How Robots and AI Are Rebuilding Solar Construction From Scratch | Soliverse Ep. 40

What does it take to build utility-scale solar at the speed the energy transition actually needs?

In this episode of Soliverse | The Future of Solar, Peter Pongracz speaks with Matt Campbell, CEO and Co-Founder at Terabase Energy Inc., about the future of solar construction, AI-driven robotics, energy storage, and the digital systems needed to scale solar from megawatts to gigawatts. Matt explains why traditional solar construction is still largely manual, why the industry needs integrated workflows instead of disconnected tools, and how Terabase is building a platform that connects yield simulation, construction management, data, AI, and robotics into one end-to-end system.

The conversation goes deep into the pressure created by data center power demand, grid constraints, battery storage, project scale, and supply chain realities. Matt also shares how Terabase approaches products like PlantPredict, Construct, and Terafab, why physical AI is changing what robots can do on real construction sites, and what it could take to build a gigawatt of solar in less than ten weeks.

Connect with Matt Campbell:
LinkedIn
Terabase Energy Website

Today’s Partner: Enerdatics

Looking for better deal timing and granular M&A insight in solar and storage?

Enerdatics helps renewable energy professionals identify transactions early, track trends, and price assets with real deal data.

Use code SOLIVERSE10 for 10% off Enerdatics products and services.
Terms apply.

Explore Enerdatics

Transcript

Coming up...

Matt Campbell 0:00

The construction is pretty antiquated. The way that projects are designed and built today is really largely unchanged for 20 years.

Peter Pongracz 0:08

That is Matt Campbell, CEO and co-founder of Terabase Energy. His company is building the platform to take solar construction from manual processes to AI-driven robotics, and they just raised 130 million from SoftBank to do it.

Matt Campbell 0:24

We can process a module every 13 seconds. So depending on the module rating, you know, you can do a megawatt or more in eight hours. It's fully automated.

Peter Pongracz 0:34

In this episode, you will learn how Terabase's integrated platform connects yield simulation, construction management, and robotics into one end-to-end workflow and why Physical AI is replacing deterministic robotics on the construction site and also what it takes to build a gigawatt of solar in less than 10 weeks. I'm Peter Pongratz. This is Soliverse, the future of solar.

Matt Campbell 1:06

So we're going to do like a half-day customer workshop where we kind of have a few panels, kind of go through some of our uh technology and product and whatnot.

Peter Pongracz 1:16

So yeah, so it should be good. No, it's definitely an interesting

Why customers are curious about solar construction robots

Peter Pongracz 1:20

place to be. And uh uh for you guys, even it comes to customers, like is it still a lot of explaining of how the product works, etc. I I assume like when we are looking into um uh terrafap, uh especially that for a European market, is a it's a little bit different fit, right?

Matt Campbell 1:37

Yeah, yeah. No, definitely there's um uh you know, we've got a few different products. Um, so we've got sort of digital products, uh, got some AI things we're working on, as well as TerraFab. And uh yeah, definitely it's something very different. So people are people are all interested. People love robots, uh, and uh so people are always curious to learn more about it. Um and uh, you know, European market is different than the US market, but a lot of the a lot of the people in Europe are involved in projects throughout the world, right? Uh Africa, Middle East, India. And so, you know, in the U, in the Europe, you don't see quite the big ground-mounted systems except for Spain. Um, but uh so the really the market for terrafab would kind of be Spain, but then, you know, MENA region, things like that.

Peter Pongracz 2:27

For sure. And I mean, I I I would think that it would make more sense in a place like Australia, for example, right? Where you have a lot of like flatland and and big pieces in one parcels, right?

Matt Campbell 2:37

Yeah, definitely it's a great product for Australia. And

Are gigawatt-scale solar projects becoming the new normal?

Matt Campbell 2:41

and I think if you look at what's happening, uh, I would say outside of Europe, um, the projects are getting bigger and bigger, right? So in the US, there's multiple two gigawatt projects that are in development or early construction. Um, in the Middle East, there's a what the five gigawatt in the UAE. Um, and increasingly the projects we see on the drawing board are two gigs, four gigs, six gigs, and and I think that's the future, um, whether it be to serve the load of a large data center, and the data center clusters are getting huge, right? A gigawatt, even um our partner, SoftBank, just announced 10 gigawatt site in uh Ohio and the United States. Um, so it it uh the trend is towards these massive load centers, which would require huge amounts of solar if you want to use solar to power them. And then and then I do I still believe that the future will involve green hydrogen and its byproducts like synthetic fuels. Um, and to do that requires just a massive scale up from what we know today. So I think that's really where when we think about robotics, we think about going from you know, the world of 50 or 100 megs to the world of 500 megs to five gigs. And whether it's Australia, Morocco, United States, South America, I these projects are are are here more or less.

Peter Pongracz 4:05

So yeah, for sure. And I mean, you've touched on the market a little bit already, but and also your your partner uh softbanks. So you guys faced uh round C about a year ago, right? So that time the industry narrative was really about urgency, you know. Uh the power demand was and still is exploding, data centers couldn't wait, the construction timelines need to get cut in half or even quarter, right? Uh now that was about a year ago. Fast forward

Why the urgency around power demand is accelerating

Peter Pongracz 4:34

to today. What is the your take on like how that urgency actually has played out?

Matt Campbell 4:40

I I think the urgency is accelerating. I think that in the and I mean it's been unbelievable in the last uh 12 to 18 months where uh the momentum around these big data centers has just accelerated. And everybody is desperate to bring power, you know, say speed to power. How quickly can you get power online? And we've hit the constraints of the grid. Um, so yeah, there are certain uh grids in the United States where they won't allow the interconnection of new data centers. And even huge grids like uh ERCOD in Texas, um, they're hitting their limits. And um, so then, you know, how do you do more behind-the-meter solar? How do you integrate batteries? How do you how do you serve the loads of the data centers? And and so I think the energy around that has uh increased a lot. Now, people are hybridizing, hybridizing with uh natural gas and other forces, so for sources of generation. But even then, I mean, I think the answer is sort of all of the above. Um, even the Republican Party uh seems to have shifted in the last uh three to six months, uh, where maybe there's a little bit more political support for solar that wasn't there um earlier last year. So definitely I think a lot of it, a lot of interest in the market hasn't slowed down at all in the United States.

Peter Pongracz 5:58

Yeah, for sure. And I mean, from from my experience and what I hear from people in the market is that, you know, as long as the maths add up, you know, it doesn't matter what political headwinds you face, you know, what the issue is, like everybody's voting with their bullets, right? So if you are still able to generate the cheapest power source in the world in the fastest period of time in terms of deployments, right? Uh if you look at the graph of like how fast can you actually churn out a large solar project, it's I think the last data was somewhere around 16 to 18 months. So that's uh that's huge. And I think that as as you can appreciate as well from uh the Terabase's perspective, most of that time is actually not really the construction or the actual physical build-out, right? It's more about the permitting and all the paperwork around it that slows the whole project down.

Matt Campbell 6:47

Yeah, I mean, like sort of the challenge that we've set is how do we build

Can solar be built at gigawatt scale in ten weeks?

Matt Campbell 6:50

a gigawatt in 10 weeks? You know, and and and it sounds a little bit aggressive, but I think it's completely doable. I mean, provided you have your permits and all those things, but physically, how do you build really fast? And um, because the demand is there, and um, you know, and I think this recent war is is exposed once again our vulnerability to geopolitics and different sources of fossil fuel and you know, whether it's the effect on Europe or or really anywhere. And I think solar uh with storage has really been the big trend, which um I think is maybe we're really seeing accelerated in the United States. Um, so you know, today it's lithium-ion batteries, but even long duration storage is coming quickly. And uh, and as soon as you've got cheap long-duration storage, um, there's you're sort of unlocking uh unbelievable amounts of demand.

Peter Pongracz 7:47

Yeah, for sure. And I mean, you've touched on the longer-term storage, right? That's that's that kind of ties into the 24-7 problem, right? That we have with uh solar data centers, they require 24-7 power. And then that needs to be a hybrid type of solution, whether that's solar plus storage plus even uh gas or or wind, you know. Basically the standalone solar days are over, right? We are only looking at hybrid projects which will help us bring that flexibility to the grid. So uh what's your take on that?

Matt Campbell 8:18

Yeah,

Are the days of solar without storage numbered?

Matt Campbell 8:19

I totally agree. I think the days of doing a project without storage uh are numbered. I mean, I think um, you know, most of the projects we see now include lithium-ion battery, and and um but I think the and then coming a few years later will be the long duration and um you know, and there's some different technologies that are being you know piloted right now, but uh but I think it's it's it's gonna come and then and then you can sort of get to 95% solar and the rest you can use natural gas as a backup or something. So I I think that um I think we're getting really close. And I think you can do that for a LCOE that's less than 10 cents a kilowatt hour for baseload solar, which is I think if you're in the eight cent or seven cent range, you're you're good because you know the price of building a um, you know, natural gas combined cycle power plant has increased a lot. And you know, the price of natural gas is low in the United States, but it's very volatile and exposed to price shocks that will certainly happen over the next 20 years. And um, now of course Europe's already feeling the price shocks. Um, so so it's you know, uh if you can have certainty over your cost and you're below 10 US dollar cents a kilowatt hour, I think that's pretty, pretty competitive.

Peter Pongracz 9:43

Yeah, I mean that's definitely a speed spot when you know you've touched on the the geopolitical side of things of what's happening in Iran and especially you know how that that's impacting Europe, you know, and that's kind of pushing us in the direction of all of these more localized supply chains, right? That the manufacturing is going to come home in some way, shape, or form. So when it comes to that, of course, we are never going to be cost-competitive with China, but hopefully with uh the right tools and the right tool set, we will be able to manufacture, for example, modules like we are seeing in Texas, et cetera. And then it's going to be easier and and better to meet at least the domestic demand, right? It's never going to be cost competitive to be exported, but at least the domestic demand and hence the whole cycle of geopolitics and and energy self-sufficiency will be covered, right?

Matt Campbell 10:35

The solar module supply chain is a is a is an

What solar manufacturing needs from supply chain diversity

Matt Campbell 10:38

interesting question. I mean, I think Europe, Japan, and the United States have seeded leadership in solar panel manufacturing. Um, you know, there's these um you've got you know folks like First Solar that have done a good job scaling up in the United States, and then there's some smaller silicon module production facilities by Q cells and others, um, which are which are good. Um, but you know, I think it seems that there needs to be a lot more. Now, Elon Musk has made a big announcement around 100 gigawatt of U.S. manufacturing. That is probably the scale you need to think about. Um, you know, right now we have an unfortunate situation where solar panels in the US cost three times what they do outside the US. And uh, you know, and that's not good for the overall LCOE. And um, but at the same time, we um we need a diversity of supply chain. I I used to make a joke like, you know, we wouldn't want all of our solar panels in the world to be made in France, right? Nothing against the French, but it you just never want to concentrate for for any number like what if there's an earthquake? What if there's this, you know, so you want diversity of supply. Um, and uh right now it's it's not the case. So um, and and there have been various attempts to try to build consortiums across different countries to kind of bring up a upstream supply chain. And I think those attempts have flailed. Um so you know, we'll see India is doing some things, which is good. Um, the US is tariff heavy, um, but um, but I I don't think it's really getting to the root of the issue.

Peter Pongracz 12:13

Oh, for sure. And and of course the LCO dictates all of these mechanisms, right? So how do you build and at what cost do you build, which is the most important part? So uh maybe that's a good question disagreeing to your kind of approach. So can you explain the audience how did you guys build a company which has a vision of building the full stack of solar deployment?

Matt Campbell 12:36

Yeah, I mean, so when we started the company uh seven years ago, as we spun out of Sunpower, um, you know, we we kind of posed the question, well, the first question was what's the for future of utility scale solar? And our theory was it's gonna get really a lot bigger, um, both in terms of overall global market size as well as individual project size. And uh, and it's really, as I as I say, there's no choice, right? The solar market has to get bigger because we need that source of clean energy. And there's really no other source other than fusion, uh, but that may be a ways off. Um, solar is cheap and abundant, and uh there's no reason we shouldn't be harnessing terawatts of it. Um, when we looked at that and looked at the market in the future, um, we said, you know, to get to terawatt

Why Terabase is building the full stack of solar deployment

Matt Campbell 13:26

scale solar, you know, that's where the name of the company, Terabase, comes from, terawatt baseload solar. Um, as we look into that future, we needed to find ways to scale the industry faster. Um, and we needed the way to cut the cost. Because although solar is cheap, it's still too expensive, in my view, uh, too expensive from the standpoint of we want to make it cheap enough to make synthetic fuels economically and things like that. Like if it's just solar on the grid, it's already cheap enough. But if you want to make molecules, I would argue it's not cheap enough. And uh, because all the efficiencies and conversions and all the other things that take place. So we we said, okay, we want to build a platform that allows us to help the industry grow as well as cut the cost. And then you say, well, how do I solve that problem? Well, and you say, well, the solar panel's cheap already, right? Like I like to make a joke, it's cheaper to buy a solar panel and put legs on it than to go to IKEA and buy a desk. You know, the the solar panel is cheaper than the IKEA desk on on an area basis, right? So the the panel has been, you know, wonderfully uh high performing and low cost. But the construction is pretty antiquated. Um, the way that projects are designed and built today is really largely unchanged for 20 years. Um and uh and there's a lot of reasons for that, but it's uh super low-tech, it's very manual, manual in the design, manual in the project management, manual in the actual performance of the construction. And and so our theory was that's got to change. And it has to change, it will change. And um, and the way to do that is is really three things: digitalization of the construction, uh, automation of the construction, and then AI integration in the operations, right? Um, and and those three things interrelate. You know, you need data for the AI to make decisions, and you need robots to implement the decisions, right? And so what I imagine is, you know, we've got today where everything is manual, and then there's a point X number of years in the future where everything is run by AI and robots, and then there's this period in between that could be 20 years or 10 years, I don't know how many years, but some period. Um, and so we said we need to build a platform that puts us on this path to this future state where solar can be built, as I said, a gig in 10 weeks, and it's cheaper and it's you know high quality and all this kind of stuff. So so that's that's why we decided to build this integrated platform. Um, you know, our our sort of goal hasn't changed since we started the company. Um, we'll start with automating the design and performance modeling, optimizing the design through uh permutations, simulations, um, and then did you know, overseeing the construction and then automating the construction, automating the commissioning. And uh, and it's you know, we we kind of have an expression which is the cost of solar dies a death of a thousand cuts. So, you know, we're building an integrated set set of bandages. Um it's um because a lot of times people want to see some magic invention. Like I've got a robot and it's gonna solve the solar problem. You know, well, no, it's not that. It it's so much more complicated than that. Because and that's where we look at digital systems, AI systems, data networks, material tracking, robotics. It's it's the synthesis of these things. Yeah.

Peter Pongracz 17:00

Yeah, and and and on that note, I I've seen you know most of the software in uh in the solar business by now. And you know, what I've uh experienced is there are too many silos, right? And uh what you guys are are proposing, and what what is essentially the kind of uh thought process behind Terabase is that you guys are trying

Integrated platform or best-of-breed tools?

Peter Pongracz 17:23

to build the the end-to-end system, which is very difficult. So maybe to play the devil's advocate. Uh if a company would want to use, for example, the best solution uh purely for energy modeling, the best solution purely for construction, the best for Scada, how how would you guys tack up? What is your kind of best uh defense against that?

Matt Campbell 17:45

Yeah, I mean it's a good question. Uh best best to breed versus an integrated platform. I think that we can see case studies like like if you take Apple, for example, so they're they've really taken an integrated platform, you know, hardware and software working really well together, and it's a really elegant, customer-friendly experience. You know, the Apple note-taking app may not be the best note-taking app. The camera app or the photo app may not be the best apps. Um, the email may not be the best, but as long as they're they're very capable, it's the integration, you know, and same thing with like Google, like they've got an integrated suite that makes it really seamless. And um, you know, I think you can come close to best in breed, um, and especially with AI, I mean, it's becoming so easy to develop software. It's crazy. And um, so the barriers um to building software are falling, uh, in which case I think the ex advantages of the integration, you know, like like even within our company, I can see it where, you know, we've got one app for our expense reports, one app for the accounting system, one app for project management, one app for um HR. And it's kind of crazy. I mean, there's so many logins, so many systems, they don't talk to each other. Um, and uh, so I think as you relate to solar workflows, I think it's okay to have, you know, sort of a broader set of skills that are integrated as opposed to, you know, maybe I'm not as good at PV case. But PV case is trying to do everything. It's trying to do rooftops, it's trying to do fixed tilt, it's trying to do like, and I don't maybe I don't care about those things. I care about big power plants with tracking systems, right? I've defined the problem more narrowly. Aaron Powell True.

Peter Pongracz 19:35

And I mean, then to a certain extent, economies of scale takes care of the rest as well, like, right? That you're focusing on one problem, you see the problems recurring in that specific vertical, and you are able to build a solution that's better than what's out there for sure. And um on that note, if we segue into probably the most tricky word in the industry, uh yield simulation. So I had uh your colleague David

What makes PlantPredict different in yield simulation

Peter Pongracz 20:01

Spieldener on the on the show before, and he's very passionate about you know bankability and yield simulation. So why don't you give us um a quick introduction about Plant Predict and why it's better than what's out there in the market?

Matt Campbell 20:14

Yeah, I mean, uh so the history of Plant Predict, this was a tool developed by First Solar um a long time ago uh to help model the performance of their modules and and the projects that they were developing when they were integrated with EPC. Um, we acquired it from them and we've you know enhanced it substantially to include um you know three-dimensional performance analysis, uh, you know, sub-hourly analysis and other things. So we're really trying to say, okay, in the real world, there's topography and there's row-to-row shading, and there's short timescale events that affect performance and stuff like that. So we're really trying to bring in a level of uh fidelity by leveraging the compute of the cloud so that we could do things that you can't do on desktop software. Now, now the interesting thing about it, and obviously PV Syst is is the industry standard. Um and uh and and one question that always comes up is why? They're like, oh my goodness, we're still using PV Syst. How come it hasn't gone on the cloud? And and I think in in a lot of industries, and it I've I see this in civil engineering, hydrological engineering, electrical engineering. There are tools that were developed decades ago that have become the standard. And even though technology has advanced massively, uh they use the old tool because everybody knows it, the banks know it. It's sort of the comfortable solution, even though everybody knows it's not state of the art, right? And uh, and I think I think the solar industry, you know, no offense to the great people uh at uh at PV Sys, but but I think that it's sort of where the industry is right now. Even though obviously harnessing cloud computing, we could do things you could never do on your desktop, right? So um, so in terms of plan predict, I mean, the goal is to have, you know, extremely accurate yield simulations. And then the other thing that we're doing is calibrating it off real world performance, right? So we're building the SCADA system for dozens of power plants. And what's crazy is that most performance models, you know, they're calibrated based on a small test array, but they're not calibrated on a gigawatt power plant, right? And most of the power plants are underperforming in the United States and elsewhere. So uh so clearly there's something that's getting missed in the modeling. Um, and uh so I think if you truly want a bankable system, you've got to calibrate it with the real world.

Peter Pongracz 22:54

But it's funny how there's no uh true feedback loop, right? What you designed and actually what came out uh as a project and how it performs, because uh again, this whole process is siloed as well, right? I I designed a project, somebody else builds it, somebody else operates it. And then basically the stakeholders have very different uh motivation and very different uh views of angle, right? Of how this should be actually working. So then there is no real

Why solar still lacks a real feedback loop

Peter Pongracz 23:23

world feedback of how it's actually performing in the real world.

Matt Campbell 23:26

Yeah, totally agree. And it's it's crazy. And and there's a lot of examples within solar where there's no feedback loop. And and part of it is it's the siloing. It's it's uh, you know, the module company makes modules, the construction company attaches the modules, you know, a different engineering firm designs the project, an owner operates the project, and and everybody's kind of guarding their data and it's not flowing. And and so you're really missing a lot, a lot of opportunity to do things better, which is unfortunate. Yeah.

Peter Pongracz 24:00

You guys are trying to build an ecosystem, right? So the first step would be within that ecosystem, the yield simulation within plant predict. Uh, if we move to the next step, uh how about the construction? How does, for example, a product like Construct work and

How Construct creates a digital twin of the construction site

Peter Pongracz 24:17

why is it different than uh then the typical uh use case or how it's typically built?

Matt Campbell 24:22

So basically the what construct is, is we take the um construction grinds and create a digital twin, and then we bring in the official schedule for the project, which in like in the United States, it's like a primavera P6 schedule. We utilize drones, we utilize other data acquisition technologies, mobile phones, other things, uh, to basically bring in all the data about what's happening on the construction site. And then the owner and the EPC kind of have a clear view into the what is happening. And um, it's a like a single source of truth. And then we track, you know, productivity, we track quality, uh, we track, you know, were things done correctly? Like if you graded the site, was the grading performed correctly? Um, and we fly drones every week or even every day increasingly. Um, and so like right now we're active on 62 different power plants. Um, the total fleet that we've addressed is 25 gigawatts. And it's and it's you know, the alternative mostly, I mean, there's a couple other competitors that that do somewhat similar things. Uh, and then also people just don't do it. They don't, they don't build a digital twin or they fly drones to take pretty pictures, but they're really not diving deep into the analytics, right? Um, so so that's been great. I mean, I think we've seen huge growth of it, you know, like we've uh doubled year over year because people really see the value into that clear view into what's actually happening on the project.

Peter Pongracz 25:54

That's interesting because you know we we sometimes fall into this fallacy in in solar that we are either in the physical dimension or the digital dimension, right? And it's quite rare to see that people would combine, you know, hardware and software together under one roof because they are slightly different toolkits and kind of uh skill sets as well to build software company to versus to build a hardware company, right? And but if you get

Why AI needs a bridge between software and the physical world

Peter Pongracz 26:22

the two to work together, that's where the magic happens, right?

Matt Campbell 26:26

Absolutely, yeah. And and you know, I I I remember talking to investors in the past, and they say, Are you a SaaS company? And I'm like, No, we're absolutely not a SaaS company. And and they don't like it because everybody loves SaaS companies, you know, it's no hardware, no CapEx. Um, but the reality is SaaS is getting commoditized, right? Like a superintelligence in a year could build a yield simulation tool in a day, right? Like, I'm not sure if it'll be accurate or not, but I mean, you I guess you could do it today if you wanted. Um, so a lot of this stuff's gonna get commoditized, but you need to build linkages to the physical world. And like if you think about AI, AI only knows what's on the internet, right? So like the AI can go read every NREL paper and and every whatever, you know, Fraunhofer report. And um, but it's still gonna be limited on what it what is going on in the physical world. Um, and that's where, you know, I like to say the AI needs uh contextual awareness, like total contextual awareness. Where are the people? Where are the machines? Where's the material? What's happening? What's the weather? Like a real-time view into the physical world. It's like giving it sense sense senses, eyes and ears and nose, and and uh so the AI needs that. Um, and to do that, you have to straddle that physical and digital world. And then, and then the AI needs hands, it needs to build stuff, and that's where robots come in. So you need the data to come in and then the commands to go out. And so that's where I think without any doubt, it's that physical digital bridge which is the key. Um, yeah, if you want to trick if you want to change things.

Peter Pongracz 28:13

Yeah, indeed. And and you know, there is a lot of talk on the streets, especially since the explosion of AI, right? That SaaS is kind of dead, right? It's getting more, as you mentioned, more and more commoditized. But if if you are able to build linkages to the real world and to be able to differentiate yourself in another manner where AI has a blind spot because and it will potentially have this blind spot for the foreseeable five to ten years, maybe, then then it's going to really make your life much easier, which brings us to our next point, right? The the f physical AI connection to Terrafab, the version two, right? So you guys have released version one uh a while ago. What is the kind of difference between v1 and v2 and and and how is that going to position you yourself in a in a way

What changed from Terafab V1 to V2

Peter Pongracz 29:05

of churning out those terawatt hours of solar?

Matt Campbell 29:08

Yeah, so the so the v2 system is um so the v1 system, we would we did five pilot projects and it was excellent learning. Um, you know, we worked with the big EPCs and owners in the US, and you know, we had to deal with um uh rats and snakes and bees and mud and snow and dust tornadoes and hail storms, and I mean, we really got an education in trying to do robotics outside. Um, and uh, and it was good. We learned a lot, and then we took all that learning and put it into our second generation system. And what's different about the second generation system is it's designed for really high throughput. So, like we can process a module every 13 seconds. So, depending on the module rating, you know, you can do a megawatt or more in eight hours. Um, it's fully automated. So we take like a tracker, all the mounting hardware, attach the modules, we inspect the modules um for defects, so before you even install them. And uh, and then we deliver and and and land the tables out in the field. So uh so high speed, high quality, fully automated. We're setting it up for 24-hour operation. Um, it's very portable. So, of course, solar sites are very big. Uh, so you want to move the factory, we can move it in four hours, uh, which took a lot of design work because um it's a complicated machine. And um, and then um we also are working on, so we have this um a second generation delivery rover. Uh so this rover takes a finished good from a factory and brings it out to the field for installation. And so we've released the second generation of that, which has a number of improvements. And that was a vehicle we developed from the ground up. And um, you know, and it's all integrated with our sort of digital manufacturing control software. And um, and then we have an autonomous

How physical AI accelerated Terabase’s robotics roadmap

Matt Campbell 31:10

version already operating. Um, so it's not commercially released yet, but we have an autonomous version operating. And then as it relates to physical AI, uh, the physical AI tools have rapidly accelerated our roadmap. So um, we recently posted a video online, and in that video, we showed a real test that we did where um the rover, our our placement rover, automatically swaged a torque tube and set it down. And that was a technical goal for the end of this year, but we did it last summer, and it was because we trained an AI model to recognize all the features, and we put that on top of this very complicated actuation system, and just in a sort of an open loop control system, we were able to do this outside in different lighting conditions and with millimeter accuracy. It was incredible. And, you know, I kind of had two aha moments with physical AI last year. One was when we did that. Uh, I was like, oh my gosh, like that was way easier than I thought it was going to be. And then the second was uh Tesla released some updates to self-drive. I've a Tesla, I'm I'm embarrassed to admit, but um, so Tesla introduced uh these updates to self-driving and they blew my mind. They were so good. And it was because of the physical AI that they've integrated with the self-driving. Um, so so like we've experienced it ourselves, I think throughout the industry, it's of course it's become a buzzword, but the reality is it is transformative because the robotics that we were using previously are uh deterministic. I mean, it's like you pick something up and you put it here, and you pick it up and you put it here. But in outside world is dynamic.

Peter Pongracz 32:57

True, especially when you mention weather conditions and and you know, hail and storm and uh, et cetera, it's it really changes, you know, how even a human perceives the world, right? Not let alone the robot. And uh, which brings me to I I think it begs the question of how do this actually juggle so many different verticals, right? Because these products are essentially like almost like small startups within the Terabase uh umbrella, right? And and it they do require very different uh skill sets of

How Terabase manages multiple hard problems at once

Peter Pongracz 33:30

people to be able to operate on their own, let alone tie the whole thing together.

Matt Campbell 33:35

Yeah, well, we uh we're fortunate to have you know really you know a large and a high performing team. Um so you know, we have you know a couple hundred engineers. So I think it's it's about um you know having a well-organized plan and then executing to the plan. Um, you know, I think everybody at the company's extremely passionate about climate and uh scaling up solar. And uh so I think we've been able to, you know, build a really high performance uh team. Uh, but it is a lot. And you know, we constantly have to say, you know, what what are the most important things that we're doing and what are the less important things and you know, prioritizing accordingly because it's true. We are we are doing a lot. I think it's just inherent in what we're trying to do. It's it's sometimes I use the analogy of SpaceX, right? Like SpaceX could have said, I'm gonna develop software that designs better rockets, and then I'm gonna call Boeing and Lockheed and try to get them to buy my software. And like, would that have been successful? And the answer is no. You know, they had to they had to design their own rocket, their own machines to manufacture the rocket, they had to build their own boat to catch the rocket and their own uh place to launch the rocket. And that, you know, and and maybe ours isn't as crazy as SpaceX, and they have a lot more money than we do. Um, but if you really want to make a difference, sometimes it takes a huge effort, a huge investment. It doesn't, you know, otherwise you're just like, I made this, you know, oh, I'm automating a CAD design. Okay, I made it, I made it better. That's great. Um, but it it's just a small piece of the overall puzzle, right? And um, as I like to say, there's no free lunch. You you sometimes to make a step change in an industry, you you gotta take, you know, same thing with Tesla, right? They had a, you know, who would have thought that you could create an electric car company from scratch? I mean, it it was crazy, right? And then um, but they did it. And sometimes you sometimes it's just the only path, I guess.

Peter Pongracz 35:38

Sometimes you need to take those little shots, right? I guess in the person's mind, uh the idea is always a little bit easier when it's born, right? But when you actually step on the path of executing, then then you realize okay, how many things can go differently or wrong or whatever the case might be compared to what you actually thought it would do. And and that's the the difficult part, right? The execution and to to stay on path and to be able to execute on your vision and mission, regardless of what happens around you.

Why scaling solar requires a solution as big as the challenge

Peter Pongracz 36:08

It has always been an ambitious goal.

Matt Campbell 36:11

Um but when you think about what's trying to be done here, you know, the world needs 75 terawatts of solar. And that's more infrastructure than has ever been built. I mean, it's just it's like all the descriptive statistics are off the charts, right? Like in terms of, you know, enough solar panels to go to the sun and back, and you know, and you know, massive amounts of steel and copper and aluminum. Like, well, what needs to be done is just vast. And therefore, the solution needs to be pretty big. I think it just has to be done. And, you know, so we're in a good position to help do it. And um, you know, I think things have proven to be harder than I would have expected. I'm I'm kind of an optimist. I tell my wife I'll be home for dinner at seven and it's always eight, right? No matter I'm convinced it's gonna be seven, but it's always eight, right? Like, and uh, you know, it's just that as an entrepreneur, you you tend you have to have kind of an optimistic view on the world. And uh, and some of the things have proven to be hard. I mean, the outdoor robotics is extremely hard. And sort of I, you know, now it's it's becoming so trendy to have, you know, robots and humanoid robots. And I kind of chuckle about it because wait till these robots go out in the mud.

Peter Pongracz 37:24

It's yeah, it's it's going to be a different uh different battle for sure.

Matt Campbell 37:29

Yeah, I mean, anybody can make a YouTube video, but it's it's a totally different thing to actually do real work.

Peter Pongracz 37:35

Yeah, maybe if we uh pick a product from a European perspective, let's say construct in Europe, how did that stack up versus your experience in the US? So when it comes to the EPC structures, the interconnection process, maybe the tracker systems as well, right? Because inherently we we haven't been really using that many trackers in Europe. Now it's becoming more or more and more of a standard because of

How Construct compares across Europe and the US

Peter Pongracz 37:59

the uh Docker, right? But uh how how is that different uh between the US and Europe?

Matt Campbell 38:05

You know, construct is uh it's pretty uh similar, I think, in Europe, because more or less the design of the power plants, even if it's fixed tilt, is pretty similar, right? You've got rows of panels and electrical collection systems and combiner boxes and inverters and medium voltage collection and a substation, right? So the ingredients are really the same. And um again, even if the projects are smaller, you know, 50 megs, 25 megs, whatever, the software really doesn't change that much. So I think I think we see, I would say pretty, pretty common, pretty common approach. I think it's it's more of like when you start to think about robotics that that you could see differences. And that's really where, you know, like we've designed ours specifically for bigger type deployments. Um but uh you know a lot of Europe is a heterogeneous set of a lot of fixed hill just owing to the latitude, yeah.

Peter Pongracz 39:04

True. And uh when it comes to the the one continuous workflow of an asset owner, like how is that differing, let's say, from the financial modeling through energy modeling uh or maybe operations and and how are you guys treating that within Terabase?

Matt Campbell 39:24

The sort of the workflow, you know, again, we're kind of coming not from a financial model side, but from the physical asset side. You know, we'll have uh customers, you know, they do their energy model and plant predict, and then they build it and construct, and they they know the geolocation of every module and its flash test and all of that. And um uh and then we perform a capacity test with plant predict, which looks at the yield of the modules and the circuits and comprehends the topography, and that you know, could get you into an instant capacity validation. Uh, and then would go into the operation. And then you can take the operational data and then compare that to what Plant Predict told you it's gonna do. So I think that's that's sort of the life cycle. And um, you know, we've stitched those pieces together and um and really try to make that whole process much easier for the owner, right? Like even today, 99% of the power plants have somebody that's got a spreadsheet and they're loading all the flash data and trying to figure out, you know, how much how much flash is on each inverter, and you know, like then pulling out some equation and trying to determine if it's performing correctly. And yeah, we're trying to make that all seamless.

Peter Pongracz 40:45

Yeah, as we mentioned at the beginning, you know, the the planning is one thing, but the real world is a very different one, right? For sure. Yeah. And Matt, uh, we are actually um approaching the end of our time. Is there any special way that people should reach out either to yourself or Terabase? What are you guys looking for? What type of people would you like to speak to in the industry?

Matt Campbell 41:06

Yeah, I mean, we we always love to speak to all types of companies, whether or, you know, it's uh an owner, an EPC, an equipment OEM, um, another software company, another robot company. Um, you know, for me personally, I'm I'm just passionate about this industry. Um, someday I have a goal to open a solar museum because I don't think there's one that exists. Um so I'm I'm really first and foremost about the industry. And uh, and so we're always open to, you know, how can we help our customers? How can we collaborate with other tech companies? Um, and then and and really how can we drive the industry forward. So uh, and in terms of connecting with us, um, of course, you can always uh come through our website. And then we also participate in, you know, we'll be at InterSolar Europe, uh, obviously the big trade shows in the US, uh, and Australia. And so, yeah, reach out to us. We'd love love to chat.

Peter Pongracz 42:07

We'll link to all of those in the show notes below. And uh, I will definitely see you in uh in Munich for sure as well. So uh on that note, Matt, thank you very much for your time. I appreciate you being here and sharing your knowledge. So can't wait to see you again.

Matt Campbell 42:19

Thanks. Yeah, great to talk and have a good day. Yeah, YouTube.

Peter Pongracz 42:23

Bye.