Show notes
Infrastructure was passé…uncool. Difficult to get dollars from Private Equity and Growth funds, and almost impossible to get a VC fund interested. Now?! Now, it’s cool. Infrastructure seems to be having a Renaissance, a full on Rebirth, not just fueled by commercial interests (e.g. advent of AI), but also by industrial policy and geopolitical considerations. In this episode of Tech Deciphered, we explore what’s cool in the infrastructure spaces, including mega trends in semiconductors, energy, networking & connectivity, manufacturing Navigation: Intro We’re back to building things Why now: the 5 forces behind the renaissance Semiconductors: compute is the new oil Networking & connectivity: digital highways get rebuilt Energy: rebuilding the power stack (not just renewables) Manufacturing: the return of “atoms + bits” Wrap: what it means for startups, incumbents, and investors Conclusion Our co-hosts: Bertrand Schmitt, Entrepreneur in Residence at Red River West, co-founder of App Annie / Data.ai, business angel, advisor to startups and VC funds, @bschmitt Nuno Goncalves Pedro, Investor, Managing Partner, Founder at Chamaeleon, @ngpedro Our show: Tech DECIPHERED brings you the Entrepreneur and Investor views on Big Tech, VC and Start-up news, opinion pieces and research. We decipher their meaning, and add inside knowledge and context. Being nerds, we also discuss the latest gadgets and pop culture news Subscribe To Our Podcast Nuno Gonçalves Pedro Introduction Welcome to episode 73 of Tech Deciphered, Infrastructure, the Rebirth or Renaissance. Infrastructure was passé, it wasn’t cool, but all of a sudden now everyone’s talking about network, talking about compute and semiconductors, talking about logistics, talking about energy. What gives? What’s happened? It was impossible in the past to get any funds, venture capital, even, to be honest, some private equity funds or growth funds interested in some of these areas, but now all of a sudden everyone thinks it’s cool. The infrastructure seems to be having a renaissance, a full-on rebirth. In this episode, we will explore in which cool ways the infrastructure spaces are moving and what’s leading to it. We will deep dive into the forces that are leading us to this. We will deep dive into semiconductors, networking and connectivity, energy, manufacturing, and then we’ll wrap up. Bertrand, so infrastructure is cool now. Bertrand Schmitt We’re back to building things Yes. I thought software was going to eat the world. I cannot believe it was then, maybe even 15 years ago, from Andreessen, that quote about software eating the world. I guess it’s an eternal balance. Sometimes you go ahead of yourself, you build a lot of software stack, and at some point, you need the hardware to run this software stack, and there is only so much the bits can do in a world of atoms. Nuno Gonçalves Pedro Obviously, we’ve gone through some of this before. I think what we’re going through right now is AI is eating the world, and because AI is eating the world, it’s driving a lot of this infrastructure building that we need. We don’t have enough energy to be consumed by all these big data centers and hyperscalers. We need to be innovative around network as well because of the consumption in terms of network bandwidth that is linked to that consumption as well. In some ways, it’s not software eating the world, AI is eating the world. Because AI is eating the world, we need to rethink everything around infrastructure and infrastructure becoming cool again. Bertrand Schmitt There is something deeper in this. It’s that the past 10, even 15 years were all about SaaS before AI. SaaS, interestingly enough, was very energy-efficient. When I say SaaS, I mean cloud computing at large. What I mean by energy-efficient is that actually cloud computing help make energy use more efficient because instead of companies having their own separate data centers in many locations, sometimes poorly run from an industrial perspective, replace their own privately run data center with data center run by the super scalers, the hyperscalers of the world. These data centers were run much better in terms of how you manage the coolings, the energy efficiency, the rack density, all of this stuff. Actually, the cloud revolution didn’t increase the use of electricity. The cloud revolution was actually a replacement from your private data center to the hyperscaler data center, which was energy efficient. That’s why we didn’t, even if we are always talking about that growth of cloud computing, we were never feeling the pinch in term of electricity. As you say, we say it all changed because with AI, it was not a simple “Replacement” of locally run infrastructure to a hyperscaler run infrastructure. It was truly adding on top of an existing infrastructure, a new computing infrastructure in a way out of nowhere. Not just any computing infrastructure, an energy infrastructure that was really, really voracious in term of energy use. Nuno Gonçalves Pedro There was one other effect. Obviously, we’ve discussed before, we are in a bubble. We won’t go too much into that today. But the previous big bubble in tech, which is in the late ’90s, there was a lot of infrastructure built. We thought the internet was going to take over back then. It didn’t take over immediately, but there was a lot of network connectivity, bandwidth built back in the day. Companies imploded because of that as well, or had to restructure and go in their chapter 11. A lot of the big telco companies had their own issues back then, etc., but a lot of infrastructure was built back then for this advent of the internet, which would then take a long time to come. In some ways, to your point, there was a lot of latent supply that was built that was around that for a while wasn’t used, but then it was. Now it’s been used, and now we need new stuff. That’s why I feel now we’re having the new moment of infrastructure, new moment of moving forward, aligned a little bit with what you just said around cloud computing and the advent of SaaS, but also around the fact that we had a lot of buildup back in the late ’90s, early ’90s, which we’re now still reaping the benefits on in today’s world. Bertrand Schmitt Yeah, that’s actually a great point because what was built in the late ’90s, there was a lot of fibre that was built. Laying out the fibre either across countries, inside countries. This fibre, interestingly enough, you could just change the computing on both sides of the fibre, the routing, the modems, and upgrade the capacity of the fibre. But the fibre was the same in between. The big investment, CapEx investment, was really lying down that fibre, but then you could really upgrade easily. Even if both ends of the fibre were either using very old infrastructure from the ’90s or were actually dark and not being put to use, step by step, it was being put to use, equipment was replaced, and step by step, you could keep using more and more of this fibre. It was a very interesting development, as you say, because it could be expanded over the years, where if we talk about GPUs, use for AI, GPUs, the interesting part is actually it’s totally the opposite. After a few years, it’s useless. Some like Google, will argue that they can depreciate over 5, 6 years, even some GPUs. But at the end of the day, the difference in perf and energy efficiency of the GPUs means that if you are energy constrained, you just want to replace the old one even as young as three-year-old. You have to look at Nvidia increasing spec, generation after generation. It’s pretty insane. It’s usually at least 3X year over year in term of performance. Nuno Gonçalves Pedro At this moment in time, it’s very clear that it’s happening. Why now: the 5 forces behind the renaissance Maybe let’s deep dive into why it’s happening now. What are the key forces around this? We’ve identified, I think, five forces that are particularly vital that lead to the world we’re in right now. One we’ve already talked about, which is AI, the demand shock and everything that’s happened because of AI. Data centers drive power demand, drive grid upgrades, drive innovative ways of getting energy, drive chips, drive networking, drive cooling, drive manufacturing, drive all the things that we’re going to talk in just a bit. One second element that we could probably highlight in terms of the forces that are behind this is obviously where we are in terms of cost curves around technology. Obviously, a lot of things are becoming much cheaper. The simulation of physical behaviours has become a lot more cheap, which in itself, this becomes almost a vicious cycle in of itself, then drives the adoption of more and more AI and stuff. But anyway, the simulation is becoming more and more accessible, so you can do a lot of simulation with digital twins and other things off the real world before you go into the real world. Robotics itself is becoming, obviously, cheaper. Hardware, a lot of the hardware is becoming cheaper. Computer has become cheaper as well. Obviously, there’s a lot of cost curves that have aligned that, and that’s maybe the second force that I would highlight. Obviously, funds are catching up. We’ll leave that a little bit to the end. We’ll do a wrap-up and talk a little bit about the implications to investors. But there’s a lot of capital out there, some capital related to industrial policy, other capital related to private initiative, private equity, growth funds, even venture capital, to be honest, and a few other elements on that. That would be a third force that I would highlight. Bertrand Schmitt Yes. Interestingly enough, in terms of capital use, and we’ll talk more about this, but some firms, if we are talking about energy investment, it was very difficult to invest if you are not investing in green energy. Now I think more and more firms and banks are willing to invest or support different type of energy infrastructure, not just, “Green energy.” That’s an interesting development because at some point it became near impossible to invest more in gas development, in oil development in the US or in most Western countries. At least in the US, this is dramatically changing the framework. Nuno Gonçalves Pedro Maybe to add the two last forces that I think we see behind the renaissance of what’s happening in infrastructure. They go hand in hand. One is the geopolitics of the world right now. Obviously, the world was global flat, and now it’s becoming increasingly siloed, so people are playing it to their own interests. There’s a lot of replication of infrastructure as well because people want to be autonomous, and they want to drive their own ability to serve end consumers, businesses, etc., in terms of data centers and everything else. That ability has led to things like, for example, chips shortage. The fact that there are semiconductors, there are shortages across the board, like memory shortages, where everything is packed up until 2027 of 2028. A lot of the memory that was being produced is already spoken for, which is shocking. There’s obviously generation of supply chain fragilities, obviously, some of it because of policies, for example, in the US with tariffs, etc, security of energy, etc. Then the last force directly linked to the geopolitics is the opposite of it, which is the policy as an accelerant, so to speak, as something that is accelerating development, where because of those silos, individual countries, as part their industrial policy, then want to put capital behind their local ecosystems, their local companies, so that their local companies and their local systems are for sure the winners, or at least, at the very least, serve their own local markets. I think that’s true of a lot of the things we’re seeing, for example, in the US with the Chips Act, for semiconductors, with IGA, IRA, and other elements of what we’ve seen in terms of practices, policies that have been implemented even in Europe, China, and other parts of the world. Bertrand Schmitt Talking about chips shortages, it’s pretty insane what has been happening with memory. Just the past few weeks, I have seen a close to 3X increase in price in memory prices in a matter of weeks. Apparently, it started with a huge order from OpenAI. Apparently, they have tried to corner the memory market. Interestingly enough, it has flat-footed the entire industry, and that includes Google, that includes Microsoft. There are rumours of their teams now having moved to South Korea, so they are closer to the action in terms of memory factories and memory decision-making. There are rumours of execs who got fired because they didn’t prepare for this type of eventuality or didn’t lock in some of the supply chain because that memory was initially for AI, but obviously, it impacts everything because factories making memories, you have to plan years in advance to build memories. You cannot open new lines of manufacturing like this. All factories that are going to open, we know when they are going to open because they’ve been built up for years. There is no extra capacity suddenly. At the very best, you can change a bit your line of production from one type of memory to another type. But that’s probably about it. Nuno Gonçalves Pedro Just to be clear, all these transformations we’re seeing isn’t to say just hardware is back, right? It’s not just hardware. There’s physicality. The buildings are coming back, right? It’s full stack. Software is here. That’s why everything is happening. Policy is here. Finance is here. It’s a little bit like the name of the movie, right? Everything everywhere all at once. Everything’s happening. It was in some ways driven by the upper stacks, by the app layers, by the platform layers. But now we need new infrastructure. We need more infrastructure. We need it very, very quickly. We need it today. We’re already lacking in it. Semiconductors: compute is the new oil Maybe that’s a good segue into the first piece of the whole infrastructure thing that’s driving now the most valuable company in the world, NVIDIA, which is semiconductors. Semiconductors are driving compute. Semis are the foundation of infrastructure as a compute. Everyone needs it for every thing, for every activity, not just for compute, but even for sensors, for actuators, everything else. That’s the beginning of it all. Semiconductor is one of the key pieces around the infrastructure stack that’s being built at scale at this moment in time. Bertrand Schmitt Yes. What’s interesting is that if we look at the market gap of Semis versus software as a service, cloud companies, there has been a widening gap the past year. I forgot the exact numbers, but we were talking about plus 20, 25% for Semis in term of market gap and minus 5, minus 10 for SaaS companies. That’s another trend that’s happening. Why is this happening? One, because semiconductors are core to the AI build-up, you cannot go around without them. But two, it’s also raising a lot of questions about the durability of the SaaS, a software-as-a-service business model. Because if suddenly we have better AI, and that’s all everyone is talking about to justify the investment in AI, that it keeps getting better, and it keeps improving, and it’s going to replace your engineers, your software engineers. Then maybe all of this moat that software companies built up over the years or decades, sometimes, might unravel under the pressure of newly coded, newly built, cheaper alternatives built from the ground up with AI support. It’s not just that, yes, semiconductors are doing great. It’s also as a result of that AI underlying trend that software is doing worse right now. Nuno Gonçalves Pedro At the end of the day, this foundational piece of infrastructure, semiconductor, is obviously getting manifest to many things, fabrication, manufacturing, packaging, materials, equipment. Everything’s being driven, ASML, etc. T…
Full show notes at the publisher 
Our TOPPODCAST Picks 
Stay Connected
