2025 US Reshoring Reassessment Part 2

AI Automation and Financing the Ecosystem

TL; DR

• The key to neutralizing cost handicaps associated with reshoring is to improve US manufacturing productivity through products designed for and manufactured with automation. Although many manufacturers have previously attempted to automate their manufacturing processes, AI-enabled robotics may revolutionize the industry altogether.

• The most likely avenue for AI automation is for OEMs to deliver solutions to their manufacturing partners. Nvidia and Foxconn are an example of this collaboration. The other option is for OEMs to integrate with their own robot and manufacturing capability vertically. Amazon provides an example of this type of integration. Apple may be forced to do the same with its products.

• Reshoring the supporting component ecosystem beyond advanced semiconductors is more challenging. Our rough estimate suggests that it would cost approximately $170 billion in upfront capital expenditures to reshore enough capacity for $200 billion in consumer device consumption in the US. The US government can't fund this transition without creating debilitating inefficiencies. The current US administration favors a tariff policy that places the burden on OEMs and their suppliers to determine the most efficient path forward.

• A viable solution could rely on well-developed and sophisticated financing options. Specialized financing would incentivize US-based component makers to remain competitive and push the boundaries of the leading edge. It is also a clear opportunity for the finance industry as the US erects barriers to global competition with no sign of this trend abating in the foreseeable future.

Introduction

This note is the second part of a two-part series on the changes affecting reshoring initiatives in the US. The first part of this research is here.

As Cathal Nolan describes in The Allure of Battle, most wars are acts of prolonged stalemate and attrition. The victors are determined by which side can outlast the other (as seen in the Russia-Ukraine conflict, for example). If there were ever a US-China conflict over Taiwan, the US would likely be unable to outlast China's production advantage, which relies on low-cost labor and a network of supply chain partners supported by government subsidies.

The cost handicap and the lack of a supporting ecosystem are the main barriers to reshoring electronics assembly. But reshoring doesn't have to mean replicating the current products and manufacturing infrastructure originally designed for Asia.

To win both in the market and on the battlefield, the U.S. must boost its manufacturing productivity, offsetting higher labor costs, stricter environmental rules, and tighter workplace regulations. To achieve this, advanced consumer electronics must be designed for automated manufacturing.

Today, most smartphones and consumer devices are not designed for automated final assembly. That is why Foxconn has thousands of workers at the end of their iPhone SMT lines. The parts are tiny and complex, and the cable routing is tricky. Yet the end result fits neatly into a sleek, attractive enclosure. If an iPhone were made in the U.S., it would still need to look good—but it would also need to be designed for automated assembly. The cost structure would have to tilt away from labor and toward depreciation.

If it were as simple as just replacing humans with robots, manufacturing would have never left high-cost regions. Many past efforts to automate completely have failed. General Motors' push in the 1980s is a well-known example. A more recent one is Tesla's struggles in 2018. The primary issue is that traditional automation is most effective on stable products with long life cycles. If you're still tweaking your design, or if your product changes every year, it's hard to justify the cost of automation. Robots are adept at performing repetitive tasks, but they struggle to adapt to new situations. Automation is hard and human adaptability is underrated.

Commercializing Physical AI for Electronics Assembly

AI-enabled robotics is about to change all of that. In July 2023, Google released RT-2, a vision-language AI model trained on web data—similar to large language models—but designed to control robots. RT-2 showed that general-purpose AI could power flexible robotic systems. Since then, more than 20 startups have entered the field with new AI tools for automation. By mid-2024, Nvidia CEO Jensen Huang was evangelizing the concept of "physical AI" to accelerate this shift.

The potential payoff is compelling. Several studies [here, and here, for example] suggest that AI-enabled robotic automation could enter the market at a fully burdened labor rate of approximately $10 to $15 per hour. If you assume very generous high-end configurations, the rate could double to $20-$30 per hour. Even if you consider the higher cost estimate, it is significantly lower than current fully burdened labor rates in the US (~$45/hr), assuming they can perform and adapt as well as thinking human beings.

There are certainly technology development issues (see here for a good review), But there are also market barriers, especially for electronics manufacturing. Most AI robotics startups would prefer to enter broad, "horizontal" markets, where their AI software runs on top of existing automation hardware. For a large part of the automation industry, that means working with independent integrators who have deep expertise in specific vertical sectors. But industrial automation is broad. It is comprised of a complex web of different industry structures. Some vertical sectors don't have specialist integrators. In these sectors, the challenge for physical AI companies (aside from the obvious need for the technology to work flawlessly) is that well-established players already own the channel. And they're not letting in organizations that could take a slice of their pie. The business model is strikingly similar to the same failed attempts in the development of the driverless car industry circa 2018. Automotive OEMs were never keen on just letting automation software stacks and sensors cannibalize their already thin margins.

The market entry problems are particularly challenging for electronics manufacturing, which can be considered the "Galapagos Islands" of the automation industry due to its isolation and specialization. The market is dominated by equipment companies focused on surface-mount technology (SMT) and die attach. Contract manufacturers (EMS and ODM firms) design and run their own lines. The channel for incorporating AI in electronics assembly is through these partners, with limited options for new entrants to siphon profits. It will be a great opportunity for the EMS/ODM and the assembly equipment value chain if they can corral the AI capabilities to create a new ecosystem.

One avenue for AI electronic assembly automation is for the OEMs to deliver automation solutions to their manufacturing partners. The first example of this is Nvidia, which is reportedly working with Foxconn to build an AI factory in Houston, Texas. The other option is to vertically integrate so that the AI model maker, physical robotic automation supplier, and manufacturing process owner are under one organization. Amazon has vertically integrated commercial service robotic capabilities. It wouldn't be surprising to see Amazon spin this business off into a separate automation division in the future, much like Amazon Web Services evolved.

Apple may be forced to do the same for its iPhone products. We would be surprised if there weren't already a design in the works within Apple engineering for an iPhone that is American-made and designed for automated manufacturing. Whether that manufacturing line will use fully humanoid AI-enabled robots or more traditional automation enhanced with AI is a function of timing. And that is anyone's guess if the trigger is a geopolitical event that forces the issue.

Financing the Ecosystem

The other major challenge for reshoring is that the component ecosystem supporting electronic assembly is mainly based in China. The US will never be self-sufficient in Tech hardware unless it can reshore these component sectors as well.

The geographic shift to China was gradual. Market pressures prompted companies to establish operations there, taking advantage of cheap labor and generous subsidies. The shift to Southeast Asia and India is more rapid, brought on by the first round of tariffs and de-risking strategies. The 2025 trade war is expected to reshape the global economic landscape further. The US is spiraling towards a policy where component makers will need to be in the US for the US. The change is likely to be more abrupt. Component makers will have to reshore some portion of that production capacity and effectively strand some assets (utilization) in Asia.

It will be costly. During the lead-up to the CHIPS Act, several studies [here, here, and here] attempted to estimate the amount China has spent to support its tech hardware industry. While the data isn’t perfect, RCD Advisors estimates that national and local Chinese subsidies totaled about $250 billion over the past 20 years across the supply chain.

RCD Advisors estimated the costs of reshoring consumer electronic devices for US self-sufficiency below. The methodology is simple enough. We estimate that the US consumes approximately $200Bn worth of consumer electronic devices (smartphones, wearables, notebooks, tablets, etc.). We then calculated the approximate capex-to-revenue value ratio for each industry sector and adjusted for the US's higher regulatory costs and increased automation requirements. We also accounted for utilization and industry inefficiencies. The analysis is conservative because it neglects the value of intermediate assembly steps (RF modules, camera modules, etc.). It also doesn't include layers further back in the supply chain, such as critical materials, which may have very high capex-to-revenue ratios.

Nevertheless, it does provide a reasonable ballpark estimate of the level of investment required to rebuild state-of-the-art electronics manufacturing in the US. A roughly ~$170Bn investment would be needed to produce $200Bn worth of device revenue for US consumption and ensure the DoD has access to the latest technologies. Roughly 70% of that investment would be in specialty semiconductors for sensors, RF, and power.

How would this massive reshoring effort get funded?  It is pretty clear that the US cannot copy the playbook of developing nations. The current US administration favors a tariff policy that places the burden on OEMs and their suppliers to determine the most efficient path forward.

Passives, interconnects, and even a large portion of semiconductor suppliers don't have the balance sheets to allocate all the capital for building regional manufacturing bases. It is too inefficient for a 21st-century electronic component supplier. Compounding the issue is that over the last twenty years, many of these organizations have made significant bets on overseas manufacturing. It wasn't only investments in China. Only a few years ago, companies were compelled to de-risk their investments by establishing alternative manufacturing locations in Southeast Asia. Many of these suppliers took advantage of government subsidies to offset some of the investment costs when they built their facilities.

OEMs (Apple, Nvidia, HP, Dell), who will be forced to reshore, could fund some of the transition costs for their supply chain. But that will inevitably pull investment away from other opportunities (like AI development) that have higher returns. There is no magic. There are opportunity costs associated with reshoring in the Tech hardware ecosystem. OEMs will certainly use whatever bargaining leverage available to force some suppliers to fund the reshoring costs. It is hard to see a scenario where the consumer doesn't end up paying for this in some way.

The US could sweeten the deal with additional subsidies, although that seems unlikely under the current administration. This consultancy has been critical of some of these subsidies in the past, and the reality is that even though the CHIPS Act was successful at onshoring some capability (TSMC in Arizona), there have been several "less than stellar" outcomes so far (Wolfspeed, Intel, etc.). Subsidies seem to work best when there is a foreign supplier with existing production capability that can "copy exactly" onto US shores. It is less effective as a boost for suppliers riding up the learning curve.

There are also limits to the efficacy of subsidies for long-term security interests. In developed economies, subsidies and tariffs foster uncompetitive markets. They allow organizations to become complacent and inefficient, preventing them from generating animal spirits to establish a competitive advantage. In a competitive rivalry with both commercial market and national security implications, the US can't afford to have a complacent Tech hardware component supply chain.

A viable US-centric solution could rely on well-developed, sophisticated financing options, such as private credit. Specialized financing would incentivize component makers to remain competitive and push the boundaries of the leading edge. There are already some examples of these arrangements in the tech hardware industry. Intel partnered with private equity firms to help fund the expansion of their semiconductor fabrication facilities. The Financial Times recently reported that Meta is in advanced talks with a group of leading private credit investors to raise $29 Bn in financing for its expansive US-based AI data centers.

These partnerships involve selling minority stakes in manufacturing facilities to private firms, allowing manufacturers to retain majority ownership and operational control while freeing up capital for other investments. The details can be complicated, including setting up special-purpose vehicles and using synthetic leases. Some of these financing options have lost appeal over the last few years, as accounting rules changed in 2018, requiring companies to capitalize all leases. Nevertheless, they remain viable options for accessing capital, and more importantly, they put enough competitive pressure on organizations to drive efficiency and innovation. It is also a clear opportunity for the finance industry as the US continues to erects trade barriers. Although it is tricky because of the possibility of perverse incentives, the US government could provide a modest backstop that shares some risks and lowers interest terms.

Conclusion

AI and drone warfare is forcing the US to shift the reshoring goal away from minimum viable capacity and closer towards self-sufficiency. The good news is that an AI robotic revolution is about to begin that could level the playing field. The bad news is that it will be expensive. A lot of capital will be needed to rebuild regional production.

If your organization is involved in electronic manufacturing, it's essential to begin thinking and planning now. If there is one thing the ongoing tariff saga has taught us all, it is that geopolitics is unpredictable. But the tea leaves are all pointing in one direction. It is best to craft the "Break Glass in Case of Emergency" plan now. That plan could include creating well-defined contingency plans for regional expansion or establishing a healthy inventory buffer. It is a risk assessment and judgment call that all Tech hardware decision makers will have to make. We can help.

If you find these posts insightful, subscribe above to receive them by email. If you would like to learn more about our consulting practice and how we assist organizations in the Tech hardware supply chain, please get in touch with us at info@rcdadvisors.com.