You must have seen Cade's LinkedIn post
The original Industrial Revolution changed the world by giving humans machines that could produce, move, and build at a scale that was previously impossible. Steam power, electrification, factories, and mechanization transformed not just individual products, but the underlying systems of production.
The thesis
AI may represent the next version of that transformation. This time, the breakthrough is not simply giving machines more power, but giving physical systems intelligence: the ability to perceive, decide, coordinate, optimize, and increasingly act on their own. That is the idea behind IndustrialRevolution.ai. The most important impact of AI may not be confined to software at all; it may be the rebuilding of factories, energy systems, logistics networks, mines, ships, and other critical infrastructure around intelligent machines.
For the last two decades, some of the most valuable technology companies were built by placing software on top of existing industries.
The model was familiar: find a large, inefficient industry, build software that makes it more productive, and sell that software back to the companies already operating within it.
That model is not going away.
But a new one is emerging.
The next generation of software companies may not simply sell technology to industrial companies. They may become the industrial companies themselves.
These companies look very different from the software businesses that defined the last cycle, but the underlying advantage is often the same: software, data, AI, and automation allow them to operate physical systems in a fundamentally different way.
The difference is that they are choosing to capture that advantage directly.
The shift
Historically, software companies sat above the physical world.
They sold tools to manufacturers, utilities, logistics companies, miners, and construction firms. The industrial company owned the assets. The software company collected a subscription.
That made sense when software improved an operation incrementally.
If a piece of software could make a factory 10% more efficient, selling that software to thousands of factories was an attractive model.
But AI, autonomy, robotics, and modern software systems can create a much larger operating advantage.
If your software allows you to run a factory dramatically faster, reduce labor requirements, improve throughput, make better capital allocation decisions, or automate processes that historically required large teams, the economics begin to change.
At some point, selling the software may become less attractive than owning the operation.
If you can operate a billion-dollar physical asset materially better than the incumbent, why settle for a small percentage of that value as software revenue?
That question sits at the center of this new industrial model.
Case — Defense
Anduril is one of the clearest examples.
The company began by building autonomous defense systems and the software infrastructure required to coordinate them. Over time, that software advantage pulled the company deeper into manufacturing.
The company did not stop at building autonomy software for existing defense contractors. It built drones, autonomous systems, factories, and now major industrial infrastructure.
Its Arsenal facilities are designed around the idea that defense manufacturing itself can be rebuilt using modern software, automation, and production systems.
The significance is larger than defense.
Anduril is showing what happens when a software-native company decides the physical manufacturing system is not simply a customer to serve, but something to redesign from the ground up.
Case — Manufacturing
Hadrian takes the thesis even further.
The company has built a software-driven manufacturing platform designed around automation, precision, speed, and repeatability.
But it does not simply sell factory software.
It builds and operates the factories.
That distinction matters.
The old model would have been to develop factory automation software and license it to manufacturers.
Hadrian's model effectively says: if our software gives us a meaningful manufacturing advantage, the best way to capture that value may be to manufacture the parts ourselves.
Software becomes the operating system of the factory rather than the product sold to the factory.
That is a profound shift in business model.
Case — Mining
Mining offers an even more extreme example.
KoBold Metals applies machine learning, large geological datasets, and advanced modeling to mineral exploration.
Its core insight is largely an information problem: the world may contain the resources we need, but identifying them is extraordinarily difficult.
A traditional software company might package those models into an exploration platform and sell it to mining companies.
KoBold instead acquires exploration rights, identifies deposits, and participates directly in developing the resources.
The software creates the edge.
The mine captures the economics.
The company therefore looks less like a traditional SaaS business and more like a technology-native mining company.
Case — Energy
Energy is another category where this model becomes powerful.
Fervo is using modern drilling techniques, sensors, software, data, and increasingly AI to improve geothermal development.
But Fervo is not selling geothermal optimization software to utilities.
It builds power plants.
The company's technology advantage is embedded directly into the process of finding, drilling, and operating geothermal assets.
This matters because many industrial industries have historically been treated as one-off engineering projects.
Factories, power plants, mines, and infrastructure projects were custom-built, slow, and difficult to replicate.
Software changes that.
Software allows physical systems to become more standardized, measurable, programmable, and repeatable.
Once that happens, industrial operations can begin to scale more like technology businesses.
Case — Shipbuilding
Shipbuilding may be one of the most interesting categories of all.
Traditional shipyards are enormous, capital-intensive, slow-moving operations with long development cycles.
Saronic is approaching the problem differently.
The company is building autonomous vessels, but it is also building the industrial infrastructure required to manufacture them at scale.
Its Port Alpha shipyard is being designed around software, automation, modular production, and data-driven manufacturing.
The key idea is not simply to make a traditional shipyard more efficient.
It is to ask what a shipyard would look like if it were designed from the beginning by a software-native company.
That question can be applied to almost every heavy industry.
The answer may be very different from the incumbents we know today.
Convergence
Several technologies are converging at once.
Together, these technologies change the economics of operating physical businesses.
The software is no longer merely helping a human operator make a slightly better decision.
Increasingly, the software can coordinate the operation itself.
That makes vertical integration far more attractive.
The flywheel
For much of the software era, vertical integration was considered unattractive.
Capital-light businesses were preferred.
Factories required too much money. Hardware created supply-chain risk. Industrial projects took too long to build. Investors preferred recurring software revenue with high gross margins.
Those concerns are still real.
But the advantages of vertical integration may be growing faster.
And perhaps most importantly, it can capture the full economic value created by its technology.
This can create a powerful flywheel:
The physical asset becomes both the product and the data engine.
For investors
This may force investors to rethink how technology companies are identified.
Some of the most important software companies of the next decade may not report 90% gross margins.
On the surface, these businesses may resemble industrial companies.
Underneath, their operating leverage may come from software.
The software margin is effectively embedded inside the asset.
That can make these companies harder to recognize using traditional software investing frameworks.
The question is no longer simply:
"Is this a software company?"
A better question may be:
"Does software create a structural operating advantage in this industry?"
If the answer is yes, the company may deserve to own much more of the value chain.
The point
Marc Andreessen famously argued that software was eating the world.
That idea defined an era.
Software entered industry after industry and digitized existing workflows.
But the next phase may look different.
Software is no longer just eating industries from the outside.
It is beginning to rebuild them from the inside.
The first generation of enterprise software companies sold tools to industrial companies.
The next generation may use software, AI, robotics, and automation to become the best industrial operators themselves.
Factories, mines, ships, power plants, and infrastructure may increasingly be built by companies that think more like software startups than traditional industrial conglomerates.
The most consequential technology companies of the next decade may therefore look surprisingly physical.
They may own fewer dashboards than we expect.
And far more factories, robots, mines, ships, and power plants.
Stay current