Schneider Electric’s VC Arm: The AI Buildout Is Creating A New Industrial Investment Cycle

This is an ongoing series on investors focused on rebuilding the physical layer. Previous interviews in the series were with ex-Meta CTO Mike Schroepfer, founder of Gigascale Capital , and Peter Barrett, a decade-long investor at Playground Global . Schneider Electric has spent nearly two centuries adapting to successive industrial revolutions — evolving from a 19th century steel and heavy machinery company into a global leader in energy management and automation. Now, through its $1 billion venture arm, SE Ventures , the company is betting that the next transformation will be driven by AI’s collision with the physical world, from data centers and power grids to robotics and industrial automation. Amit Chaturvedy, SE Ventures global head and managing partner. (Courtesy photo) For Amit Chaturvedy , who joined SE Ventures in 2022 after leading corporate investments at Cisco , AI’s biggest opportunities extend well beyond software. As demand for compute strains energy infrastructure and accelerates reindustrialization, the firm is backing startups building the technologies that underpin the AI economy — investing in everything from data center infrastructure and grid resilience to robotics and industrial AI. Crunchbase News spoke with Chaturvedy about where those opportunities are emerging, why energy has become AI’s defining constraint, and how industrial technology is being reshaped by the AI era. “We were set up with the intent to figure out where the market is headed,” he said. The significance of the energy and industrial sectors has grown with AI, and that has led even traditionally tech-focused venture investors to rush into the space. “Today, the scarce resource in this entire space is the capacity to build — building, real estate, energy, power and electrification gear,” Chaturvedy said. SE Ventures counts eight unicorns in its portfolio , and has notched 12 exits including its most recent, Fabric8Labs , a 3D metal printing technology acquired by Tokyo-based electronic manufacturer TDK Corp. The firm will often take board seats or board positions and work to bring value to its portfolio companies. Around 80% of the startups in its portfolio have some level of commercial relationship with a business unit of Schneider Electric. Most often that’s as a partner servicing Schneider’s customers, which is the holy grail, according to Chaturvedy. Sometimes it’s as a vendor, although that remains a smaller set of use cases. In our conversation, we spoke about power scarcity, the electrical grid, workforce training, reindustrialization and notable portfolio companies. The interview has been edited for length and clarity. Gené Teare: Which sectors or investments are you focused on? Where there is a lot of drive or interest because of what is happening in AI? Chaturvedy: Three things come to mind, especially in terms of the areas we invest in versus the broader construct of the market. First, AI is getting embedded, and you need to train models, whether open source or proprietary. Model training has upleveled to inference so you need AI infrastructure. Together AI is a great example of that. Five years out, when this CapEx cycle starts to come down and new data centers are perhaps not getting created, data center efficiency will become a hot topic. We are also investors today in a company called Hammerhead AI , which focuses on that problem. That will come three, five or seven years out. It is going to come. It is not a problem today because we are on the upswing of the CapEx cycle. Together AI and Hammerhead AI are very interested in partnering with Schneider Electric, because Schneider Electric is a leading electrification player in the data center space. It makes a lot of gear and equipment that go into these data centers. Today, the scarce resource in this entire space is capacity to build: buildings, real estate, energy, power and electrification gear. The other market impacted by the emergence and growth of AI is the interplay with the grid. There are more demands on the grid beyond the electrification of vehicles, and it is 100x or 1,000x bigger than what we saw with vehicles needing to get charged from inside houses. The grid could not keep up with that capacity in the past, and it certainly cannot keep up with these demands today. More project developers are coming in and setting up renewables or other types of capacity, but again, the interplay is still with the grid. Anything that helps with grid resilience is clearly an area for us to invest in. The third thing is the transformative impact of AI on the world of industrials. That is where we are quite excited. Robotics is one clear area where a general-purpose model can allow the same robotics hardware to do multiple different tasks that were not possible in the past, because cognition and inference were not possible at the edge before the advent of large language models. Companies like Skild AI in our portfolio — which is one of the most exciting companies at the intersection of robotics and AI — are market-leading indicators of where this world is headed. There is also an element of using AI to deliver better use cases in the field. Companies like Axion in our portfolio essentially capture warranty data, analyze it and feed results back to design engineers in big corporations. There are a lot of OEMs and hardware companies looking for select use cases where AI can actually be very transformative. That is what customers are looking for: How can AI be transformative for my business? Whichever startup is working with me in that transformation journey is the startup that will move from POC to adoption overnight. That is essentially the world of successful startups. Overlaying on top of this is a confluence that we see and watch from our vantage point. When you think about the energy efficiency that needs to happen in these industrial worlds, energy technologies and industrial technologies have to collaborate and deliver those use cases while being energy efficient. That was not the case in the past. Energy was cheaper and more readily available. Now industrial is taking off. There is more AI adoption. The workforce is getting older, and there is no way to overnight train a workforce in America, so you have to rely on AI. You are going to consume more and more AI for industrial use cases, which was never a business imperative in the past. This is where the worlds of enterprise and industrial are colliding very quickly in the world of AI. Increasingly, what I hear is that the bottleneck for AI at this point is energy. Are you seeing some short-term solutions that help with this? What about longer-term technologies? Chaturvedy: It’s very clear that from a short-term basis — and this isn’t quite an energy-related solution — it’s more about tokens. If you think about the unit economics of an AI data center, it’s the tokens. To generate a token, it costs electricity. To train your model, or infer from a model, you need a lot of tokens. The bigger the model, the bigger the data set, and the more complex the use cases, the more tokens. Ultimately, it’s a battle of producing tokens cheaply and also consuming fewer tokens through the models that exist today. That’s where optimization is happening, but that’s more in the enterprise space: How can I write clever versions of software that allow me to do essentially that? The longer-term solution is going to be about — actually, maybe there is a middle layer also — beyond the tokens: When I’m running my data center, can I push inference to a different point in time so I’m not consuming peak electricity rates? Can I manage my HVAC better? You need cooling systems to cool your data center environment, and there are techniques that work really well there. Schneider has also bought some assets in the past. Then the longer-horizon cycle is really about creating new generation capacity, la