The Trump administration is considering a plan to impose new tariffs on semiconductors and the products that contain them. Politico reported on August 27, 2026, citing eight people familiar with internal administration discussions. Under the plan, companies would receive duty-free import quotas based on their commitments to invest in domestic production, and finished products such as servers would also be brought into scope. The current measure, which took effect in January, exempts imports destined for data centers—but that exemption could be removed in a second phase.
The administration is trying to pursue two goals at once: bringing semiconductor production back to US soil and supporting the rapid buildout of AI data centers. But domestic supply capacity cannot immediately replace imports. How much of today's computing demand can a duty-free quota tied to future factory investment actually cover? Whether the tariff plan succeeds will depend on whether it can bridge that timing gap.
Tying duty-free import quotas to domestic production pledges
According to Politico, the plan favored by Commerce Secretary Howard Lutnick would let companies import a certain volume of semiconductors duty-free based on the amount of production they pledge to bring to the United States. Officials are also considering varying tariff rates and import quotas by country. Another option under discussion is phasing in the new tariffs gradually to avoid a sudden cost shock.
The scope could extend downstream from chips themselves. Candidates include laptops, gaming consoles, and the servers used in data centers. If servers are taxed, the impact would extend beyond companies that procure chips directly to cloud providers that import finished systems.
However, none of this is settled policy. As of the report, the tariff rate, covered products, the formula for duty-free quotas, and country-specific terms remain undecided, and could change significantly over weeks or months of negotiation. Whether the data center exemption will be eliminated is also still unresolved. What is clear is that the administration is weighing the next phase of the two-stage measure it began in January—but the details of the system are not yet known.
The basis for the duty-free quota will also shape how this works in practice. It has not been disclosed whether the quota would be tied to investment amount, future production capacity, or actual output. Nor is it clear whether quotas would be allocated per company or distributed by country or manufacturer. Without a formula that data center operators can use to weigh quotas against demand, they cannot estimate procurement costs that include tariffs, making it difficult to factor the policy into multiyear construction plans.
The January 25% tariff includes a data center exemption
The proclamation President Trump signed on January 14, 2026, established a two-phase adjustment of semiconductor imports under Section 232 of the Trade Expansion Act. The first phase imposed a 25% ad valorem tariff on a limited set of advanced computing chips and certain derivative products, while exempting imports used in US data centers and R&D. The same exemption applies to startups, consumer uses, and the public sector.
The proclamation states that in the second phase, after negotiations conclude, "substantial" tariffs could be imposed on semiconductors, manufacturing equipment, and derivative products. It also indicates the administration intends to include a tariff offset system that grants preferential treatment to companies investing in US semiconductor production and related supply chains. So the basic idea of linking domestic investment to import preferences is not new. What this latest report reveals is the possible shape of that preference—a production-volume-based import quota—along with the potential narrowing of the existing use-based exemptions.
The proclamation also lays out the rationale for reshoring production: the Commerce Department determined that while the United States consumes roughly a quarter of the world's semiconductors, it only fully manufactures domestically about 10% of the chips it needs. The administration views this import dependence as a national security risk and aims to use tariffs to shift the price gap enough to spur investment.
The same proclamation directed the Commerce Secretary to submit a report on the data center semiconductor market to the president by July 1. As of August 28, that report has not been made public. This means the supply outlook the administration used to design the new import quotas—and how it accounts for current demand and the pace of new factory ramp-ups—cannot be independently verified.
Breaking down the 15.6% construction cost estimate
The Computer & Communications Industry Association (CCIA) has published its own estimate of what could happen if the data center exemption is removed. Its calculation assumes IT equipment accounts for 78% of data center spending, that 80% of that equipment is imported, and that a 25% tariff is almost entirely passed through. Multiplying these factors together yields an effective 15.6% increase in cost per dollar of construction investment.
Based on this burden, CCIA estimates that about 20% of planned US AI data center investment for 2026–2030—$450 billion (450,000,000,000 dollars)—would be canceled, delayed past 2030, or shifted overseas. From that, the group projects an annual GDP loss of $90 billion (90,000,000,000 dollars) and risk to 243,000 jobs.
These figures are not government forecasts. CCIA is an industry association representing technology companies, and this scenario was built to support its policy argument against the tariffs. It assumes an 80% import ratio, near-total pass-through of tariff costs, and no double taxation from separately importing chips and then re-importing them as part of finished systems. Different policy design choices would change the outcome.
CCIA's sensitivity analysis shows which conditions move the estimated burden most:
| Policy scenario | CCIA's estimated annual GDP loss |
|---|---|
| 25% tariff, including data centers | About $90 billion (90,000,000,000 dollars) |
| Tariff rate lowered to 10% | About $45 billion (45,000,000,000 dollars) |
| 25% tariff retained, servers exempted | About $36 billion (36,000,000,000 dollars) |
| Rate lowered to 10%, servers also exempted | About $16 billion (16,000,000,000 dollars) |
Even under the same tariff policy, lowering the rate and exempting servers reduce the burden by different amounts—because how broadly finished products are taxed has a major effect on capital spending. Rather than treating the $90 billion (90,000,000,000 dollars) central estimate as a certain loss, it's more useful to read it as a benchmark for measuring whether the server and data center exemptions survive.
What to check before the tariff rate: import quotas and taxable boundaries
Even if duty-free quotas are tied to domestic production, AI operators will still pay tariffs on imports to cover any shortfall until factories come online. The current proclamation itself acknowledges that the United States fully manufactures domestically only about 10% of the chips it needs. Tariffs can shift the price gap between imported and domestically produced chips, but they cannot immediately expand domestic production capacity.
Expanding the scope to finished products would also complicate customs enforcement. In a joint letter in May, 19 industry groups including CCIA pointed out that importers of finished products often cannot determine, at the time of customs clearance, the country of origin, value, wafer fabrication location, or assembly and testing location of the embedded chips. That information is often held by component suppliers several tiers upstream in the supply chain, and it can change during manufacturing.
Industry groups are calling for a de minimis threshold that would exclude products with minimal semiconductor content, clearly defined product scope based on tariff classification codes, and safeguards against overlap with existing tariffs. If a standalone chip is taxed, and then taxed again after being assembled into a server overseas, the cumulative cost could exceed the level needed to spur domestic investment. Which documentation qualifies for duty-free quotas will also shape procurement planning.
The US AI Action Plan set out a policy of turning America's current advantages in data center construction, computing hardware, and AI models into long-term leadership. Reconciling semiconductor tariffs with that goal depends not just on the published tariff rate but on how import quotas and taxable boundaries are designed. What needs to be confirmed in any formal plan is whether the data center exemption survives, whether duty-free quotas will cover demand until domestic supply ramps up, where servers will be taxed, and whether double taxation on the same chip can be prevented. Allowing necessary imports until domestic supply increases would let AI infrastructure construction continue uninterrupted while still building demand toward future domestic production.