$230 billion buys the factories. It cannot buy the workers.

Making 10 categories of consumer technology entirely in the United States could require $230 billion in factory investment alone, according to an industry study of the cost of forced reshoring. ^[2]

The Consumer Technology Association, a trade group representing technology companies, and consultancy Kearney released the estimate on October 6, 2026. Their model examines fully domestic production by 2031, from components to finished smartphones, laptops and eight other types of devices. The estimated capital bill runs from $185 billion to $230 billion, before the higher manufacturing costs flow through to buyers. ^[1][2]

The report arrives as President Donald Trump demands that 50 percent of semiconductor manufacturing be brought to the US by the end of his term. That demand is separate from the study's scenario of entirely US-made consumer devices. The Genesis Mission, his manufacturing and science initiative launched on November 24, 2025, does not require those products to be made wholly in America. No White House response to the CTA report was found in the available sources. ^[1][3][6]

The economic argument is straightforward: domestic manufacturing capacity can improve supply security, but compulsory self-sufficiency would consume resources that technology companies also need to expand. For AI, the collision would be particularly sharp. The industry needs imported chips and more electricity while Washington wants more chip production at home. A policy that makes both inputs harder to obtain would favour builders with the deepest pockets.

The CTA has an interest in making that warning loud. Its members include Apple and Google, and small businesses account for 80 percent of its membership. This is an industry lobbying against costs it expects to bear, using engineering estimates rather than observed US production costs. That limits the precision of its forecast. It does not make the physical demands disappear. ^[1][2]

What the $230 billion buys

The study covers computer monitors, laptops, robotic vacuums, smart speakers, smartphones, smartwatches, televisions, video game consoles, wireless earbuds and wireless headphones. It excludes AI infrastructure, enterprise hardware, networking equipment and automotive electronics. The upper estimate is therefore a bill for a defined slice of technology manufacturing, not a price for making the entire technology industry American. ^[1][2]

Capital expenditure pays for the productive assets needed to make those goods. It is separate from the recurring expense of running factories and from the price a customer pays at the checkout. Treating $230 billion as the total cost would miss much of the study's warning.

Under full domestic production, Kearney estimates that smartphone manufacturing costs would rise by 152 percent. A phone that costs one unit to manufacture under the model's baseline would cost 2.52 units in the domestic scenario. Laptop manufacturing costs would rise by 93 percent, almost doubling. Smartwatches would rise by 97 percent, video game consoles by 58 percent and televisions by 41 percent. ^[2]

Those are modeled manufacturing increases, not announced retail prices. The study assumes companies pass between 25 and 50 percent of the increases to consumers. Under that assumption, smartphone retail prices rise by 38 to 76 percent, laptop prices by 23 to 46 percent and television prices by 10 to 21 percent. ^[2]

The assumption matters almost as much as the factory estimate. A company with enough pricing power might pass through more. One facing fierce competition might absorb more, protecting sales at the expense of its profit margin. Neither outcome is free. One makes the product less affordable, the other leaves the company with less money to invest.

The CTA said the products consumers consider most essential, led by smartphones and laptops, have the highest modeled costs under full domestic production. ^[1] That makes the policy economically awkward: the largest increases would fall on devices people use for work, study and communication, rather than only on optional gadgets.

There is also a recurring electricity bill. The study estimates annual demand of 19.1 billion to 19.5 billion kilowatt-hours, roughly San Francisco's annual consumption. That demand would compete with AI data centres for power. ^[1][2] Factory spending can buy equipment. It cannot, by itself, guarantee an electricity connection when the equipment is ready.

The workforce cannot be ordered into existence

Full domestic production would require between 555,000 and 668,000 additional full-time workers, according to the study. The upper figure is more than double the entire existing US computer and electronics manufacturing workforce. ^[1][2]

That is the report's hardest constraint. Hundreds of thousands of jobs may sound like an uncomplicated political victory, but a hiring requirement is not a hiring plan. The study does not establish where those workers would come from, what wages would recruit them or how quickly they could be trained.

On a press call, CTA executive board chair Gary Shapiro described Trump's semiconductor target as "physically impossible, labour-wise impossible" within the available time. He said that even with the money provided immediately, "you just can't do it." ^[1]

The scale of the semiconductor buildout gives the timing objection substance.

The Semiconductor Industry Association estimates that $630 billion has been committed to the US semiconductor supply chain across 140 projects in 28 states. US fabrication capacity is projected to grow 203 percent between 2024 and 2032. Even after that expansion, domestic supply is not expected to meet domestic demand, according to the Center for Strategic and International Studies, a policy research organisation. ^[4]

A projected tripling of capacity still leaves a shortfall.

The precise meaning of Trump's 50 percent target also needs an answer. The available evidence does not establish whether it is measured by manufacturing capacity, revenue or another yardstick. Those measures are not interchangeable. A target cannot be checked until the base the percentage is measured against is clear.

AI faces the same constraint at a larger scale

The CTA study does not price domestic production of AI accelerators or servers. Its relevance to AI lies in the shared demands of electronics manufacturing: specialised chip fabrication, components, packaging, skilled workers and power. Packaging is the process of encasing finished chips so they can be wired into devices. The consumer-device estimate shows how expensive rebuilding those capabilities can be, without providing a dollar forecast for AI hardware.

Only about 12 percent of global semiconductor fabrication capacity is in the United States. Meanwhile, US data centres are expected to deploy more than $1.4 trillion in semiconductors by 2030, most of them imported, according to CSIS. ^[4] The immediate AI expansion therefore depends on access to factories abroad.

TSMC, the Taiwanese chip manufacturer, illustrates both the progress and the lag. It has committed $265 billion to its Arizona expansion, with 12 fabrication and packaging facilities planned. Its most advanced two-nanometre production is only now ramping up in Taiwan and will not reach Arizona for years. The process name denotes a generation of chipmaking technology, with the newest generations central to the race for more capable hardware. ^[5]

The eventual US share of TSMC's global two-nanometre capacity is expected to be roughly 30 percent. The remaining 70 percent would stay in Taiwan. ^[5] An enormous US investment can still produce a supply chain that crosses borders.

CSIS estimates that a 100 percent tariff on all semiconductors would add $1.4 trillion to the US AI data centre buildout. That is a conditional tariff scenario, not a bill already imposed, and it cannot simply be added to the CTA's consumer-manufacturing estimate. The studies cover different purchases and different policy mechanisms. ^[4]

For AI developers, the likely competitive effect is clear. If hardware becomes more expensive, developers need more capital to buy computing capacity, whether directly or through a provider. The largest companies have more room to absorb that increase. Smaller builders would face a higher financial threshold before they could train or serve a model.

The available estimates do not quantify how many developers would be priced out. The conclusion is narrower: making the required computing capacity more expensive would strengthen the advantage of capital-rich incumbents. A policy intended to increase national technological strength could make the domestic AI market less open to newcomers.

A cheaper route still carries a bill

The strongest case for reshoring is supply security. More domestic capacity can reduce exposure to disruption abroad and preserve access to important manufacturing skills. The investment already committed to US semiconductor projects shows that expanding capacity is a serious undertaking, not merely a slogan. ^[4]

But that case does not require every component of every consumer device to be domestic. The CTA models a second option: final assembly in the US using largely imported components. Its estimated capital cost is $16 billion to $19 billion, with 61,000 to 73,000 workers required. ^[1][2]

At the upper endpoints, full domestic production would require about 12 times the capital investment of assembly alone. That comparison puts the expensive part of the ambition in view: recreating component production as well as bringing the finished device together. ^[2]

Assembly alone is no bargain for buyers. The model still puts smartphone manufacturing costs 67 percent higher and laptop costs 50 percent higher. Its projected retail increases are 17 to 33 percent for smartphones and 12 to 25 percent for laptops. ^[1][2]

The CTA recommends reducing or eliminating tariffs on components, particularly those supplied by trusted partners and allies, as the measure most likely to make domestic assembly cost-effective. ^[1][2] Its proposed bargain is more American assembly supported by continued foreign production of parts. Whether that satisfies the administration's security goals is unresolved.

The report's weakest points are its precision and its assumed response to higher costs. Kearney uses bottom-up engineering estimates of what products should cost to manufacture, rather than observed prices from a completed reshoring programme. Its consumer forecasts depend on the assumed pass-through rate. The models also apply tariffs in effect as of September 2026, so further increases would change the result. ^[2][3]

Those limits justify scrutiny of the dollar figures. They do not justify treating complete domestic production as a costless command. My reading is that targeted capacity expansion offers a more credible route to supply security than a requirement to reproduce the whole chain at once.

The next decisions are concrete: how the administration defines its semiconductor target, whether it accepts imported components for US assembly, and whether tariff policy makes those components cheaper or dearer. Until then, the CTA has supplied a price range for one scenario. Washington still needs to specify which scenario it intends to build.