Reshoring American manufacturing will only work if it comes back more automated than it left. That was the central argument from Siemens AG President and CEO Dr. Roland Busch in a new episode of the Washington AI Network Podcast, recorded live at the CTA Innovation House.
Welcome remarks were delivered by Kinsey Fabrizio, President and CEO of the Consumer Technology Association, whose organization sits at the crossroads of the technology industry and the policy community and has been closely tracking the infrastructure demands of the AI era.
Busch spoke with BBC chief anchor Sumi Somaskanda about the rise of industrial AI and what it will take for the United States to lead the next phase of industrial automation.
Busch drew a sharp line between the general-purpose models consumers use every day and the kind of AI that can run a factory floor. Chatbots are non-deterministic by design — an asset for open-ended work, a liability where the same precise outcome is required every time. “If you want to deploy AI technology on the shop floor, it has to be deterministic,” Busch said, which means models trained on industrial data.
He pointed to Siemens’ Eigen Engineering Agent, which can program an industrial robot to weld a car door from a simple prompt — pulling design documents, writing and compiling the control code, and iterating until it works. “Normally you need a trained engineer,” Busch said. “Now you do that literally in minutes.” It is already in use by more than 100 customers.
Industrial AI, he added, is “the biggest lever to monetize all the investments which are currently spent on AI.”
Asked about rebuilding domestic capacity in pharmaceuticals, semiconductors, and defense, Busch was direct about the constraint: “You cannot re-industrialize this country with the same kind of labor intensity which you had in the past.”
The result, he argued, is different jobs rather than fewer — higher-value roles requiring investment in training and in the supplier ecosystem underpinning any real manufacturing base. The raw inputs, he said, are already here: domain expertise, semiconductor and pharmaceutical know-how, the AI technology itself, and people to deploy it. What’s missing is workforce training.
Siemens has put roughly $1 billion into U.S. manufacturing capacity in recent years, including a new production line in Fort Worth, Texas, and cited partnerships — among them work with Microsoft on data center infrastructure — as essential to assembling the full stack reshoring will require.
Looking ahead, Busch predicted a step change in industrial foundation models and a proliferation of autonomous robots on factory floors — though not the kind currently getting attention. “If you talk to our manufacturing people, they say, I don’t need a humanoid — take the legs away,” he said. On a flat factory floor, wheeled and task-specific machines are simply more efficient.
He also expects more inferencing to move to the edge, with models trained in data centers deployed directly onto shop-floor hardware.