MANUFACTURING

Reindustrialization and the realities of manufacturing

By: Del Costy, President and Managing Director, Siemens Digital Industries

The resurgence of American manufacturing encompasses far more than what happens on the shop floor.

Reindustrialization requires organizations to master the entire innovation lifecycle from product conception through engineering, validation, production, and optimization. The United States has long led the way in excelled at innovation. The challenge now is harmonizing that innovation with modern manufacturing capabilities so we can continue to innovate with speed.

When we think about in the industrial segment, we still are in the early stage of the curve of adopting technology. There are huge gains to be had—and people are going to play an important role. And, of course, you can't have any conversation without asking: What does it mean to adopt AI in manufacturing?

That mentality has shaped how I think about what is happening in manufacturing today. Everywhere I go, people are talking about AI. They are eager to talk about what AI can do, yet fewer are talking about what it is that AI needs to work effectively, and that is context.

Manufacturing requires a clear understanding of complex systems, processes, materials, and constraints. This is where AI in manufacturing can be incredibly powerful. It can elucidate hordes of data, finding patterns in the insights to make recommendations for product, system, process and plant improvement. The caveat is this only works if AI is grounded in the reality that manufacturing is a precision environment. Without context, intelligence becomes little more than prediction—and prediction alone is not enough to transform industry.

Why Industrial AI Is only as good as the data behind it

By now, everyone has heard the promise: “AI will optimize production lines, accelerate product development, improve quality, and predict failures before they happen.” Those opportunities are real. But AI can only scale in manufacturing when data is contextualized and connected to the physics of engineering and production.

You can have the most sophisticated model in the world, but if it lacks an understanding of your tolerances, materials, production constraints, validation history, and operational requirements, the results will be unreliable at best and detrimental to your business at worst. Manufacturers cannot afford the risk of implementing overly confident AI They need AI that can make accurate and trustworthy recommendations every single time

That is where many Industrial AI conversations break down. Too often, AI discussions fail to reflect the constraints, priorities and conditions of the shop floor. Organizations make arguments that sound convincing but are disconnected from the physical world. And in manufacturing, confidence without certainty is dangerous.

What often goes unsaid is that many AI deployments are built on someone else's model, running on someone else's assumptions about how a business operates. In effect, manufacturers are renting intelligence that they should own.

At the same time, product development remains burdened by decades-old processes. Bringing a new product to market can take between 40 to 50 months. On top of that, up to 60 percent of development costs are often consumed by physical prototyping. Too much of the R&D budget is spent proving designs rather than innovating them.

If American manufacturers want to compete and win, we must compress the entire industrial cycle from concept to production. Industrial AI can help achieve that goal, but only when it is built on a foundation of data that accurately reflects how products are designed, manufactured, and operated.

How to build the digital foundation for reliable Industrial AI

Last year, Siemens invested $15 billion in U.S. software companies to further ensure customers thrive in the next era of digital and data-driven transformation, and we’ve spent decades building the related data foundation.

Our approach starts with product lifecycle management systems that capture and preserve every design revision, engineering change, and validation result within a customer's own environment. That institutional knowledge remains where it belongs: with the manufacturer.

We combine that with physics-based simulation technologies that validate AI recommendations against real engineering models rather than relying solely on statistical outputs. When AI suggestions can be tested against the laws of physics, manufacturers gain confidence that they can act on the results.

The third piece is automation. With Siemens technology connecting design and production, data does not stop at the engineering stage. It extends into operations, creating real-time visibility into how products, lines, and plants are performing in real-time. Together, these capabilities create something no stand-alone AI platform can replicate: contextualized data connected across the entire lifecycle.

We call this the digital thread. It connects design intent to production reality and preserves organizational knowledge as products and systems evolve. It provides AI with the context it needs to generate meaningful outcomes.

And that distinction matters. Reliable Industrial AI cannot be bolted on as an afterthought. Manufacturers cannot outsource their domain intelligence and expect better results. They need a connected foundation that links products, processes, and production so Industrial AI can operate across the enterprise rather than inside isolated silos.

The value of that approach becomes clear when we look at how companies are putting it into practice.

How we need to execute on reindustrialization

The United States has set an ambitious goal to reindustrialize its economy.

American industry has already made tremendous progress through automation, software, and digital twin technology. These innovations have helped manufacturers become more productive, more resilient, and more adaptable.

But our customers are entering a new phase. They're trying to build new capacity, modernize existing operations, and remain globally competitive—all while facing persistent shortages of skilled labor.

Industrial AI builds on that strong digital foundation, helping customers expand what their people are capable of. It gives engineers, operators, and technicians new tools to become more productive, accelerate innovation, and accomplish more.

The challenge isn't inventing AI anymore. It's deploying AI at enterprise scale. Today, only a small percentage of organizations have successfully integrated AI across their operations. Closing that deployment gap is one of the defining industrial challenges—and opportunities—of the next decade.

That's what Siemens is focused on: helping our customers move from AI ambition to AI deployment, and translating the promise of AI into real productivity gains across American industry.

How PepsiCo uses digital twins and AI to accelerate manufacturing performance

With one of our customers, PepsiCo, we have built a replicable framework that combines Digital Twin technology, Industrial AI, and a connected IT/OT architecture to help scale improvements across manufacturing, warehousing, and logistics operations.

The results have been significant. At one U.S. Gatorade facility, PepsiCo achieved a 20-percent increase in throughput within just three months. The company also estimates a 10- to 15-percent reduction in capital expenditures by using virtual validation to uncover hidden capacity before making physical investments.

The digital twin has helped compress facility design and optimization timelines from months to days. By creating physics-based models using real operational data from multiple facilities, teams can establish performance baselines, identify bottlenecks, understand material flow, and pinpoint where and why energy is being consumed.

More than 90 percent of production issues can be identified before production even begins.

This is the power of a digital-first approach that PepsiCo harnesses. Engineers optimize layouts before equipment is moved. Material flow is validated before a conveyor is touched. Energy-intensive processes are identified before they create operational challenges. Decisions are driven by data rather than assumptions.

When product innovation, process development, and production execution are connected, organizations gain the speed and confidence needed to modernize legacy infrastructure. That is exactly how manufacturers win the race against time.

While reindustrialization is certainly about improving existing operations, another key component is building entirely new models for production.

How Haddy demonstrates the future of distributed manufacturing

Another example of a company leveraging Siemens integrated software, hardware and automation portfolio is Haddy, an AI-powered robotic 3D-printing company, that is fundamentally changing the future of how we manufacture.

Haddy's MicroFactory model demonstrates that reality, combining industrial production capability with highly efficient economics to produce everything from amusement park decor to boats for the Navy. This company, a recent winner of a Siemens Techcellence award, bridges commercial and defense requirements, helping connect enterprise innovation with national industrial readiness.

What makes the model Haddy uses compelling is how AI continuously improves manufacturing performance through a closed feedback loop:

Step 1—The part talks: A camera detects a tiny surface indentation nearly impossible for a human notice.

Step2—The system listens: The system recognizes a pressure spike four minutes before the defect appeared. AI connects those signals and identifies a relationship.

Step 3—It traces the root cause: Analysis reveals a material degradation occurred upstream in the drying process before printing began.

Step 4—It fixes itself: The dryer temperature is automatically adjusted. Production continues without interruption.

Step 5—It verifies: The next ten layers print successfully. The issue is logged and recorded as resolved.

Step 6—Every machine gets smarter: The solution is shared across every Haddy printer globally, making it permanent and repeatable.

This loop is not possible through automation only. It is a combined hardware, software and automation stack that facilitate institutional learning at industrial scale.

Reindustrialization requires connecting innovation to production

The good news about efforts to reindustrialize across American industries is that manufacturers do not need to reinvent engineering to achieve breakthrough results. By combining proven engineering frameworks with simulation and Industrial AI, extraordinary gains in efficiency become possible.

The organizations that lead the next decade of manufacturing will be the ones that built the strongest foundation for industrial AI. They will be the companies that retained their domain expertise, connected design and production, and ensured every AI-driven decision could be validated against physics. Turning probabilistic AI recommendations into deterministic outcomes.

That is the lesson we see across companies like PepsiCo, Haddy, and many others who have realized that this is not only a competitive advantage, but the path to accelerating US manufacturing.

American cities and businesses that are already committing to reindustrialization did not wait for a signal to start to rebuild their future. American manufacturing cannot afford to wait either.

Build the foundation, own the intelligence, and partner with organizations that understand how to connect innovation to production and together we can accelerate the next era of American reindustrialization.

Published: August 17, 2026