At Data Center World 2022, Autodesk vice president of industrialized construction Amy Marks stressed the need to move away from traditional building methods for data centers and towards productizing the building process.
The following transcript has been edited for length and clarity.
Data Center World 2023 will be held in Austin, Texas, from May 8-11.
Amy Marks: So, I like to think of industrialized construction as the house that I see industrialized construction built.
And so, it starts with technology enablers. So, we see IoT sensors on everything. We want to attach IoT sensors to lots of pieces of equipment. Now we're attaching them to assemblies and the buildings, so we can have digital twins and things like that. But big data and analytics is part of industrialized construction. And we also want to think about the cloud.
And then there are some process enablers. And I know some of you are not exactly from the design or construction space or in architecture. But we like to think of manufacturing in the Lean manufacturing terminology. BIM, for those of you who don't know, that acronym is "building information modeling.” It's basically 3D, 4D, and 5D drawings – models, really – to make sure that we understand how these buildings are going to be put together, not just in flat 2D drawings. And we'll talk a little bit about how we need everyone to get on BIM but you have to make the “I” intelligent. It's not enough just to model things.
A lot of you hear this terminology DfMA, design for manufacturing and assembly. Everybody talks about how they want to have prefab and modular data centers and “I want assemblies.” And how many of you've heard, “I want kit-of-parts”? But you have to design it that way. And so, we've taken a word from manufacturing called DfMA -- design for manufacturing and assembly. It's a concatenation of two words: ease of manufacturing and ease of assembly. And in our world, what we messed up a little bit is that the makers in manufacturing actually define DfMA, how to make it easiest, and how to assemble it easiest. So, we heard the word “design,” and we said, “It must be the architects that understand it. They have to be the one that designs this so that I could manufacture it the best way so that I could do a kit-of-parts.” But it's not about design. It's about the data for manufacturing and assembly that we sort of missed in that. And we've expected architects to be able to know everything. And that's kind of impossible, right? Nobody can know. Not even architects can know every Piece Part, every proprietary system that we need to manufacture to make these data centers work. It's impossible.
So, there is on the physicality of pieces and parts that we see in data centers, there are products that we think of like advanced building products, like Victaulic, where we don't have to weld pipe anymore. You'll see single trade assemblies, meaning pipes, spools, or electrical components and things like that, that have been put together in factories. And then we are doing lots and lots of multi-trade assemblies. Most of you have seen skids, or even big open racks with multiple trades on them – electrical, plumbing, all that stuff on the rack – coming from the factories. And then we have volumetric modules. Make sure you know that this is not a hierarchy. Kits-of-parts are just as good as volumetric modules, especially if your module is going far and there's lots of air in it. You don't want to ship expensive air.
And we want to make sure that we understand that continuum, and the data for all those pieces and parts and where they're coming from so that we can build things with robotics and automation. We don't want to weld onsite anymore. We don't want to cut things on-site.
Sustainability is really important to me. And we have to do this. We have to understand this to get a more sustainable world. We want to start building things in the most advanced way. We don't want to do things the way we did them 100 years ago, right? And why do we have to do that kit-of-parts? We have to productize them.
Why Now? The Dynamics Are Changing
So, let's talk about why now. The world has changed. In this report by McKinsey, they said in the next 15 years – this was two years ago – 45% of all the construction and design in the world is going to be done in industrialized construction fashion. Forty-five percent. It's pretty much past the tipping point at that point. And why is it happening? Because we have these emerging disruptions that are happening. New materials are happening. We've seen digitization. How many of you are getting off paper? Hopefully, you're all off paper, and you've converged your business model with digitalization. And we see a lot of new entrants in the marketplace. And so, we see people buying different things. They're aggregating companies. Lots of change, right?
But what McKinsey said is the most impactful thing that would happen in the next one to three years, in the future industry dynamics, was going to be a product-led approach, not design-led. A product-led approach. That was almost two years ago.
And so, 66% of those survey respondents think that industrialization, this industrialized construction, and digitization are going to have the highest impact compared to any of those emerging trends. That's really important for you to understand.
But we have a problem beyond that, just in our normal practices: People design things that can't be made. They make assumptions, and then we can't make them, and then we have risk and uncertainty. And it's iterative, and it's expensive. And that's why it's called “an estimate.” And that's why you have a bunch of line items called “designed contingency” and “construction contingency,” and the owner puts “contingency” on there. We have very little certainty. But here's what I do know: We need digital infrastructure. That I'm certain about. And I know that we need a lot of it. And we need to keep moving it along and getting everybody online.
So, what is the real opportunity here for all of you to be part of this ecosystem? It’s that you and I together, and Autodesk, all of us, have to connect these tools. We have to connect these workflows and make them usable at scale.
Here's what we do today with contractors. In this case, I was looking at a football stadium. What are they interested in? How big is a stadium? What does it look like? How cool does the entrance look? That little Piece Part right there is the guts behind the bathroom wall, and it's MEP –mechanical, electrical, plumbing. It's a little box. In the beginning, they're like, “I don't know. We need bathrooms.” They stick a dumb empty box behind that wall. By the way, in stadiums, people drink a lot of beer. It's not so pretty, but we need those bathrooms, right?
So, we take it off our desktop, which is Revit in our case, which is building information modeling software. It's static. We have a space allocation for it. We're like, “I don't know. It was 20. I took the last one I did – it was 12 bathrooms – and I just stretched it.” And then you give it to the GC with pretty much no information or some information or wrong information or an estimate or something like that – incomplete data. And then my poor friend over here gets it months later, and they're like, “You got to do this for me. Give me a price.” And he's like, “Wait, that's a 12-stall size for this bathroom, and you want 20 bathrooms in it.” That's not going to work.
And so, what do those guys do? They hand it to their fabrication shop manager. If you don't know, all the large MEP contractors these days have fabrication shops. And then he gives it back to the GC, who gives it back to the architect, and it obviously doesn't meet their requirements.
By the way, if you're one of those building product manufacturers, it's even worse. You're like two, three, four levels down with distribution. You don't even know it's the same job. You just gave them some parts that you didn't even aggregate what you're doing, because you're so far downstream.
Here's the bad news: We do this for thousands of interdependent assemblies every single day in data centers. Thousands. What gets reconciled first? I don't know. Steel has to be bought, the GC is self-performing, I'm doing this, my friend is doing this. I'm going to reconcile this one thing that negates the possibility of all these other things. It's done in that way with our mouths and our feet, but not so often with technology, right?
There is a difference between products and prefabrication. Prefabrication happens usually downstream. It could be very good. Products are things like generators. They come in 250s, 200s, 300s.
We have to stop saying, “Just give me an empty space, and I'll make it a unique snowflake however you want it.” We can't do that, because we're not getting scale. And that's creating waste.
The Formula for Change
In order to change, you have to have dissatisfaction. After that, you have to have a vision of what's possible. Don't give up.
And after you have a vision, it's not enough to be inspired. You then have to do something. You have to have a couple of first concrete steps of action to get where you're going to go. And this is the important part: To overcome resistance, both internal resistance at your companies, which is usually harder, and external resistance in the world – By the way, I'm a woman in construction and manufacturing and tech. You don’t think I know resistance?
We can overcome this as an ecosystem. We have to start listening to people that don’t look like us or talk like us or didn’t come from our industries.
And so why is it happening now? Let's ask ourselves. It was happening before COVID. We're seeing convergence happen.
It's funny. When I used to build subsea cable landing stations, we used to say, “Data centers and subsea – that technology is converging.” We've seen technology converge, right?
Processes converging. We're not just contractors anymore. You're not just doing whatever somebody gave you in the drawings, right? You're trying to get up to the front of the process to influence it. So, you’re trying to understand those processes that are converging.
Whole industries have converged. Industrialized construction, prefabrication, the modular companies. By the way, I know owners that own manufacturing facilities. You know architects that own manufacturing facilities now. We're seeing manufacturers understand how to get digital twins in the front of the design process. We see this happening all the time. So, what does that mean?
Business models have converged and changed.
Productization Is Happening
This was last year. GSK said, “This is how we make manufacturing facilities now. Not how we want to make them. Not how we're thinking of making them. This how we make them.” They picked kickball teams. Do you know what that means? Most of the data center guys are picking kickball teams now, too. That's not designed by an architect gone out to bid and then understanding what it could be in that snowflake. That is a product that gets integrated into that manufacturing facility, every time.
Partnering. Are any of you in procurement? Stop that nonsense. It's about 90% of our waste, right?
So, what is happening? At a conference last year, I asked people, “Now that I described convergence business models, how many of you consider yourself a convergence customer?” Eighty-two percent said, “That's me.” And 79% of that audience said they're on a journey to productize.
It's a lot. If you're a contractor right now, if you're an owner that doesn't understand these products and your kits-of-parts, if you're an architect that wants to understand how to move these around, you should be afraid if you didn't know this was happening.
And where's it happening? Mostly, think data centers – heavy MEP, where it's 60%, 70%, and 50% of the cost. Any application where it's heavy MEP, heavy structure, hardened structures, big central utility plants that I don't want to build on the roof anymore because I want to build them concurrently -- that's where it's happening.
So, our companies are changing, and I'm going to go kind of fast through these. But we're seeing owners become “serial owners.” I coined that phrase. Why? Because I thought of it like serial killers back in the day. They want to kill with the same fingerprint, right? “I don't want 50 different data centers. I want the same data center, I want the same hospital, so operationally I'm consistent.” So, they become serial owners.
Building product manufacturers want to do bigger and better assemblies. They're becoming solution aggregators. They're not just happy to be way, way downstream anymore.
And we have contractors. They want to become solution aggregators eventually as they become manufacturers, but right now they're more like products and systems integrators.
And architects and engineers are becoming generative designers.
Quickly, owners again: They don't have one building. They have many buildings. Certainty of cost, schedule and savings, operational consistency, flexibility, and sustainability are more important to them than beating up an electrical contractor for one little bid. They talk about things like this: Design for Manufacturing and Assembly, speed to market, kit-of-parts, reduced outcome variability, supply chain understanding and management and variability that's going on right now, especially now. They want integrated factory models. They want to understand the digital twin of not just the equipment but the buildings themselves.
Those building product manufacturers are becoming more solution aggregators, right? They don't want strut. They want the whole rack. So, they want to get in the front of the process, and they want to aggregate more of their equipment into these. That's why skids happen now. They say things like product platform, protecting the spec, manufacturing, informed design, and design assist. These are all the words they use in trying to get up to the front.
Contractors are becoming more product and systems integrators. “We have to become more like product and systems integrators because that's now where the only value is that we have. We have to understand what things we're integrating.” What do they want? They want standards. They want vertical integration. They want to move more work offsite. They talk about fabrication to manufacturing.
And so, architects and engineers, they're not the generators of this data. That's what I just told you. They don't have the make information yet. So, they want to arrange that information, understand the true parameters of that, organize it using some technology, and disseminate it in the best possible way. That's actually how they make money in the future state. They become almost like embedded technology companies, not just architects, right? They say things like this: Computational design. I think less than 3% of architects at the moment have computational design experience, right? And they say things like manufacturing and assembly, digital twins, platforms, standards, and kit-of-parts.
I wrote this transformation framework a couple of years ago, because I was like, "Listen, it's okay to want things.” I say this to my 13-year-old daughter: “You can want things, but you got to know how to get there. If you want to be a doctor, you have to do a little bit of foundational study.” If you want to be a manufacturer, you can maybe afford to buy a six-axis robot right now, but maybe you shouldn't, because foundationally, you don't have the right skills and tools. You don't have the right people that know how to do it. You don't have enough scale to make it work. And I always say, “It'll end up like your treadmill. If you're not ready for it, it’ll just have dust sitting in the corner or it’ll make scrap faster.” Robots don't make things faster if you don't have enough things to feed them and do them at scale or understand the parameters of what they can do for you.
So, we need some readiness – platform readiness, as well. Once you cross over the silos, we have to understand, “Am I ready to share data across these silos?”
And then you have that productization, both physical and digital. And we have to digitize that. You don't go right to digitization without understanding how to lay it all out first. It's like crawling before you fly. In the best manufacturers in the world, automation happens in the last 5%, not the first 5%.
And this is going to be a gendered comment: Stop buying toys when you don't know what you're making yet. Stop buying land for factories when you don’t know what you’re making yet.
You got to have some digitization, then you need to connect it. And only then can you optimize things like generative design. And then we get to true circularity.
Digital waste, everything we draw that never gets made, exponentially leads to physical waste.
Why Is Data Important?
So, why is data important? Well, obviously, everyone says we want better insights. We want better visibility. You want efficiencies. It's okay to want things. But this is the reality of it: We've moved to remote work, we have more data-driven projects, and we lose and gain our money as owners in the asset lifecycle management, not in beating somebody up for 2% less on that piece of equipment.
So, there's massive information loss as you get to the operational side of the business. And we are not digitized upfront, as much as we need to be to share information across the front of the process, which leads to the fact that we have no information at the end.
And so yes, we show this all the time. And we're like, “You can use the data to inform the planning.” And I'm like, “Well, not if you only have one to build.”
These are data workstreams of shared data for these types of customers. On the far right, you're going to see the manufacturers and the subcontractors. As you move to the left, you're going to see the sub-consultants, the consultants, and the engineers. And then at the far left, you see the owner. Now, think everyone on that far-right uses a million different kinds of software. Then they're using a different workstream, then they don't have a standard, and then they're compressing it and massaging that data. How good is that data at the front of his process?
We have to impact change. We have to allow for some standards and some platform understanding for this guy, let's call him a data center owner, and these players, to have data that is usable.
Product vs. Platform
There isn't one ring to rule them all when it comes to platform anyway, right? You have to understand it's not going to be one product. So, I go into a lot of meetings with CEOs, and they’re like, “Just sell me the solution.” I'm like, “How do I sell you the solution to that problem? I can get you on the journey. I can certainly do that.”
There's a difference between product and platform. And this is important: Products are tools, and platforms are portfolios of products and the surrounding ecosystem of products. They let come in not just their products but others’ and competitors’ products. Products have everything contained in it, like one kind of project management software. But with platforms, the value is from sharing that information across many products. And the feature is the connectivity, not the “point and click and I get this feature.” And then there may be challenges to adding tools and connecting.
That's where we're at right now if you just have a product understanding, of why you want data to rule your life. These products can't be disparate. They have to talk to one another.
With platforms, it's about the ability to connect these with external teams and our data so that we can understand some standards and get some clean data up to the front so that you could use it again and again.
Productization of Data
So, let's talk about the productization of data. Last year, you might have heard Autodesk’s CEO say something unprecedented, I thought. He said we could fix the uncertainty in the phases of designing a building and building a building if we had manufacturing-informed design. And he showed Inventor, our mechanical CAD tool. There are other ones out there, like SolidWorks – competitor tools. And if you could define the rules and constraints of things that were made, designers could consume it in our building information modeling software. And that would be great.
But the fact that we thought manufacturing in front of design is mind-blowing for those of you who are not nerds like me. And we just did a very large investment now in manufacturing informed design because we need to enable this certainty. By the way, listen up, for those of you who are like, “I don't want the same thing every single time”: No, it's one-to-many mapping because we know you're going to do one skid that has certain parameters. You can make 15 skids out of that same model, as long as you can still make it.
So, what does that mean? We go to this opportunity again, and we want to make sure we enable new tools, so that this product when it's made, we start thinking about it in advance. We start understanding that this product needs rules and constraints.
So, this is all happening. We have to think about not just how it's made, but what is its sustainability criteria, because, ultimately, we want machine learning and AI to tell us which product is the right product to pick. That's called generative design. You can't do that unless you have the data backbone for that, for things to be understood. So, we need to understand the physical Piece Part. And then you have to understand the digitized workflow that it has to take. There are going to be thousands of mini workflows like neurological pathways, because electrical needs different functionality than precast, than structural steel.
We have to productize some things and start digitizing these workflows. But we need to generate one-to-many, right? One-to-many. One model, many versions. That's how you actually can get something different. And we have to implement that by enabling dynamic content. People say standards. I say dynamic content that I can adjust within those standards.
That little skid that we talked about needs a little “Dora the Explorer” backpack. It needs a little data backbone. So, when it travels across all these things, from concept all the way to assemble, all the way to commissioning and then operating, I go, “I need sustainability information,” and just like in Dora the Explore, she's like, “I need rope,” and then the rope would pop out of her backpack. She didn’t have to take the backpack off and dump it on the ground. And no, I just said, “I need rope,” and then it pops up. So, you need to build that into the criteria of the productization so that we can get that little skid with its little backbone on it across so that you, as owners, can reuse that data and improve that data and use that product again.
We have to figure this out. You want these things, and you cannot do any advanced workflows and automation, utilizing not just our tools but other people’s tools, to figure out how to do this stuff better using machine learning and AI – not our mouths, not our feet. I'm tired of hearing contractors saying, “Get me to the front of the process, and I'll just fix everything.” No, you cannot. Because even if you're the smartest guy at your shop, you're not smarter than a generative design tool that knows every permutation you've ever done and anything acceptable and allowable, that tells them, “Under these constraints, this is what you should do.”
You could use all these things – if you had a product, physical and digital.
So, we have to build upon manufacturing-informed design and building information modeling, and we have to standardize not just on the BIM standards but on the data structure so those thousands of flows you saw going back up to that owner and data centers are real. You want them to be real so we can make data-driven decisions about them.
How Do We Get There Together?
How do we get there together? Well, I gave up my own business a couple of years ago, after several years of talking to Autodesk, because I realized I was too small, using my mouth and my feet. And I started enabling a group in the company called Convergence Consulting. We help you if you want to be on a business transformation. Not software. Not like, “Hey, you should buy this software,” but more like, “Hey, what do you want to be when you grow up in a business? What keeps you up at night? What are the critical factors of that? What are the business outcomes you're looking for? And what are the capabilities you have now? What are you ready for? What are you not ready for? Are you platform ready? Are you manufacturing-ready? Are you Lean ready? Culturally ready?” I love benevolent dictators, by the way, when it comes to culture. If you want to get stuff done, find a benevolent dictator.
And then you have to connect platform workflows, right? It's not enough to have disparate little pieces of information as we saw. You have to connect that information in a meaningful way.
We have to connect these. And we have to enable platform capabilities. So that we have new partners and new software and more apps.
This is a quick case study before I end. We just did this with a company. There's lots of data, lots of disparate tools, lots of things that they had to connect, and they're all disparate. This is like every company I know, just about.
And then we had to understand what other data they have – not just Autodesk’s, but everybody else's – what software they’re using, what capabilities. And then, what do you want to see it look like at the end so that no one has to massage anything? And how do you apply technologies so nobody has to do it by hand? This is a real job. I just can’t show you the name.
And so, it created a next-gen use case where you could get all this stuff for free. What's the right kit-of-parts approach? What are the design options I should use?
So, at the end of the day, they can get instant product information and what's going on in the manufacturing facility and simulate it. And we want to make sure that you can sell it, so nobody has to sell something that you can't make. Things like configuration upfront to understand the parameters of the customer, so you don't screw it up, so somebody can't make it later. That exists in this case study.
And in fact, here's some of the data stream that we were mapping out. I showed you a simple version of it, because trust me, like in software, this stuff gets so I can't even read it. I'm like, “What does that thing say?” But so, basically, you're mapping out all the databases and all the standards they want to have and where it comes from and what the inputs and outputs of that have to be and where it goes. And what you get out of that, you get these amazing calculators. In this one, they were doing carbon calculation, ease of manufacturing, and ease of availability of those pieces and parts. These were customs for them, that they want, that are part of their business.
By the way, then you can understand data-driven decisions when you're looking at embodied carbon like this one versus this one. That's not my opinion. That's a fact. Why? Because I keep making that product over and over again.
Forty percent of what's in our landfills is from construction waste. Forty percent. Productization creates complete waste avoidance, of cutting and garbage and dumpsters. We have to stop this nonsense.
Thank you so much, you guys, for listening today, and I appreciate your time.