In a fabrication shop outside Lexington, KY, MiddleGround Capital builds robotic work cells: self-contained stations where robots perform tasks with minimal human involvement.

The PE firm's automation team—roughly a dozen engineers and fabricators, many recruited from the Toyota plant up the road—assembles these cells near the firm's headquarters, then trucks them to the shop floors of its portfolio companies.

A typical build, founding partner Scot Duncan said, is "a robot arm on a little dolly." It opens the door of an industrial workstation, removes and measures a finished part, loads new material and starts the next cycle. "The human interaction is really almost nothing when it's actually running," Duncan said.

The setup is futuristic-looking, yet it is not generative AI, it is automation. That distinction is becoming increasingly important—and blurred—as PE firms race to put AI at the heart of an American manufacturing revival.

The private sector has made nearly $2 trillion in spending commitments to the US manufacturing industry since 2025, much of it spurred by investors' belief in AI's potential to boost productivity, according to trade publication IndustrialSage.

In March, the Wall Street Journal reported that Amazon founder Jeff Bezos held early discussions about raising as much as $100 billion for a vehicle to acquire manufacturers and accelerate their automation using AI. His startup, Prometheus, closed a $12 billion funding round in June at a roughly $41 billion valuation.

On Monday, the Institute for Supply Management's manufacturing index, which measures sector sentiment, hit 55.6, its highest reading since May 2022.

Yet when it comes to AI implementation, interviews with investors, manufacturers and investment consultants suggest that while AI is spreading across finance, sales, forecasting and engineering, productivity improvements on the factory floor still come from familiar sources, such as implementing robotics and lean manufacturing principles.

"Everyone's making very audacious claims about their use of AI and how it's going to—or already has—revolutionized their business," said Miles Arnone, chief executive and co-founder of Re:Build Manufacturing, a permanent-capital company that owns a group of US engineering and manufacturing businesses. "A lot of that is overblown at the moment."

MiddleGround built its robot shop to solve a problem caused not by technology, but labor. During the pandemic-era hiring crunch, of 40 applicants for a plant job, around 10 would qualify, and six of those would quit within days, Duncan said.

According to PitchBook estimates, automation makes US production competitive precisely because labor becomes a smaller share of cost—5% to 10% in highly automated facilities against 25% to 40% in older models, according to PitchBook's Q2 2026 Manufacturing Launch Report.

MiddleGround is experimenting with AI, including a system that matches purchasing requests with other portfolio companies. But Duncan described those as later phases—shop-floor automation came first.

"We think of AI use and automation as two different things," Duncan said.

At portfolio company Race Winning Brands, an automated forging machine is expected to generate roughly $9 million in equity value. A separate $100,000 wristpin-automation project at the company is estimated to create $1.27 million in value, the firm said.

Re:Build runs Toyota-style rapid-improvement projects across its factories, and a digital specialist now joins each one, using AI to code small tools on the spot.The gains so far, Arnone said, come "much more [from] the implementation of lean," and not yet the implementation of AI.

Re:Build's co-founder and co-chairman is Jeff Wilke, Jeff Bezos' longtime consumer chief at Amazon and once a leading candidate to succeed him. Bezos' family office participated in Re:Build's $120 million Series B funding round in April 2025, according to PitchBook data.

Manufacturing also presents specific challenges to the adoption of AI tools.

Operational data may belong to customers, fall under export controls or consist of sensitive designs that customers don't want exposed to a large language model. An error in an aircraft or nuclear component has consequences far beyond an inaccurate chatbot response, Re:Build's Arnone said.

Isolating an AI tool's contribution to productivity growth "is going to be a lot harder than people think," he added, using on-time delivery as an example. "There are probably 30 variables" that affect that. AI "is not going to increase profitability in engineering and manufacturing industries by 50% or anything like that," he said. It is "a wrench", a tool that works only inside a disciplined process.

Michael Psaros, co-founder and managing partner of industrial investor KPS Capital Partners, is bullish about where AI is headed.

The word "artificial intelligence" never entered the conversation with KPS investors or management teams until 12 to 18 months ago, he said.

Now, he calls AI "all upside" for manufacturers and said it "can't happen fast enough," particularly in demand forecasting, inventory planning and factory scheduling. He describes predictive demand as "the holy grail," capable of freeing up capacity without having to invest a dollar of new capital.

Though asked whether KPS would include a single dollar of AI-driven EBITDA improvement when underwriting an acquisition today, Psaros answered: "Hell no."

"You cannot today, in August 2026, put capital at risk behind a real or imagined efficiency associated with AI—at least that we would underwrite," he said. "We could have this discussion at Christmas. Things are changing pretty quickly."

Transom Capital Group, a Los Angeles PE firm that backs lower-middle-market industrial businesses, is rolling out AI across accounting, customer service, forecasting and software development. But co-founder and managing partner Russ Roenick said Transom has not used AI to guide factory-floor decisions. AI-enabled roll-ups of old-line industries have become a fashionable middle-market thesis, but one he has rarely seen executed.

"It's very popular and it kind of makes intuitive sense," Roenick said. "It's more of a concept than it is something that's actively happening."

Roenick expects AI to influence valuations before its operating impact is established. This may, in fact, be where the technology is doing its most powerful work.

"If you are trying to sell a business today in 2026 and the business is not prolific at utilizing AI in its day-to-day operation, you're going to sell for a discounted multiple to your industry peers that are," he said.