Onshoring manufacturing with robots and AI

Onshoring (also called reshoring) means bringing manufacturing back to the country where products are designed and bought, in this case the United States, after decades of offshoring. This page serves the futurism and abundance and post-scarcity goals of this wiki, with a secondary look at problems in living such as jobs, skills, and communities. Its central claim is narrow. Robots and artificial intelligence can shrink the labor cost that pushed factories abroad, but they do not carry back the knowledge of how to make things. That knowledge, and the web of nearby suppliers that holds it, is the real bottleneck, and automation changes which kinds of knowledge matter most.

Why labor cost is only half the story

US manufacturing employment peaked at 19.6 million in June 1979. In September 2026 it stood at about 12.65 million.

In 2009, Harvard Business School professors Gary Pisano and Willy Shih argued that outsourcing erodes an "industrial commons": the collective research, engineering, and manufacturing capabilities shared by a company and the suppliers of advanced materials, tools, production equipment, and components around it. They stressed that such commons are geographically rooted. In Producing Prosperity (2012) they argued that once manufacturing capabilities go away, much of the ability to innovate and compete goes with them.

The consequence for automation is direct. A humanoid robot or a robotic cell can replace a wage. It cannot replace a missing supplier of display glass, a process recipe built from years of yield learning, or a technician who knows why a line drifts out of tolerance.

Three verified 2026 examples

Tesla Optimus. On October 21, 2026, Tesla said it had started producing its Optimus humanoid robot in Fremont, California, in space once used for vehicles. Its shareholder letter called the units "the first production-quality bots that we have manufactured" and said Tesla would keep iterating on the hardware and the manufacturing process. The robot is a product of domestic manufacturing before it is a tool for it, so its own actuators, sensors, and chips are part of the onshoring question.

Terafab. The project sometimes heard as "TerraFab" is Terafab, a chip manufacturing effort by Tesla and SpaceX. In August 2026 the companies named Grimes County, Texas, as the site, with an initial phase of about $16.8 billion and a research fab at Giga Texas in Austin. Electrek noted that SpaceX's IPO filing called Terafab a "general framework" with no binding commitments. In October 2026 Musk targeted volume production for the third quarter of 2028 with "big error bars." Reuters reported that Intel has joined the project and that acquiring equipment from ASML, Applied Materials, and KLA is a central challenge. The know-how arrives through a partner and the tools through equipment makers.

Anduril BlueChip. On October 27, 2026, Anduril and Qualcomm announced BlueChip, starting with a laptop "designed and assembled in the United States," with circuit board assembly planned at Anduril's Arsenal-1 facility in Ohio and orders from 2027. The release describes designing the enclosure for US machining, the battery system for existing American battery technology, and a bonding process for US-made display glass. Anduril told The Verge the prototype was about 32 percent US-origin by component cost, then committed to exceeding the 65 percent Buy American threshold at launch. The processors are to be made at TSMC in Arizona, and the company is programming FPGAs to stand in for power management chips it normally cannot source outside China.

BlueChip is the clearest public illustration of the commons problem. Almost none of the reported engineering concerns assembly labor. It concerns redesigning a product around the suppliers that still exist.

A know-how stack for onshoring

One useful way to read any onshoring announcement is as a stack of five layers.

1. Repetitive labor. The layer robots and AI replace most directly.
2. Process know-how. Recipes, tolerances, and yield learning, much of it tacit knowledge carried by people.
3. Supplier depth. Components and materials within trucking distance, such as glass, batteries, and power chips.
4. Tools and equipment. Machine tools, lithography systems, and the robots themselves.
5. Integration and maintenance. The people who install, program, and repair automated lines.

Automation compresses layer 1 and enlarges the importance of layers 4 and 5. That creates a twist. In 2024, Chinese manufacturers supplied 57 percent of the robots installed in China, up from 47 percent a year earlier. Onshoring that depends on imported robots moves the dependency up the stack rather than removing it. A practical test for any claim is three questions: what share of procurement cost is domestic, which layer is still imported, and who on site can fix the line when it stops.

What the data shows so far

The Reshoring Initiative counted about 247,000 US manufacturing jobs announced through reshoring and foreign direct investment in 2025, roughly matching 2024, and projected about 309,000 for 2026. Respondents named tariffs, geopolitical risk, and proximity to customers as top reasons.

Kearney's 2026 Reshoring Index, which compares imports from 14 Asian low-cost countries with domestic output, stayed negative at minus 91 (from minus 115). US imports of manufactured goods rose 4.6 percent, manufacturing investment tripled over four years while capacity grew only 1.5 percent, and laptops and smartphones still cost less to make in typical source countries even after tariffs. The gap between announcements and output fits the commons thesis: money arrives faster than capability.

The most cited capability gap is in chips. In July 2023 TSMC pushed production at its first Arizona fab to 2025, citing a shortage of specialist workers, and said it was sending experienced technicians from Taiwan to train local workers. The know-how traveled in people, which is how industrial commons usually move. The fab was among the most prominent projects of the CHIPS Act era.

On robots, the International Federation of Robotics reports 2024 densities of 1,220 robots per 10,000 manufacturing employees in South Korea, 307 in the United States (8th worldwide), and 166 in China (22nd, after revised labor data). Density hides scale, though. China installed about 295,000 robots in 2024, 54 percent of the world total, and holds an operational stock above 2 million. The United States installed 34,164, about 6 percent. Robots are therefore not a uniquely American advantage. The country with the deepest commons is also automating fastest.

Implications for abundance and problems in living

For abundance, production near demand shortens shipping chains and removes single points of failure. If automated domestic lines reach scale and good yields, the time price of goods can fall. Early onshored goods usually cost more, so the payoff comes after the learning curve, not at the ribbon cutting. Distributed tools such as 3D printing can complement large plants for spare parts and low-volume items, a step toward the conditions described in Post-scarcity economy.

For problems in living, the returning work will not resemble 1979. Expect fewer line positions per unit of output and more technicians, robot integrators, process engineers, and toolmakers. Terafab's first phase is expected to employ at least 3,000 people, and the site plan already drew local concern about water use. Communities benefit only if local training pipelines exist before the jobs arrive.

What learners, teachers, and researchers can do

Learners can target skills in layers 2, 4, and 5 of the stack: PLC and robot programming, machine vision, metrology and statistical process control, CNC machining and tooling design, design for manufacturing, and semiconductor equipment maintenance.

Teachers can build case units around BlueChip: give students a laptop bill of materials and ask them to map which parts have a domestic supplier, then redesign around the gaps. Simulation and scored challenges can make this kind of training more engaging, in the spirit of Making productive work as engaging as games.

Researchers can ask:

• How quickly does tacit process knowledge transfer when experienced technicians train a new workforce, and what shortens that time?
• Does heavy robot adoption increase or decrease the number of distinct supplier firms in a region?
• Which products have tightly linked design and process, and therefore depend most on a local commons?
• Can "commons depth" be measured, for example by domestic content by cost plus supplier count within a day's drive?

Discussion questions, essay ideas, and learning related AI prompt ideas

• If humanoid robots made assembly labor nearly free, which remaining layer of the know-how stack would most limit a return of laptop manufacturing to the United States, and why?
• Compare what the Reshoring Initiative job announcements and the Kearney Reshoring Index each measure, and propose a combined indicator that would show real onshoring.
• Should public money for onshoring go mainly to factories, or to the technicians, toolmakers, and component suppliers that let factories run?
• How does reliance on robots built or installed mostly in China change the resilience argument for robot-led onshoring?
• What would a town near a new chip fab need in its schools within five years to fill the skilled jobs locally?
• At what point would automated local production lower the time price of everyday goods, and what evidence would show that it has happened?
• Design a learning path from beginner to maintaining a robotic work cell, and identify which steps simulation or an AI tutor could teach well.

See also

• Robotics
• Manufacturing
• Manufacturing engineering
• Artificial intelligence
• International business
• Post-scarcity economy
• Time prices and measuring abundance
• 3D printing
• Making productive work as engaging as games