The current U.S. administration has justified its aggressive tariff policy as a way to spur American innovation, drive manufacturing jobs back to American soil, and increase wages. In reality, most manufacturing that left was manual assembly, the floor for which is $20-$30 per hour in the U.S. as set by employers like Amazon at its fulfillment warehouses (if not minimum wage in some areas). Given that developing countries charge a mere tenth of that, it’s simply not realistic to compete. It’s become increasingly clear that the only cost-effective way to bring manufacturing back to U.S. soil is to automate unskilled labor via robotics while human operators, technicians, and managers oversee complex production systems.

This is not labor replacement but a state where humans direct and optimize robotic automation. While the path to building hundreds of automation lines in the U.S. is not simple, the trade conflict puts manufacturing investment at an inflection point. After clearing certain technological and manufacturing hurdles, we can seize the opportunity to onshore by utilizing today’s class of AI-enabled robotic solutions.

Flexible robots

The first hurdle is probably the biggest but presents a tremendous opportunity—automating manufacturing flows for goods with complex materials and processes. A great example is clothing: Fabric is hard to manipulate due to its lack of rigidity, seams have to be stitched, assembly is often intricate, and sizes vary widely. It’s a process that requires significant skill and is highly variable, and the smallest defects result in unsellable waste.

These complex scenarios are exactly where AI will have the biggest impact on robotic capabilities and augment human labor. Imagine robots with the dexterity of human fingers and the ability to move rapidly from one task to another. Such capabilities are still in development and not immediately available for implementation; however, capabilities like “few-shot learning”—where robots can replicate a task after seeing it only a few times, like a human—are already being demonstrated in labs across the country. It’s only a matter of time until this ability makes its way into production.

Training challenges

The second is labor availability and job training. Manufacturing and production hubs across the U.S. have withered and died off, resulting in a shortage of skilled workers able to pick up where the U.S. left off before globalization. For companies to hire, they need a talent pipeline similar to the one driving semiconductor companies like Intel and TSMC to expand their footprint in Arizona, where there’s established, high-volume semiconductor manufacturing and Arizona State University graduating ample engineers. One option is creating labor training programs where the manual production of each type of good comes with complex training requirements.

The alternative is strategically deploying AI robots to make humans more productive and create higher leverage for a limited labor pool. In a world of automated manufacturing, skilled labor only needs to learn how to operate, maintain, and fix the robotic tools doing the production. This is still challenging, but far less so than training millions of laborers on tens of thousands of specialized manual tasks.

Supply chains: Where to begin

The third is the global supply chain. Even though a Ford F-150 or Honda Accord may be made in the USA, a good chunk of the bill of materials (BOM) is not. These components could be as simple as electrical fuses or as complex as entire transmission systems. Replicating such a supply chain for every major good—from clothing to automobiles to semiconductors—will take decades and hundreds of billions of dollars of investment, and even then, there are certain components or tools that may never be fully built in the U.S.

The rational way to start the incremental onshoring process is by using automation and robotics to manufacture the highest value and highest tech components first because they would have the greatest tariff impact on the final sales price of finished goods. After going down the list, manufacturers may realize that the final ~10% of a BOM is not even worth replicating in the U.S. and will simply have to be an embedded cost. Alternatively, in the very long term (20+ years), if manufacturers wish to fully onshore the supply chain, robotics may offer a path to replicating even the production of the lowest-cost commodity components.

AI robotics investing

Finally, manufacturing poses a significant return-on-investment issue. It’s hard, complicated, capital intensive, and very low margin. Why would rational investors choose this profile over investing in technology companies with sticky 75%+ gross margins and relatively capital-efficient scaling? After all, the marginal cost of “shipping” software is nearly $0 (excluding customer acquisition costs) while the marginal cost of shipping a commodity widget is typically 80% to 90% of its sales price. There’s a reason Elon Musk described Tesla’s troubles as “production hell.”

However, there are robotics business plans and financial models being developed that show software-like margins at a steady state. In addition, unique business models such as robots as a service (RaaS) offer predictable, recurring revenue streams and strong payback periods for the upfront capital expenditure. This assumes robotics reach economies of scale, achieve broad suites of capabilities, and are priced more like labor replacement than cost-plus commodity widgets. Regardless, this creates an attractive investment profile—first for early-stage VCs and eventually, for generalist investors.

Tariffs as a near-term tailwind

While there is tremendous opportunity to increase productivity with AI robotics, we need to be realistic and understand that not everything can, will, or should be onshored. But, as we saw during the COVID pandemic, short-term supply chain imbalances in areas such as shipping and logistics can make such innovation more palatable than the steady state.

As we consider the future of manufacturing in the U.S., it’s critical to prioritize solutions designed not just for early gains but also for enduring impact and adaptability. I believe (as do other early-stage investors) that today’s robotics companies offer a promising path to returns, regardless of tariffs. Tariffs may be a near-term tailwind, but it’s still highly uncertain what impacts they will have long term. Only time will tell, but one thing is sure: Robotics were already seen as an attractive investment, and tariffs are likely to make them even more so.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Read more:

Nearly 4 million new manufacturing jobs are coming to America as boomers retire—but it’s the one trade job Gen Z doesn’t want
Americans want more U.S. factory jobs—as long as they don’t have to work them
I connect U.S. manufacturers with their workforce. A career in manufacturing provides purpose, connection, and work-life balance
How to make manufacturing cool again

The only way some manufacturing jobs will return to America is if AI-empowered robots do them—and that’s a good thing