Beyond the Internet: The AI Industrial Revolution

The Collapse of Legacy Workflows and the Dawn of AI-First Factories

The internet had a huge economic impact: new industries were created and existing industries quickly embraced it. Whether it was for communicating, idea sharing, or client acquisition, it altered the way businesses operate. Today, all leading industrial companies use the internet in some way, but was it a game-changer for everyone? In this article, we argue that the internet had an incremental impact on the industrial sector, whereas AI will force a wholesale restructuring of it.

Internet and Manufacturing

At its heart, the internet is a communications platform. The internet changed how companies communicate, exchange information, and by extension, how supply chains and distribution networks are managed. The impact of the internet on industrial companies has been very limited—cars connect to phones, but how cars are used and manufactured did not change much. The internet’s impact was limited to offering new features, while often core products remained the same. The introduction of new features created value but was not enough to cause massive disruption.

Manufacturing and AI

We believe AI will impact everything. Hardware companies are going to be fundamentally disrupted as AI changes how we design, build, and operate hardware. Smart CAD will accelerate product development cycles; AI-run, robot-operated factories will reduce manufacturing times; and large systems will be autonomously operated by AI agents. The internet did not penetrate the physical production world, but AI will.

Lessons of the Past

Today's manufacturing leaders are software-first companies. They designed their hardware around software, offering seamless integration. As software became an important component of every product, we saw legacy manufacturing companies generally struggle with software development and deployment. Perhaps most surprisingly, industrial manufacturing titans of the tech industry also suffered the same fate. Today's leading compute company is software-first. NVIDIA sells most of the GPUs deployed for AI. One of NVIDIA's most potent weapons is CUDA (Compute Unified Device Architecture), a programming module for NVIDIA's GPUs. The internet never required manufacturing titans to become software-first until the AI era. With CUDA as standard software, legacy manufacturers do not have an adequate product to challenge NVIDIA.

If we believe that AI will have a huge impact on how we design and build hardware, why shouldn't history repeat itself?

It should, which is why we believe legacy companies will integrate AI into their workflows, but it will be too little, too late.

Incumbent Disadvantage

The big disadvantage incumbents have is, ironically, their well-functioning business units. If those organizational workflows work, why change them? Even when it is obvious that radical changes need to be made, making changes for large corporations is difficult and, in most cases, impossible. It takes a founder mindset to make significant changes.

Car companies are an example—their core competency is the development of internal combustion engines (ICE). ICE departments have the most political and financial capital. While it is highly likely that the future is electric, legacy OEMs significantly lag Tesla in EV market share. Even if executive leadership of legacy OEMs saw the potential of EVs, taking on the ICE department would be a heavy lift. This dynamic explains why large systems are prone to technical disruptions, even when those disruptions are obvious.

Source: Marlinspike research

AI-first companies are not just about integrating AI into workflows; they are about completely changing workflows and product life cycles. Product design, manufacturing, and product operations will all be data-driven. Insights from the physical world (operations layer) will be quickly integrated into manufacturing and product design software.

However, it is important to stress that only by starting with a clean sheet (team, product design process, and factory layout) can the full benefit be realized. A clean-sheet approach will ensure that each step of a product's lifecycle is optimized for AI—for example, AI CAD tools will optimize designs to take maximum advantage of AI-operated robots. Those processes will be continuously improved as more data is generated by manufacturing and deployment of products. As workflow feedback loops generate data, a flywheel will be in effect: AI systems will improve, resulting in legacy systems failing to keep up.

As we have seen in the example of Tesla, using modern tools and clean-sheet designs can lead to approximately a 3x increase in productivity, perhaps even more given the latest advances in AI.

The implications for the industrial economy are massive.

AI-Powered Industrials

The opportunity here is massive. Industrials in the S&P 500 account for approximately $3.4 trillion in market capitalization and generate $1.1 trillion in annual revenues. As new AI-first industrials emerge, whole markets will be disrupted, and ultimately current S&P industrials will be replaced by newcomers.