Even with Semiconductor Exemptions, Tariffs Are Creating Cost Pressures That Will Ripple Through the AI Supply Chain, Impacting Every Stakeholder

While semiconductors have fortunately been exempted from the latest round of tariffs, the broader Artificial Intelligence (AI) ecosystem is still poised to face considerable headwinds due to the newly announced 10% baseline global tariff on U.S. imports (excluding China). The most immediate pressure will fall on AI infrastructure. Original Equipment Manufacturers (OEMs) operate interconnected global supply chains, with most U.S.-based firms primarily focusing on domestic assembly. Take Hewlett Packard Enterprise (HPE), a leading U.S. AI server OEM—it sources components and materials from Mexico, China, Taiwan, India, Singapore, Malaysia, and beyond, while also relying on commercial operations in the Czech Republic. Imposing tariffs on these foreign-sourced components and raw materials will substantially increase the cost of manufacturing AI servers. Even companies like SuperMicro, which emphasizes its “Made in the USA” branding, will not be immune. Despite a stronger domestic manufacturing presence, companies still depend heavily on overseas-sourced components.

This increase in production costs creates a difficult choice for server OEMs: absorb the costs and see profit margins shrink or pass them on to customers through server price hikes. In the United States, customers may have limited negotiating power and little choice other than to accept higher prices, but internationally, buyers will have greater flexibility and may be able to pivot to non-U.S. alternatives with more cost-competitive offerings (e.g., Lenovo, Huawei). This buying power will likely force OEMs to absorb some of the tariff costs into their bottom-line. But some costs will be passed on and added to the likely increases in labor and raw material cost, with the end result being large increases in data center Capital Expenditure (CAPEX) (and Operational Expenditure (OPEX) for hyperscalers, neo-cloud providers, and stakeholders (e.g., telco operators). The net impact will be a slowdown in data center expansion, and a subsequent reduction in demand for key components, especially high-performance semiconductors. There will also be a downstream effect on AI software and services. Infrastructure led by CAPEX expansion will result in increased prices cascading down to end users, increasing the cost of software, platforms, and services.

Beyond AI supply-related costs, AI-focused manufacturing projects face risks. Over the last several years, companies like TSMC and Intel have committed significant investment in reshoring semiconductor manufacturing and building capacity within the United States. Yet, these tariffs now threaten these projects by increasing the cost of construction—labor, imported raw materials, and foreign equipment (such as ASML’s lithography systems). The likely outcomes are grim: projects could be paused in hopes of riding out policy changes or canceled altogether due to diminishing Return on Investment (ROI). While a theoretical solution could involve reinvesting tariff revenue to support domestic manufacturing, that scenario appears politically unlikely. The most probable outcome is a long-term pullback in AI-related infrastructure investment. That means slower growth in server manufacturing, reduced expansion of data center capacity, and a broader decline in the U.S. position in the global AI market.

The temporary relief granted to many of the United States’ global trading partners has not extended to China. In response to the United States’ aggressive tariff stance, China’s decision to retaliate has triggered a sharp escalation—potentially setting the stage for an epoch-defining trade war. As of now (though conditions are rapidly evolving), the United States has implemented a 145% tariff on Chinese imports, while China has countered with tariffs of up to 84% on U.S. goods.

Semiconductors remain exempt for the time being, but the uncertainty around their future inclusion presents real risk—particularly given the heavy reliance many U.S. chipmakers have on the Chinese market. Based on 2024 revenue estimates, roughly 30% of Intel’s (US$15.5 billion), 25% of AMD’s (US$6.25 billion), 43% of Marvell’s (US$2.5 billion), 13% of NVIDIA’s (US$17.1 billion), and 46% of Qualcomm’s (US$18 billion) revenue is tied to China. If semiconductors are drawn into the trade war, the financial consequences across the sector will be significant.

Even without direct tariffs on chips, the semiconductor players will already be impacted. Vendors with exposure to consumer devices, especially smartphones and AI Personal Computers (PCs), will see immediate disruption. As demand in China contracts, companies like Intel, AMD, and Qualcomm will feel the ripple effect through key OEM partners such as Dell and HP. Apple, too, faces serious headwinds, with 17% of its revenue (US$66.8 billion) coming from Greater China—most of which could be jeopardized under current tariff conditions. Meanwhile, Chinese OEMs that have made strong inroads into the U.S. market are equally exposed. Lenovo, now holding approximately 15% of the U.S. PC market, and IEIT SYSTEMS, a rising server player in Western markets, both risk seeing their recent gains wiped out as the trade war intensifies.

The escalating U.S.-China trade war—and broader imposition of global tariffs—poses major strategic challenges for the AI industry, even as semiconductors remain temporarily exempt. Three main challenges impact stakeholder decision-making:

• Strategic Uncertainty: With tariffs shifting unpredictably, long-term planning becomes a high-risk endeavor. AI ecosystem stakeholders—from OEMs to cloud providers— will struggle to build resilient strategies when core component and import costs are in flux. Uncertainty will also impact semiconductor vendors as demand for leading chips will be impacted by tariffs directly impacting the cost basis of their technology and commercial partners.
• Structural Friction: Domestic manufacturing may offer a path to stability, but the reality is more complex. Building or scaling manufacturing capacity, particularly for high-tech infrastructure, requires years of lead time. Around a 5-year lag between investment and meaningful output presents a major friction point for players seeking short-term cost control or supply chain realignment.
• U.S.-Centric AI Supply Chain: U.S.-assembled AI infrastructure is set to face cost increases of around 10% due to higher tariffs on imported components, with fully imported servers facing similar hikes. This margin compression poses serious challenges for U.S. vendors looking to compete internationally, particularly in cost-sensitive regions like Europe where domestic and Asian alternatives may hold pricing advantages. These challenges are amplified by the current U.S. dominance of the AI value chain. While China continues to advance and Europe gains traction through players like Mistral.ai and Aleph Alpha, the lion’s share of AI chip vendors, server OEMs, hyperscalers, and model developers remain rooted in the United States—making the entire ecosystem more vulnerable to increasing costs.

In response, ABI Research  expects to see several strategic shifts across the AI supply chain, which, if implemented, could significantly reshape AI power distribution globally:

• U.S. AI server OEMs (e.g., HPE, Dell) will aim to expand production capacity in offshore facilities—particularly in Southeast Asia and Eastern Europe—to serve international markets and lessen the impact of U.S. tariffs on cost basis.
• Hyperscalers and cloud service providers will ramp up investment outside the United States, as they look to expand AI data center infrastructure, while retaining greater control over costs and ROI.
• Chipmakers will deepen partnerships with non-U.S. OEMs and Original Device Manufacturers (ODMs) to circumvent U.S.-centric price increases and remain competitive globally.
• International cloud service providers will diversify suppliers, turning to vendors like IEIT SYSTEMS  that are less exposed to U.S. tariffs.
• Software and service vendors will look to move workloads to adjacent regions outside of the United States to avoid higher inference costs that will likely be passed down to them from infrastructure vendors.

But the challenge isn’t limited to the AI supply chain. U.S. enterprises are also being forced into difficult decisions. Increasing input costs will put pressure on Information Technology (IT) budgets, and Chief Information Officers (CIOs)—already under scrutiny—will need to reassess their AI roadmaps. For some, AI will be seen as an expensive “nice to have,” especially for early-stage deployments that haven’t delivered ROI. These projects are likely to be paused or scrapped. Others may double down on AI to drive operational efficiency in response to tighter budgets. ABI Research expects that the net effect will be a near-term slowdown of AI adoption as financial constraints push organizations to prioritize ROI-positive initiatives.