Industrial AI: The New Wave of Innovation Hitting Heavy Industry

Our last article introduced industrial software and why Energize has been investing in it for nearly a decade. This article dives into the industrial value chain-and how AI is changing the path from concept to product.

‍The energy transition relies deeply on physical industry. From solar panels to electric vehicles to batteries, every product that is deployed (and every product that goes on to power the decarbonized economy) must move through an industrial lifecycle, progressing from design to manufacturing and, in many cases, automation.

Because physical industry is so central to the energy transition, industrial software is a crucial piece of Energize's investment strategy. In the past five years, we've seen immense demand growth for energy and manufacturing infrastructure-including the physical backbone for renewables, grid systems, and batteries-and software will be core to meeting that demand efficiently and accurately.

The newest class of industrial software, industrial AI, embeds intelligent inference and automation directly into industrial workflows, and it is poised to dramatically impact manufacturing. In this Deep Dive, we outline where we think the most significant change is happening in the space, and what AI tools we see leading the next chapter of innovation.

‍
What's driving the growth of industrial AI?

In recent years, several macroeconomic forces have bolstered demand for software and AI-enabled solutions across the industrial value chain:

‍Labor constraints. In August 2025, roughly 409,000 manufacturing positions went unfilled in the U.S. The gap is projected to widen: By 2033, industry growth, coupled with a shrinking pool of workers, is expected to create 3.8 million new manufacturing jobs. About half of those positions are likely to go unfilled. This is largely the result of an aging workforce, rising technical skill requirements, and training pipelines that have not kept pace. Manufacturers who can't hire enough people are looking at software to avoid falling behind.

Investment in new capacity. A decade of policy-including, most recently, executive actions on supply chain resilience-has directed significant capital into domestic manufacturing. Spending on U.S. manufacturing construction is up 139% since 2021. Facilities, both new and old, face growing pressure to boost productivity without adding headcount.

‍Rising cost of downtime. American factories produce, on average, about 75% of what their installed capacity allows. The cost of this under-utilization, which is caused largely by facility downtime, is high and growing quickly: Across sectors, the price of one hour of unplanned downtime increased 2-4x between 2019 and 2025. In automotive manufacturing, a single hour of downtime on a large production line can cost $2.3 million. As a result, factories see a clear economic case for maintenance and monitoring through software.

‍Better connectivity. Today, 79% of factory assets are connected in some form through IoT sensors, edge computing, and the cloud. Meanwhile, the cost of that connectivity has dropped sharply; IoT sensor prices fell 71% between 2004 and 2020. And "smarter" factories are now becoming data-rich environments where software and AI can operate more easily. The bottleneck has shifted from data collection to data usability.

‍Validation from early adopters. Some manufacturers have already successfully deployed advanced industrial technologies. These factories, called "lighthouses," have boosted AI adoption rates on key use cases from 14% in 2018 to 55% in 2023, and they report 2-3x productivity gains and a 99% reduction in defects. For other manufacturers, this success is concrete evidence that these tools are worth investing in.

Amid these trends, the heads of operations at many industrial facilities are looking with renewed interest at their software budgets, and seeking both strategic investment and faster adoption of new tools.

Emerging opportunities in industrial AI

At a high level, bringing any physical product to market requires key steps in design, manufacturing, and often automation, each of which relies on the labor and expertise of a diverse array of technical experts. Mechanical, electrical, and systems engineers drive design, working through component specs, simulations, and documentation. Plant engineers oversee operations and line performance on the factory floor. Automation engineers deploy and integrate machines.

For these engineers, industrial AI will have far-reaching impacts on day-to-day work, from the earliest stages of design to the management and eventually AI-native automation of production on the factory floor. Here's where AI is creating the most significant change in design, operations, and automation, and the companies we're watching:

Product design and simulation

In manufacturing, hardware design has historically been a slow process involving multiple teams of experts - including industrial designers, product managers, mechanical engineers, manufacturing and supplier teams, and regulatory and compliance staff - and multiple stages, from requirements documentation to mass production.

That rigor exists for good reason, as design errors are expensive to fix once manufacturing begins. But it's also costly. Engineering hours are expensive, and engineers tend to spend a lot of their time on repetitive, bureaucratic tasks-documenting, iterating, and validating-rather than on design itself. For example, in interviews with the Energize team, engineers estimated that they spend an average of 70% of their time on administrative and other tasks and just 30% on design during the product validation and testing phase.

To address this inefficiency, software that works alongside established design tools can shorten workflows and automate repetitive tasks, like documentation. Today, the CAD and simulation tools that engineers use, like Autodesk and SolidWorks, are deeply embedded and costly to switch. As such, solutions that integrate with existing workflows - rather than displace them - may be on a quicker track to market penetration.

In PCB design, for example, iteration cycles tend to be long, slowed down by validation steps, design changes, and prototyping. AI that helps automate those loops and manages design changes can compress the timeline considerably.

AI is similarly poised to disrupt documentation. In our conversations, mechanical engineers identify documentation, like workflow records and technical instructions, as a primary pain point. AI-enabled software that generates documentation automatically, and outputs it in the right formats, saves engineers time. But output must be accurate: Mission-critical designs cannot afford to have errors.

‍Companies to watch: Diode Inc., Quilter, PhysicsX, Foundation, DraftAid, Flow Engineering.

Data-driven production

After a design is planned, validated, and tested, it moves into production, scaling from small prototype batches to hundreds or thousands of units a week. The engineering challenge shifts from getting the design right to making things work on the factory floor.

In manufacturing today, factories typically generate large quantities of unstructured data from diverse sources, like sensors, cameras, maintenance records, and inventory logs. This data, when high-quality and properly structured, can identify bottlenecks and optimize equipment uptime.

‍

‍

But few industrial organizations use this data to its full potential. This underutilization stems from an array of challenges common to industrial facilities. A plant's data flows through many complex layers, and each layer typically stores data in its own format. Data generated at the machine level, in proprietary protocols, is difficult to extract insights from. Even within the same company, data architectures differ enough across sites that cross-facility analysis is difficult to do.

To solve this problem, software solutions pull data from disconnected systems, structure and contextualize this data in a common format, and connect it to the right teams and workflows. Software companies are tackling these problems in diverse ways, from processing time-series data to creating unique data lakes across facilities.

AI, meanwhile, is changing both the speed and capability of this work. Where other software connects and structures industrial data, AI can contextualize and parse true insights from the "noise" of this data. And AI agents, with humans in the loop, can monitor operations, flag anomalies, and surface the right insights to the right person at the right time. AI is also giving rise to a new services industry, where firms use AI to build and operate customIndustrial solutions.

In all these applications, software companies that own the underlying data infrastructure are especially well-positioned: Once embedded, solutions are difficult to displace.

Companies to watch: Nominal, SIFT, ethon.ai, Nixtla, Manex AI, Nexxa AI, Juna AI.

Automation

As manufacturing matures and scales, facilities tend to invest more in automation, which bolsters efficiency and output while reducing labor shortage pressures. While previous waves of innovation focused on reducing hardware costs, that pain point is now changing; the cost of industrial hardware has fallen roughly 80% since 1995. Instead, per our analysis of large-scale automation projects, 65% of costs now go to the "hidden costs" of software and services-that is, programming, commissioning, and systems integration. The specialists who perform this work are in short supply, and managing automation across multiple production lines is a challenge.
‍

‍

This represents an opening for software. Consider the task of programming a new machine for the factory floor, which traditionally required specialists to write control code, or PLC logic-a bottleneck for deployment. Software that generates that code automatically, and lets manufacturers test how machines will behave before they're physically installed (virtual commissioning), can remove that bottleneck.

Once systems are live, software can continuously manage deployed codebases across installations, serving as the GitHub for industrial automation. It provides version control, update management, and auditability across every deployed unit.

As robot deployments scale, software has the opportunity to step in as the orchestration layer, monitoring fleet performance, managing code across installations, and giving operators the visibility they need to maintain and improve systems in the field.

Companies to watch: Xaba, Augmentus, ThoughtForge (deployment); Copia, Foxglove, Rerun.io (orchestration); Field AI, Physical Intelligence, Generalist (robotics data and foundation models).

The big picture

Energize has been investing in industrial software since 2018, and we've watched firsthand as industrial software has evolved from spreadsheets, to the cloud, to machine learning, and now AI. Throughout these waves, our belief has stayed consistent: Investing in a decarbonized economy means investing in the industrial economy, too. The physical infrastructure of the energy transition must be manufactured by facilities that can keep up with the pace of demand.

This work has ripple effects. Better data reduces downtime and informs the next design cycle. A product that's better designed is easier to manufacture. Better automation software makes the whole system more productive. Across all these phases, software is the common denominator, and cumulative improvement across the industrial value chain has far-reaching consequences for the industrial economy and a decarbonized future.

Now, we're excited to partner with the next generation of companies delivering industrial AI to critical industries.
Are you building or investing in this space? We'd love to hear from you, or see you in person at Energize Connect: Industrial AI on June 22 in Chicago.

‍