President Donald Trump issued a significant tariff announcement on Wednesday targeting the global semiconductor industry, declaring a 25% ad valorem tariff on certain imported semiconductors, high-end semiconductor manufacturing equipment, and related derivatives, effective from the 15th. Concurrently, the Trump administration announced a 25% tariff on specific high-end chip products, including NVIDIA's (NVDA.US) H200 AI chip and AMD's (AMD.US) MI325X AI chip. However, this specialized semiconductor tariff policy does not apply to chips imported to support the construction of the US technology supply chain and to strengthen domestic manufacturing capacity for semiconductor derivatives. Furthermore, the Trump administration emphasized it may impose broader tariffs on semiconductors and their derivative products in the near future, alongside a corresponding tariff offset plan to incentivize the development of domestic high-end semiconductor manufacturing, as previously announced. The White House stated in a release: "These latest tariff measures targeting NVIDIA and AMD do not apply to chips imported to support the construction of the US high-tech supply chain and enhance domestic US manufacturing capacity for semiconductor derivative products." "In the near future, President Trump may impose broader tariffs on imports of semiconductors and their derivative products and initiate a matching tariff offset plan to incentivize domestic high-end manufacturing, as announced previously."
This semiconductor tariff policy announcement is directly related to Section 232 of the Trade Expansion Act of 1962. The US Secretary of Commerce has also completed a Section 232 investigation into the importance of domestic production of semiconductors and high-end semiconductor manufacturing equipment. According to the latest semiconductor investigation results: "The current volume and circumstances of imports of semiconductors, semiconductor manufacturing equipment, and their derivative products threaten national security." The White House stated: "Today's actions will address the threat to national security in several ways, including by incentivizing large-scale semiconductor production in the United States and reducing our nation's dependence on foreign sources and foreign semiconductor supply chains. Semiconductors are vital to US economic, industrial, and military strength, and supply chain disruptions reliant on imports could undermine US industrial and military capabilities." "President Trump recognizes that restoring domestic production capacity for semiconductors, semiconductor manufacturing equipment, and their derivative products is critical for both economic and national security," the White House said in a statement.
On Tuesday local time, the Trump administration formally approved allowing the sale of NVIDIA's H200 series AI chips to China and established a new regulatory framework, potentially paving the way for larger-scale future shipments of H200 chips, albeit contingent on payment of a tax fee to the US government, despite strong opposition from some Washington hawks. Under the latest H200 export regulations, these chips will be reviewed by third-party testing laboratories before shipment to China to confirm their AI technology capabilities; the quantity of chips received by China must not exceed 50% of the total volume sold to US customers. Following the news of the tariffs targeting NVIDIA's H200 and AMD's MI325X, NVIDIA's stock price fell nearly 3% intraday before paring losses to close down 1.44%, with its market capitalization settling near $445 million; AMD's stock showed no major fluctuations, rising steadily throughout the session to close up 1.19%, with its market capitalization hovering around $364 billion.
What does Trump's latest tariff policy mean for the data center businesses of NVIDIA and AMD? The Trump administration's announcement of a 25% tariff on NVIDIA H200 and AMD MI325X chips primarily targets chips imported into the US from other countries, as well as scenarios involving importation into the US followed by re-export (transshipment)—specifically, imports of chips meeting certain performance standards that are ultimately deployed overseas (especially those potentially transshipped to other countries). The US had previously allowed NVIDIA's H200 to be exported to China under specific conditions, with the 25% fee/tariff acting as a "quid pro quo." This arrangement represents a policy compromise, allowing exports while generating revenue. In contrast, higher-end AI chip products, such as those based on the Blackwell architecture, are still considered more sensitive technologies at the US policy level and are therefore not currently within the scope of export licenses. This means they are simply not permitted for export and thus not subject to such tariff policies.
For the two companies, this largely means, first, an increase in cross-border sales costs. If these high-performance AI chips enter through US customs before being transshipped to overseas markets (especially regions with high demand like China), the tariff becomes an actual cost, raising the effective procurement cost for customers and potentially suppressing overseas demand. Secondly, it creates pressure on profits and pricing, as well as pressure to restructure semiconductor strategic supply chains. The Trump administration's latest tariff policy emphasizes "reducing dependence on foreign semiconductor supply chains," which directly intensifies significant pressure for fabless chip design companies like NVIDIA to relocate advanced node chip manufacturing processes entirely back to the US, a move that would substantially increase manufacturing costs and related capital expenditures. It is important to note that this semiconductor tariff policy targeting NVIDIA and AMD excludes chips intended for use in domestic US data centers, consumer devices, and industrial applications; meaning these tariffs will not apply to H200/MI325X chips directly deployed for use within the US. This latest semiconductor policy does not target more powerful AI chip series like Blackwell, reflecting a strategic compromise by policymakers between controlling exports and allowing some commercial exports.
Some media reports indicate that Chinese tech companies have already ordered over 2 million H200 AI chips, with a procurement price of approximately $27,000 per chip, far exceeding NVIDIA's inventory of 700,000 H200 chips. Last week at the International Consumer Electronics Show in Las Vegas, NVIDIA CEO Jensen Huang stated that strong demand from China and other parts of the world is driving up long-term lease prices for H200 AI chips in hyperscale cloud data centers, and the company is currently working with TSMC to increase H200 AI chip production.
A major boon for US semiconductor equipment manufacturers? Semiconductor equipment manufacturers—such as KLA Corp (KLAC.US), Lam Research (LRCX.US), and Applied Materials (AMAT.US)—are likely to benefit directly from the Trump administration's 25% tariff on certain imported semiconductors, semiconductor manufacturing equipment, and derivatives. The tariff move could prompt TSMC, Intel, Samsung Electronics, and SK Hynix to massively shift towards purchasing US-made semiconductor equipment for their new and expanded chip fabrication plants in the US. Furthermore, US semiconductor equipment manufacturers are also poised to be the largest beneficiaries of the rapid expansion in AI chip (AI GPU/AI ASIC) and DRAM/NAND memory chip production capacity. Citigroup predicted in a research report that the global semiconductor equipment sector is entering a "Phase 2 bull market cycle," suggesting a new bullish trajectory following the super cycle of 2024-25. A Citigroup analyst team led by renowned Wall Street semiconductor analyst Atif Malik forecasts that, with surging demand for AI chips and memory chips, the world's three largest chipmakers—TSMC, Samsung Electronics, and Intel—along with memory chip maker SK Hynix, will significantly raise their capital expenditure (capex) guidance for 2026 and beyond in upcoming earnings disclosures. This leads them to predict that global Wafer Fab Equipment (WFE) spending in 2026 is more likely to align with their "most optimistic forecast scenario."
As the global hyperscale AI data center construction boom, led by tech giants like Microsoft, Google, and Meta, intensifies, it is comprehensively driving chipmakers to accelerate expansion of 3nm-and-below advanced node AI chip production, CoWoS/3D advanced packaging capacity, and DRAM/NAND memory chip capacity. The long-term bullish thesis for the semiconductor equipment sector appears increasingly robust. From Citigroup's perspective, the "computing power-storage-advanced chip manufacturing" chain driven by the AI infrastructure frenzy ensures that semiconductor equipment capex stickiness is stronger than in any previous cycle: massive computing power demands for AI training/inference not only boost demand for advanced-node logic chips but also significantly increase demand intensity for high-end memory chips (especially HBM/enterprise SSD related); against the backdrop of rising chip manufacturing process complexity, the number of "front-end advanced process steps" per wafer increases, making sustained demand and improved order visibility more apparent on the equipment side.