It is reported that the new round of tariffs will no longer be limited to chips themselves but will cover laptops, gaming consoles, and data center servers. The U.S. technology sector is on high alert as the AI infrastructure boom directly confronts significant policy barriers.
The Trump administration is considering launching a new round of broad-based semiconductor tariffs. According to POLITICO, which cited eight informed sources, the proposal is still under discussion and may be adjusted in the coming weeks or even months, but its potential impact has already raised significant alarm within the U.S. technology sector.
These tariffs may no longer be limited to chips themselves but could expand to products that heavily utilize chips, including laptops, gaming consoles, and data center servers. Commerce Secretary Howard Lutnick favors linking tariff exemptions to corporate investment in chip manufacturing within the United States, aiming to drive the reshoring of advanced production capacity, while the government is also considering establishing a phased implementation transition period.
The contradiction lies in the fact that the United States is currently experiencing a boom in AI data center construction, amid already tight supplies of high-end chips. U.S. technology firms argue that raising the cost of imported chips at this juncture would not only increase data center construction costs but could also force some companies to cut back on investment.
Jonathan McHale, head of digital policy at the Computer & Communications Industry Association, stated that the scale and capital investment in current data center construction have been compared to building the transcontinental railroad, noting that "increasing costs and reducing predictability" will make investment more difficult.
The White House has emphasized that promoting the reshoring of semiconductor manufacturing is a top priority for the Trump administration, claiming that related policies have already facilitated hundreds of billions of dollars in investment.
Linking tariff exemptions to domestic investment may exacerbate supply shortages
Technology companies are recently intensifying their lobbying efforts, hoping that the new tariffs will align as closely as possible with the proposals from early this year, which retained broader exemptions for domestic uses such as data centers.
However, informed sources reveal that the tone of recent negotiations is deteriorating. Government officials have signaled that exemptions previously granted for uses such as data centers, research and development, startups, consumer applications, civilian industry, and the public sector may be eliminated in the future.
Lutnick's preferred approach involves setting a quota for duty-free chip imports, with the exemption volume determined by the number of chips companies commit to producing in the United States. The issue is that domestic U.S. production capacity is far insufficient to meet demand, which could create a larger gap between the duty-free supply available and actual demand.
Michael Sobolik, a senior fellow at the Hudson Institute, also pointed out that shifting large-scale chip production to the United States "would be very expensive," given that production costs are lower in other parts of the world.
At present, the Department of Commerce has not yet determined specific tariff rates and other key details. One proposal under discussion involves setting different tariff rates and quotas for different countries, along with separate policy guidelines for major semiconductor manufacturers.
As bets on AI rise alongside costs, policy contradictions come to light
The technology industry does not oppose the goal of domestic manufacturing; rather, it argues that U.S. advanced chip production capacity cannot fill the gap in the short term. More than 90% of the world's most advanced semiconductor production capacity is concentrated in Asian supply chains.
Investments in advanced chip fabs often run into billions of dollars, with construction cycles typically spanning several years. Although Taiwan Semiconductor (TSM) has committed to investing $265 billion in its Arizona facility—a project regarded as the largest foreign direct investment in U.S. history—the company expects that even after full completion, only about 30% of its most advanced production capacity will be located in the United States, with significant volume not coming online for several years.
This means that before domestic supply in the United States truly expands, tariffs will first hit companies reliant on imported chips.
Beyond data centers, the costs of electronic products such as servers, computers, and televisions may also rise. Chip design firms like NVIDIA and AMD could similarly be impacted, as they rely on overseas manufacturers for chip production. Meanwhile, companies like Apple may face competitive disadvantages in foreign markets, where rivals can purchase the same chips without bearing U.S. import tariffs.
Sujai Shivakumar, an economist at the Center for Strategic and International Studies, points out that tariffs alone cannot address the structural weaknesses of the U.S. chip industry: they cannot increase the number of skilled technicians, shorten approval cycles, or expand reliable systems for electricity, water supply, and suppliers.
A more immediate issue is timing. Technology and trade officials close to the Trump administration estimate that a large-scale expansion of domestic chip manufacturing capacity in the United States will take at least five years, whereas the buffer periods Trump previously granted for other tariffs were often much shorter.
Therefore, the core of the current policy debate is not merely whether to develop U.S. chip manufacturing, but whether the United States is willing to let the AI industry bear higher chip costs to drive supply chain reshoring before domestic production capacity is fully established.
As one technology industry participant involved in the relevant negotiations stated, the volume of duty-free chips currently envisioned by the government “may not even cover the demand of hyperscale cloud service providers,” because the United States “currently lacks sufficient production capacity to manufacture these chips.”