WASHINGTON, 30 AUG 2026 — The US administration is preparing tariffs on imported semiconductors that would extend to finished hardware: data centre servers, laptops and gaming consoles. Reported as chip tariffs, the reach is what matters: they would tax assembled products, and most of that assembly happens in this region.

What is being prepared

The proposal covers an expanded range of technology products made alongside chips rather than chips alone. At its centre is a mechanism championed by Commerce Secretary Howard Lutnick that ties tariff exemptions to a company's investment in US chip fabrication: a firm could import a volume of chips duty-free, with the allowance set by how much production it pledges to bring onshore.

Rates, timing and terms are unsettled. A phase-in period is under consideration, and officials are weighing separate rates and import quotas by country, along with country-specific guidance for major chipmakers.

Industry objection has been immediate, on the argument that raising the cost of AI infrastructure hardware undercuts the same administration's stated aim of leading in AI.

Finished goodsServers, laptops and consoles, not only chips
Exemption for pledgesDuty-free allowance tied to committed US fab investment
By countrySeparate rates and quotas under consideration
UnsettledRates, phase-in and terms are all still undetermined

A tariff on finished goods is a different instrument

This is not a semantic distinction. A tariff on semiconductors taxes the country of fabrication. A tariff on servers and laptops taxes the country of assembly, and those are different countries.

Chips are fabricated in Taiwan, South Korea, the United States, Japan and increasingly China. Servers, notebooks and consoles are assembled in Vietnam, Malaysia, Thailand, Mexico and China. A measure aimed at chip supply that lands on finished goods therefore falls hardest on economies that do not fabricate anything and have built their electronics sectors on assembly.

That is Southeast Asia's position, and one several of its governments have been trying to move beyond. Vietnam has been pressing Qualcomm and Samsung for research investment rather than more assembly, and the tariff structure now under discussion would tax the activity it currently has while doing nothing for the activity it wants.

The exemption mechanism is the real policy

The duty-free allowance tied to pledged US fabrication investment is not a tariff feature. It is industrial policy delivered through customs.

It offers importers a discount proportional to how much manufacturing they move to the United States. That is a subsidy in the shape of a penalty, and because it requires no appropriation it is administratively lighter than a grant programme.

It also concentrates benefit among firms large enough to build fabs. A company with the balance sheet to pledge a US facility earns an allowance; a mid-sized importer with no such option pays the full rate. Whatever one thinks of the objective, the mechanism is regressive across firm size by construction.

Where the cost actually lands

The importer pays the tariff. The question that decides its effect is who absorbs the cost from there.

In consumer electronics with thin margins and price-sensitive demand, most of a duty passes to the buyer. That is why coverage has jumped straight to laptops and consoles getting more expensive, and it is likely right for the consumer segment.

Data centre servers behave differently. The buyer is an operator building against a capacity plan rather than a price point, and demand has been inelastic through two years of shortage. The alternative to buying is not buying less but building later. A duty on servers is therefore more likely to show up as slower build-out and higher cloud pricing than as cancelled orders.

That is the mechanism behind the industry's objection. The measure would not reduce demand for AI hardware. It would make the same build-out cost more and take longer, in a race the administration says it wants to win.

What a regional manufacturer should do now

Nothing has been decided, which should be the basis for planning rather than a reason to wait.

Establish which of your products would be classified as finished goods under an expanded scope, because that classification is the whole exposure. A contract manufacturer shipping bare boards is in a different position from one shipping assembled servers, and the tariff line is drawn between them.

Establish who bears the duty in your existing contracts. In much regional assembly the importer of record is the brand owner rather than the manufacturer, which moves the immediate cost but not the competitive consequence: a brand facing tariffs on Vietnamese-assembled servers will look at assembly elsewhere.

And watch the country-specific element, because that is where this becomes a regional story rather than a general one. Separate rates and quotas by country would turn a trade measure into a ranking of assembly destinations, and the countries with US fab investment behind them would rank differently from those without.

Why this may not arrive as described

Two things argue for caution about the reported shape.

The industry objection is unusually well aligned with the administration's own AI agenda. Tariffs that raise the price of data centre servers make domestic AI build-out more expensive at a moment when build-out speed is treated as strategic, and that tension has previously produced carve-outs rather than climb-downs.

The measure has no published rate, start date or scope document. It is a reported plan at the stage where scope contracts as it meets specific industries. We made the same point about the FCC's Chinese optical transceiver measure, which is a draft rather than a rule, and it applies with equal force here.