the different LIDARS and so on. So we can actually tell an OEM company, "Hey, this LIDAR will stop the car at 50 miles per hour, but it will not stop it at going at 90 miles per hour. So they can make those design decisions. And this is the power of simulation. We can allow our customers to do much more rapid innovation and increased quality and lower risks.
Ali Tabibian:
It's interesting Prith. There seems to be a nuance but still an important nuance between a lot of the simulations ANSYS has been doing, and for example, the autonomy one. When you simulate the performance of a wing, first of all, you've built up the performance of the wing from first principles, from those equations that you've talked about. And then the scenarios under which you test it, while there are always going to be edge cases that might be missed, but it seems like the scenarios under which that model needs to be tested are a lot more, are essentially finite compared to the scenarios under which an automobile gets tested. To some extent you have a model but maybe it's not the model isn't built up from first principles in the auto scenario. Some of it is maybe just in turn, depends on machine learning input. And then you're operating in an environment where it's harder to determine what the edge cases are. Am I drawing a distinction without a difference or is that something that you need.
Prith Banerjee:
Fantastic distinction and actually that brings me to the topic of IoT that I wanted to end my conversation with. So, what we enable our customers. So in the past, we used to allow our customers to do what is called full 3D simulation. Then solving their Navier–Stokes equation in the most detailed level. Doing the most complex meshing, having 40 million mesh binds over those wing and those airflow going over the wings and so on. So that is what we used to do which are called these 3D models, and using finite element, finite volume methods. But then there's this new technology that we have invented called creating reduced order models. And this is where it's almost like coming to AI. You do a bunch of detailed simulations Ali and based on those during the simulations, you can take that detailed model and build a simpler system level model which is called reduced order model.
So instead of having a full 3D model, a model need actually three dimensions and time four dimensions. You can extract extrapolated into a zero dimensional model or a 1D model. A Static ROM is what is called a zero dimensional model. And in 1D-ROM, redo stator is what is called a Dynamic ROM. Anyway, those are technical details. But essentially you take a really complex component like a robot, using our Twin Builder product, We can build a reduced order model of that robot, right to say, Hey, and essentially, the redo stator a model says, "If my outside temperature is 20 degrees, this thing robot will crack after five hours. If the outside temperature is 100 degrees, the robot will crack after 10 minutes. And the after outside temperature is 900 degrees, the robot will crack in three seconds."
Those three things, you did through very, very detailed 3D simulation using ANSYS mechanical finite element analysis and so on. Then what you do is you stick that robot in the outside world, and with an IoT platform, you collect the temperature. What is that outside temperature working the robot? And if the robot is outside temperatures is 10 degrees, I will predict that the robot will crack after 20 hours. And if it is outside temperature is 200 degrees, it'll crack up for 10 minutes. And if the outside temperature is 500 degrees, it will crack after five seconds. I had done that characterization before that's that reduced order model, then I tie that in with the IoT and voila I have a predictive analytics often asset with a very, very high degree of accuracy. These are the capabilities that ANSYS is providing with our Twin Builder, digital twin product.
Ali Tabibian:
And that seems to be a spectacular change in the typical industrial design process. Because what you're describing is not the old serial process simulation to prototype to production but in a sense, because you've got that sensing out there, your models are being informed and updated in real time by the sensing that's being done in the field on the finished product --
Prith Banerjee:
That is exactly true! Let me give you another example of how this is. So I'm literally imagining myself sitting on top of Golden Gate Bridge. And suppose there is, I'm trying to predict that, that bridge is going to have a crack, and because of the crack is going to fail. So what happens is, if we know there is a crack of 10 inches on a particular part of the bridge with ANSYS mechanical and with the amount of stress going on, on Golden Gate Bridge today with the 10,000 cars that are going, I can predict through simulation how much that crack will grow by next week. I will predict through simulation that crack will go from 10 inches to 12 inches. And then I have an IoT connection to the Golden Gate Bridge. I actually measured the crack next Tuesday to say, how much was the crack.
If the crack is at 14 inches, I say, "Oh, my God, I cracked more than the simulation predicted. So I need to accelerate my simulation." So my simulation model of the bridge gets more accurate with the IoT data. If the IoT said that crack was only eight inches, that means, Hey, I was wrong. I need to decelerate the simulation instead of doing the crack propagation at 12 inches, I need to go to eight inches to the next week's prediction will be more accurate. So the combination of IoT fitting into the simulation, so the power of the digital twin is, as you know, the digital twin is you have a physical asset, you have a model of the asset. And with the IoT, you have two-way information flow going back and forth. With ANSYS Twin Builder, we have the ability to get a more accurate prediction through the simulation of the digital twin, that you cannot do with IoT data analytics.
Ali Tabibian:
So this is a nice lead on to the recently announced Object Management Group and a Prith, I'm just going to turn it over to you but right before I do that, I was interested to see that Lend Lease where our good friend, Bill Ruh is involved and also comes from an IoT background was part of that Object Management Group. Why don't I just turn it over to you to tell me what that group is doing and why it's important to the future of the digital twin?
Prith Banerjee:
Absolutely. So the Object Management Group, there's a gentleman Richard Soley. He's a very good friend of mine for the last 10 years. And they have done various standards in the industry and about five years ago, they ran this organization called the Industrial Internet Consortium, the IIC, where Bill Ruh, when he was at that time head of GE digital, he was a big proponent of joining IIC and I was at ABB at the time. And so Bill and I are very good friends. We both IIC and the IIC group, were about 300 companies. And ultimately Schneider, also ABB. So this was a consortium of companies that all valued that interoperability of an IoT based ecosystem and everybody was doing predictive analytics and so on and so forth. So there was a hugely successful consortium that Richard Solely run, Bill Ruh was part of, Prith Banerjee was part of, and it was at ABB and so on. Fast forward to 2020, Bill Ruh leaves, GE goes to LendLease and now he's doing building management infrastructure and so on.
And he's facing, he say, Hey, I'm trying to build this digital twins, but there's no standards out there to build this thing and I have got digital twins in building management, digital twins in aerospace defense, digital twins in automotive and so on. Hey, this whole world is becoming this no standards, no nothing. We should have a consortium. So basically, Bill Rue goes to my friend Richard Soley, "Hey, OMG, you did this IIC so successfully five years ago, maybe we should do a digital twin consortium. And so Richard Soley calls me, I mean, it's the same characters, he says, Prith, "You are at ANSYS. Do you see value?" I said, "Absolutely." Because I'm trying to push at ANSYS this whole simulation-based digital twins and so on.
And so then I say, "Hey, I've got this good friend at Dell." I mean, he was also part of the IIC. So I call up and say, so bunch of people all got together and say, "You know what, we need to do this Digital Twin Consortium." That is how we'd got formed. A few people, I would claim my myself to be part of the... Actually we are one of the four founding members. So Microsoft, Dell, ANSYS, and LendLease are the founding members of the newly formed Digital Twin Consortium, which is coordinated by the Object Management Group. Now, after they announced, we just announced it like less than a month ago. We now have, I think 70 members and a lot of companies are joining, and this whole ecosystem is just going to really grow and expand and explode.
We've got four broad vertical groups you are going after building management is one which being obviously led by LendLease and others. Then we are looking at aerospace and defense and oil and gas and energy, and also aerospace and defense. So for very, very exciting verticals for which digital twins will be created. And this whole consortium will allow people to do lot more collaborative research and digital twins, both physics-based digital twins, combined with analytics-based digital twins.
Ali Tabibian:
Excellent. And going back to the automotive example where you can swap in different LIDARS from different vendors and, simulate the outcomes for the performance of the system which is in that case, the automobile. Presumably, in the part of the job of the OMG is to allow a plug and play ability and interchangeability of the models of their own offerings that the consortium members are providing to the simulation environment. How do you go about testing and verification and compatibility to make sure that people are submitting models that are valid and work appropriately with the tools that the group members typically use?
Prith Banerjee:
So what the... Again, we have just started the DTC, so I'm on the steering committee and we are having all meetings here. So one of the things that we have agreed to do as a group is to first of all, publish these are frameworks and standards and so on, how are the digital twins going to be created? Some of these digital twins will be based on analytics alone, some will be based on simulation, some will be based on combination and so on. And then the other thing is there's the need to create what is called a digital twin definition language. So when you look at a large system like an airplane and you can say, "Oh, I want to have a digital twin of a Boeing airplane. Well, an airplane is a large system: fuselage, tail and two wings. So you can say, okay, I can describe an airplane having a fuselage, tail and two wings, click, click. Digital twin of a wing.
So then how would describe the wing? A wing has this and there's two engines on it. Click, click. Digital twin of the engine or the engine has XYZ and here's 52 blades. Oh, I need that digital twin of a blade, click, click. So essentially, I start with trying to build a digital twin of a system which is airplane and it recursively calls functions which is the digital twin of the wing, the digital twin the engine, of the blade and within that, the skin of the blade, I mean, you can go into as much level as possible too. And that is where ANSYS and CAD models and simulations coming because we are in the heart of it. When products get built, they are built using system level tools. They're top level requirements using model-based software engineering, MBSE and requirements and so on. Then you enter it into a CAD model then that cat goes into machine, that machine goes into simulation.
So we are playing in this ecosystem. So essentially there is a language called digital twin definition language that we are partnering with Microsoft on to express digital twins of complex systems. The system can be an aerospace system, the system can be a building management, can be an oil and gas and BJ well oil fields or manufactory. Then with... So once you specified with digital twin definition language, then you go into the actual models. And all the models that are being created, you mentioned this site, how do you encourage more collaboration? Well, through open source. Just like in the open source world, the whole operating system Linux got created. And lots of people contributed to the open source Linux, and it was maintained by commercial companies like Red Hat and Sushi and so on or the map produce, created in an open source formed by Google and then it was maintained by Cloudera and Mapper and so on.
We feel that all these technologies, all these different people will contribute to an open source world, the Lean Track, and then ultimately some commercial offerings they'll say, Okay. "You can use to download that open source software or you can have it officially maintained and serviced by companies like ANSYS." And so, this is how we plan to make money out of it, but the value will be through the collaboration.
Ali Tabibian:
So that's an extraordinarily bold project and sounds like it's the right time for it from a technology and technical point of view. Now, if you go to the guy who's managing an oil platform and say, "I'm going to let you be able to double-click your way from the top of the system to the bottom of the system on this computer." Sometimes they'll the reaction is, "Could you just get out of the way for now, I want to close this valve before we go up in flames." How do you get the buy-in from the business operations to allow such a bold and clearly useful modification or enhancement to their approach to actually take hold you're an expert on driving change in large organizations. What are your views on that?
Prith Banerjee:
First, Well. I think there is no growing realization by different people in various industries that this digital twin tsunami that is coming is exactly real. So let me tell you something, since you talked about so financials and so on. The digital twin idea is actually about seven, eight years old. I mean, it's more than 10 years old. The market-
Ali Tabibian:
Right.
Prith Banerjee:
Exactly. The market for digital twins, if you look at 2020 is about three, 4 billion today but the market opportunity is going to be 26 billion by 2025. So this is like going to explode. It has less than a billion, just a few years ago, two, three, 4 million today, two 26 billion by the year 2025. Why is that? I mean, again, I can send your listeners, point them to a really good research done by Grand View Research. They can go to Grand View Research and look at statistics but in manufacturing alone, the percentage of that market, that twenty six billion dollars in manufacturing alone is about 20%. So five million dollars out of that is in manufacturing. There's about 30% in residential commercial, which is why LendLease interest. So the point is that, this VP of operations in this manufacturing company, there was a old way of doing things, which is paper and process and so on.
Well, he looked at all different injection molding machines and they laid them so on or this machine is not up and so on. That is the old world. And the new world is industry . All connected, smart manufacturing and so on. We had things are, I mean, they're not making 20 million copies of the same product. I mean, it's being, people want the efficiency of large production with the customization of one. That's the future of manufacturing. A VP of operations of such a manufacturing plant can get to that using 1,900 stick knowledge. So they have to embrace digital transformation, therefore digital twin, therefore all these technologies. So it is actually a fairly easy thing. I was just talking to... I mean, if you look at our panel, I mean, I just want to point our listeners to this simulation world conference that we held June 10th and 11th and I urge them to look at a keynote that I presented on long-term technology strategy, where I highlights all these things.
And also a panel lay moderated where the CTO of Stanley Black & Decker to the Bangalore talked about the fact that in his large factories, he has 150 factories. The cost that he can save is one hundred million dollar off saving in the injection molding machines alone. So these numbers are shared with the CEOs, the VP of operations has no choice, but to listen.
Ali Tabibian:
It's interesting. We both, as you said, live in San Francisco, I was on Van Ness street, which is a six-lane boulevard going down the center of the city. One of the construction workers that I saw, and he wasn't a manager. He was walking the construction site with tools and doing things. He had an iPad on his tool belt. He actually like, and every now and then he would turn around and refer to, I guess, to see where the underground infrastructure is, et cetera. And that goes to show you that the digital twin compared with the visualization added to the visualization that's available via that iPad or even augmented reality can really make a difference in terms of how immediately and in real time the benefit of this digital twin is delivered to the person that's onsite and is really critical in terms of the adoption of these solutions.
Prith Banerjee:
Then just to make a point. If this worker that you talked about on Van Ness street. Just think of the plumber that comes to your home. Think of the analog plumber. He comes in and he strives to fix a pipe behind the wall. And this leaking happening, the wall is leaking, he has no clue. Imagine this plumber going with an iPad goes into your home, and he's connected to the plumbing drawings of the building when it was done like 10 years ago. And he points the iPad at that wall and through GPS or Bluetooth or whatever he has exact location. He knows exactly where that pipe is and tied with similar he can say, Hey, there's a leak happening to ANSYS fluid simulation. He can simulate what is happening. He can simulate. If I were to break this at this point, puncture the pipe, what is going to happen? All of those things can be visualized through augmented reality.
Think about the digital plumber. If analog plumber with these VR glasses or forget the iPad, wearing this augmented reality glass looks at a wall, looks at the pipes behind the wall. And as he is drilling to the wall to fix the pipe, sees real time simulation of what if I cut this, how much water would flow to ANSYS simulation? This is where we are headed.
Ali Tabibian:
That's pretty impressive. So let's take that thread in the future and maybe transition our discussion a little bit further into the future. Do you think that some of these capabilities will actually, finally I would say enable outcome-based services? We talked about asset performance monitoring, predictive analytics. The goal was also to be able to just provide jet engines as a service. Essentially every piece of equipment as a service, do you think we're a lot closer now as a result of these tools than we were several years ago?
Prith Banerjee:
Absolutely. In fact, in 2015, I presented the white paper to the World Economic Forum in Davos, where we can talk to about this whole journey towards outcome with, and many other companies have also talked about it. So essentially everything as a service. Instead of giving a jet engine, doing the jet engine as a service, I transferred to the service and the value of this as a following. Like if you own a car, and as the owner of the car, you are not motivated to keep your car. Really working all the time and it fails, and now I go to the repair shop and fix it. But if the owner is motivated to keeping the uptime of the car all the time. Suppose you are Uber and you have a fleet of cars, you are highly motivated. You can have all your cars up all the time. So as you are going towards car as a service, as opposed to owning the car for twenty thousand dollars.
If I, as a customer, just rent out the outcome of transportation and a service, I pay them twenty dollars to go somewhere else. Uber is highly motivated to keep all their cars up all the time. And therefore they will need predictive analytics to make sure those things just before it fails, they will be able to repair those things and so on. So this is absolutely coming. In every industry, people are talking about energy a service, building as a service. And so literally what Airbnb and Uber have done today, it is the future outcomes and so on. And I am absolutely sure that we are very close to that end state.
And so to give an example, people talk about in the energy industry. Oh, I'm going to do IoT and so on and so forth. And I will sell this fantastic IoT technology to a city government of city of San Jose or city of San Francisco. Well, the city governments have no money. But if they hadn't the money, then they could have actually done the vision of all their lives and so on and so forth and saved a whole ton of money. The trouble is, nobody knows who will pay the money. So essentially there are these very sophisticated people who are thinking about, so for example, and a site company, I think it was TCS or Wipro. One of the companies, Indian companies, they went to the city of Pune, propose it. I will IOTIs all your lights and everything in your city of Pune, and the energy savings that you have, you don't pay me a thing but over the next 20 years, you pay me 5% of the savings.
So essentially the city of Pune said, "Hey, I don't lose anything. And if this really works, they will save energy consumption by 30%. And I'm absolutely willing to give 5% of the 30% saving to," I forget the Indian company. This is the future, outcome-based businesses. So now these future business leaders, like the future business CEOs have to take that leap, have to trust this technology and this will totally benefit society and everybody else.
Ali Tabibian:
And it's interesting. For that to work, we've kind of come back full circle, the machine learning isn't enough. You need a physical model to augment it because nobody's going to repair. Nobody's going to put in a thousand dollars of new brakes, if it's only 50/50 that it was the right time. And you certainly don't want to wait too long if you're Uber. So we've come back to that integration of machine learning and physical modeling that we started-
Prith Banerjee:
That is exactly the reason we change. We rotate our cars, our tires, every 5,000 miles or we go and do an oil change after 20,000 miles is because of what is called preventive maintenance. The car companies take statistics of all the million cars that are driving on the road, and they say on the average, every 5,000 miles, you should change rotate your car, tire or change your oil. But if you are driving as crazy as Prith drives, Prith tiers need to be rotated after 2000 miles, not 10,000 or not 5,000 or maybe as Ali drives, you are a very courteous driver I've heard and your tires need to rotate after 10,000 miles.
So through an IoT connection, you can actually model predict when your tires need rotation. And if you do it purely based on statistics, you will not be accurate. If you tie that in with the simulation, that will be an accurate model. And to your point of the 50, 50. when I was at ABB, we are having all these robots and it's two hundred million dollar assets. And we say, predictive analytics. We predict it will fail next Thursday. If the accuracy of that is 60%, which is actually a number that I have. And I replaced the part. I made a 40% error, I just made a four hundred thousand dollar mistake replacing the asset which I need not have replaced. Through simulation-based, we can raise it up to 95% on the combination, not simulation alone but the combination of simulation and analytics will actually get you to the 95%. So when we say this million dollar asset will fail, it will fail. You better replace it.
Ali Tabibian:
Prith, I mean, clearly working in a public company, you have some form of a P&L and operating incoming objectives. Is there something amongst the types of things we've talked about where you also maybe even informally measure yourself and your success at ANSYS, are there things out there you want to see some of these creations where you would say, "Wow, you know what, we made it financially successful but that's physical manifestation of my work at ANSYS is what makes me know I've been successful."
Prith Banerjee:
So, I mean, from a visionary perspective, I have been fortunate to work for a CEO, Ajei Gopal, who is very, very supportive of all the dreams I have. So one of the dreams I had for Francis was to move into a whole new vertical. Vertical of healthcare. So today we sell to our simulation products in aerospace defense, in automotive, in industrial things, in high tech and so on but not in healthcare. And the question is why? In most of the other vertical, I mean, there's... If you look at the R&D expenditures, across these different industries, all industries is one trillion dollars.
The simulation spend in a typical company is about 2% so off a one trillion, the simulation spend of R&D is 20 billion, which by the way, is the total available market for ANSYS. Right Tom. This is what we say. And we are, we get 1.5 billion of that 20 billion market. Now off this one trillion market, two hundred and forty billion dollars is in healthcare, but in health, unlike other industries like aerospace, that Boeing or Airbus before build a plane, they use simulation to test the plane. When Ford or GM makes a car, they use simulation to test the car. When... You name it. ABB makes a robot. They use simulation to test the rat except in healthcare. When drug companies invent a new drug and we are facing the COVID crisis as we speak. They do a little experimentation in the lab with chemistry things.
I take hydrochloric acid and ammonia and mix it together and I put it with this little thing in a test tube. Once that works, we tested on mice, then we tested rat, then we test it on baboons and monkeys on humans. Then one human than 10 human, then 100 humans. With the clinical trials. The cost of a new drug is two point five billion dollars because it takes such a huge cost because of the clinical trials and so on. What if you could simulate everything you could accelerate the COVID-19 vaccine. I mean, I shouldn't say COVID, but this is the vision. So what I am trying to do is to solve the societal problem. Move ANSYS towards a vertical of healthcare and essentially use simulation for pharma companies for medical device, for pharma companies like Novartis and Pfizer and so on. For medical device companies like Medtronic and Stryker and GE.
Making pacemakers and CR machine and so on. Whole new clinical applications that will drive different... For a doctor to just before they perform surgery. I mean, they will use an ANSYS simulation, the backend to say, "Hey, if I were to do this heart surgery at this point, how much blood will flow?" I mean, essentially it could do all those things with Clinical Apps. This is the grant vision that I am excited about and you can clearly feel the passion in my voice.
Ali Tabibian:
Right. I can. Well, Prith, So maybe that's a really exciting note and a futuristic and hopeful note for us to wrap this up. You've been very kind and generous with your time. And I couldn't ask for more. I really appreciate it.
Prith Banerjee:
Thank you so much. I mean, it's always a pleasure. I really enjoy these interviews with you, Ali. So thank you so much for inviting me. I wish we could have done it real life because it could have been so much more fun. As you know, I'm a very passionate guy. I just jumped up and down the stage, but this audio, the energy will still come across.
Ali Tabibian:
Believe me Prith, it definitely does. And I wish it was in person and I mean, this quite sincerely, I always feel like I'm leaving a little smarter after I talked to you and I'm going to listen to the audio a few times just so I can relive it and relearn. I really mean that. Because live in the COVID environment, of course, this is a remote and we're going to stop the recording but Prith and Tom, if you could just stay with me and not end the call, that would be great. And we'll be done 30 seconds after stopping recording.
Prith Banerjee:
Thank you.
Ali Tabibian:
Thank you so much.