Section 232 AI chip tariffs and 2026 export rules

Between 2025 and 2026 the United States built two separate legal machines around advanced AI chips, and public discussion routinely conflates them. One is a tariff on imports, imposed under Section 232 of the Trade Expansion Act of 1962 by presidential proclamation and collected at the border by U.S. Customs and Border Protection (CBP). The other is a set of export controls on outbound sales, written by the Commerce Department's Bureau of Industry and Security (BIS) under the Export Administration Regulations and enforced through licenses. Different statute, different agency, opposite direction of trade. By 2026 the two were nevertheless drawn to the same technical thresholds, which is much of the reason they are confused.

The tariff arrived on January 15, 2026, at 25 percent ad valorem, under Proclamation 11002. Its covered category is drawn so narrowly that it describes individual accelerator products rather than a class of goods, and its exemptions cover almost every domestic use, including the data centers that consume most imported AI silicon.^[2][3] The export track ran from an April 2025 license requirement that functioned as a ban on NVIDIA's H20, through a July 2025 reversal and a reported arrangement under which NVIDIA and AMD would remit 15 percent of their China AI-chip revenue to the US government, to a January 2026 BIS rule that opened case-by-case licensing for the more capable H200 and AMD's Instinct MI325X.^[6][8][10] As of August 2026 the tariff's threatened second phase had not been imposed, no proclamation modifying it had appeared in the Federal Register, and shipments of H200s to China had begun but remained, in the words of the responsible Commerce official, "very few."^[17]

Section 232 and the semiconductor investigation

Section 232 of the Trade Expansion Act of 1962, codified at 19 U.S.C. 1862, lets the Secretary of Commerce investigate whether imports of an article threaten to impair national security. If the Secretary so finds, the President may adjust imports, most commonly through tariffs. The statute gives the Secretary up to 270 days to report to the President and the President 90 days after that to decide whether to act.

On April 1, 2025 Commerce Secretary Howard Lutnick initiated an investigation into imports of semiconductors, semiconductor manufacturing equipment (SME) and their derivative products. The public notice appeared in the Federal Register on April 16, 2025 at 90 FR 15950 under docket 250414-0066, with comments due May 7, 2025.^[1] The declared scope was unusually wide, reaching "semiconductor substrates and bare wafers, legacy chips, leading-edge chips, microelectronics, and SME components" as well as downstream electronics that merely contain chips.^[1] Nothing in that scope limits the inquiry to AI accelerators; the eventual action did.

Lutnick transmitted the report to the President on December 22, 2025, inside the statutory window.^[2] The findings recited in the proclamation are broad. Commerce concluded that the United States consumes roughly a quarter of the world's semiconductors while fully manufacturing only about 10 percent of the chips it requires, that semiconductors are essential to each of the 16 critical infrastructure sectors identified by National Security Memorandum 22 of April 30, 2024, and that dependence on foreign supply chains is therefore both an economic and a military risk.^[2] A separate finding singled out AI: chips that enable artificial intelligence "are an important element of many data centers currently in use, such that the importation in current quantities and circumstances poses a threat to the national security, when such importation does not contribute to the buildout of the United States technology supply chain."^[2]

The Secretary recommended a two-phase plan. Phase one would continue trade negotiations with foreign jurisdictions and impose immediately "a 25 percent ad valorem tariff on a very narrow category of semiconductors," with no tariff where the chips support the domestic technology supply chain. Phase two, after negotiations concluded, would bring "broader tariffs on semiconductors, at a rate of duty that is significant," paired with a tariff offset program giving preferential treatment to firms investing in US semiconductor production.^[2]

The 100 percent threat and the negotiated ceilings

The loudest moment in the sequence produced no legal instrument at all. On August 6, 2025, at a White House event alongside Apple's announcement of additional US investment, President Trump said: "We'll be putting a tariff on of approximately 100 percent on chips and semiconductors." He immediately attached an exemption: "But if you're building in the United States of America, there's no charge," extending it to firms still under construction, and warning that a company which promised to build and did not would find that the duty "accumulates and we charge you at a later date."^[11] The remark was verbal. No 100 percent semiconductor tariff was ever proclaimed, and the threshold for qualifying investment was never defined. Firms with announced US fabrication commitments, among them Intel, Micron, TSMC and Samsung, were the presumed beneficiaries, and Apple supplied the occasion.^[11]

The threat also did not survive contact with trade negotiations. The August 21, 2025 United States-European Union framework statement confirmed a 15 percent ceiling covering EU semiconductors, pharmaceuticals other than generics, and lumber, which foreclosed a 100 percent levy on European chips before any Section 232 action existed. The same package recorded EU commitments to purchase at least $40 billion of US AI chips and $750 billion of energy products through 2028.^[12] A later agreement with Taiwan, implemented in the Federal Register on May 28, 2026, capped tariffs on several Taiwanese sectors and recorded a Taiwanese commitment to facilitate $250 billion of investment in US semiconductor production, though its tariff annexes did not set a semiconductor-specific rate.^[22] These ceilings matter for phase two: a broad semiconductor tariff would run into rates already fixed by agreement for several of the largest supplying jurisdictions.

Proclamation 11002 and the 25 percent tariff

President Trump signed Proclamation 11002, "Adjusting Imports of Semiconductors, Semiconductor Manufacturing Equipment, and Their Derivative Products Into the United States," on January 14, 2026. It was published on January 20, 2026 at 91 FR 2443.^[2] CBP issued implementing guidance the same evening the proclamation was signed, as CSMS #67400472.^[5]

Clause 3 imposes a 25 percent ad valorem duty on Covered Products, effective for goods entered for consumption or withdrawn from warehouse for consumption "on or after 12:01 a.m. eastern standard time on January 15, 2026," to continue "unless such action is expressly reduced, modified, or terminated."^[2] The duty is additional to any other duties otherwise applicable, and it applies on top of preferential rates under free trade agreements and preference programs.^[3]

Covered products and the technical windows

The Annex inserts a new U.S. note 39 into subchapter III of chapter 99 of the Harmonized Tariff Schedule and creates nine new headings, 9903.79.01 through 9903.79.09.^[3][4] The covered universe is not all semiconductors and not bare integrated circuits. It is limited to goods classifiable in three HTS subheadings:

• 8471.50 (processing units for automatic data processing machines)
• 8471.80 (other units of automatic data processing machines)
• 8473.30 (parts and accessories of automatic data processing machines)

Within those subheadings, an import is a "semiconductor article" only if it is "a logic integrated circuit, or an article that contains a logic integrated circuit" meeting one of two performance windows:^[3]

Scroll sideways for more →

The note adopts the same TPP formula BIS uses in its export controls: TPP equals 2 times "MacTOPS" times the bit length of the operation, aggregated over all processing units on the integrated circuit, calculated on dense rather than sparse matrix performance, and using the highest value the manufacturer claims.^[3] "Total DRAM bandwidth" is the aggregate bandwidth between the chip and DRAM, including copackaged high bandwidth memory and non-copackaged GDDR, but excluding memory reached remotely over an interconnect.^[3]

Those are not broad bands. They are narrow windows with both a floor and a ceiling, which is why the measure covers individual products rather than a product class. NVIDIA lists the H200 SXM at 141 GB of memory and 4.8 TB/s of bandwidth, or 4,800 GB/s, which falls inside window 1.^[23] AMD's Instinct MI325X sits in window 2. Analysts at the Center for a New American Security described the proclamation as "establishing a 25 percent import tariff on H200s and AMD's MI325X", meaning the H200 and MI325X pair that BIS names in the export rule issued the previous day.^[16] Parts faster than window 2, including NVIDIA Blackwell accelerators, and parts slower than window 1 are both outside the tariff.

Exemptions

Only heading 9903.79.01 carries the 25 percent duty. The other eight headings collect the ordinary rate.

Scroll sideways for more →

Three of the definitions do real work. A "U.S. data center" is defined as "a facility that requires greater than 100 megawatts (MW) of new load dedicated to AI inference, training, simulation, or synthetic data generation," which restricts the largest exemption to hyperscale AI builds rather than any server hall.^[3] A "startup" is an "emerging growth company" as defined at 15 U.S.C. 77b(a)(19), the securities-law term, not a colloquial one.^[3] The consumer-electronics heading is explicitly enumerated to include "gaming, personal computing, professional visualization, workstation applications, and automotive applications," and the civil-industrial heading to include "factory robotics and industrial machinery."^[3] The proclamation also lets the Secretary extend relief to "other uses that the Secretary determines contribute to the strengthening of the United States technology supply chain."^[2]

Because the exemptions are defined by end use rather than by importer, they are administered through the entry classification itself, and each requires the importer to enter under the corresponding 9903.79.0x heading.^[4][5]

Stacking, drawback and zones

Note 39 removes the covered goods from most other tariff stacks. Articles entered under 9903.79.01 are not subject to the vehicle, vehicle-parts, medium- and heavy-duty truck, copper, aluminum, derivative aluminum, iron and steel, or derivative iron and steel Section 232 headings, nor to the Canada and Mexico headings 9903.01.10 and 9903.01.01, nor to headings 9903.01.24 through 9903.01.76 and 9903.02.01 through 9903.02.71.^[3] Clause 5 of the proclamation states the same result at the level of executive orders: covered products are not subject to the reciprocal tariffs of Executive Order 14257 of April 2, 2025, or to Executive Orders 14193 and 14194 of February 1, 2025 on the northern and southern borders.^[2] Where a covered product would otherwise fall under two Section 232 proclamations, this one governs.^[2]

No drawback is available on the duty. Goods admitted to a US foreign trade zone on or after the effective date must be admitted as privileged foreign status unless they qualify as domestic status, which prevents zone processing from reclassifying the duty away.^[2]

Phase two

Phase two remains prospective. Clause 1 directs the Secretary and the United States Trade Representative to pursue agreements under 19 U.S.C. 1862(c)(3)(A)(i) and says that, depending on the outcome, the President "may consider imposing significant tariffs on imports of semiconductors, semiconductor manufacturing equipment, and their derivative products, as well as an accompanying tariff offset program to incentivize domestic manufacturing."^[2] Clause 2 required a negotiation update within 90 days. Paragraph 12 notes that if an agreement is not entered into within 180 days of the proclamation, is not carried out, or proves ineffective, the President may take further action.^[2] Clause 9 required the Secretary, by July 1, 2026, to provide "an update on the market for semiconductors that are used in United States data centers, so that the President may determine whether it is appropriate to modify the tariff imposed in this proclamation."^[2]

As of August 1, 2026 no proclamation broadening the semiconductor tariff, and no notice modifying the 9903.79 headings, had been published in the Federal Register, and Commerce had not published the July 1 data-center update. The 25 percent narrow tariff was still the only Section 232 semiconductor duty in force.

The export-control track: H20, H200 and the revenue share

Export controls are a separate instrument with a separate history. BIS has restricted advanced computing chips to China and Macau since October 2022. An interim final rule effective November 17, 2023 replaced the earlier "bits times TOPS" metric in ECCN 3A090 with total processing performance and a performance-density criterion, and broadened the country scope to Country Group D:5.^[7] Those thresholds are the reason vendors designed China-specific parts in the first place.

The H20 license requirement

NVIDIA built the H20 to sit below the 2023 thresholds. In its results for the quarter ended April 27, 2025 the company disclosed: "On April 9, 2025, NVIDIA was informed by the U.S. government that a license is required for exports of its H20 products into the China market. As a result of these new requirements, NVIDIA incurred a $4.5 billion charge in the first quarter of fiscal 2026 associated with H20 excess inventory and purchase obligations as the demand for H20 diminished."^[8] The company also reported $4.6 billion of H20 revenue recognized before the restriction and a further $2.5 billion it could not ship.^[8] AMD faced a parallel restriction on its MI308. In practice the license requirement functioned as a ban.

Reversal and the reported 15 percent

The policy reversed within three months. NVIDIA chief executive Jensen Huang lobbied publicly and privately to reopen the market, and in July 2025 the administration signaled that H20 licenses would be granted. Licenses were issued in the first week of August 2025, and AMD won comparable clearance for the MI308.^[24] On August 10, 2025 the New York Times and the Financial Times reported that both firms had agreed to remit 15 percent of the revenue from those China sales to the US government, an arrangement The Register described as covering "15 percent of H20 and MI308 revenues."^[24][25] Against NVIDIA's reported $10.5 billion of lost China sales in the first half of 2025 and AMD's $800 million shortfall in one quarter, the arrangement was estimated to be worth about $900 million a quarter to Washington.^[24]

The arrangement was never codified. In its quarterly filing of August 27, 2025 NVIDIA disclosed that US officials expected 15 percent of the revenue from licensed H20 sales and warned that "any request for a percentage of the revenue by the USG may subject us to litigation, increase our costs, and harm our competitive position and benefit competitors that are not subject to such arrangements." Chief financial officer Colette Kress said at the time that "to date, the USG has not published a regulation codifying such a requirement," and that although a select number of Chinese customers had received licenses, the company had made no H20 sales under them.^[10] The legal objections are structural. Article I, Section 9 of the Constitution provides that no tax or duty shall be laid on articles exported from any state, and the Export Control Reform Act of 2018, which is the statutory basis for BIS licensing, does not provide for charging exporters a share of revenue in return for a license. Whether a payment characterized as voluntary escapes the Export Clause, and who would have standing to litigate it, both remain untested.

Beijing's counter-pressure

The Chinese state then discouraged the purchases the US had just permitted. In August 2025, according to Bloomberg, Chinese authorities wrote to firms advising against using H20 accelerators for work involving "government or national security-related work"; NVIDIA replied that "the H20 is not a military product or for government infrastructure."^[25] In September 2025 the Cyberspace Administration of China ordered Alibaba, ByteDance and others to suspend testing and cancel orders for NVIDIA's RTX Pro 6000D, and Chinese authorities opened an antitrust action against the company, with domestic alternatives including Huawei's Ascend 910C and CloudMatrix 384 cited as the reason a ban was now affordable.^[26] The effect was that a legal pathway existed while commerce did not flow.

The H200 opening

On December 8, 2025, in a Truth Social post reported the following day, Trump wrote that he had "informed President Xi, of China, that the United States will allow NVIDIA to ship its H200 products to approved customers in China, and other Countries, under conditions that allow for continued strong National Security," and added that "$25% will be paid to the United States of America." He said the permission covered the H200 only, not Blackwell or the forthcoming Rubin generation, and indicated the approach would extend to AMD, Intel and other US firms.^[13]

BIS gave the decision regulatory form in a final rule filed on January 13, 2026 and published, effective the same day, on January 15, 2026 at 91 FR 1684 (RIN 0694-AK43, docket 260112-0028).^[6] The rule changed the license review policy for exports to end users in China and Macau from a presumption of denial to case-by-case review for "commodities with a TPP less than 21,000 and a 'total DRAM bandwidth' less than 6,500 GB/s, such as the NVIDIA H200 or AMD MI325X."^[6] Presumption of denial was retained for reexports and in-country transfers to Macau and Country Group D:5, and for exports to entities headquartered, or whose parent is headquartered, in those destinations.^[6]

Case-by-case review is available only if the applicant provides nine certifications, set out in a new paragraph (dd) of supplement no. 2 to part 748:^[6]

Scroll sideways for more →

The testing-lab qualification is a live control rather than a formality: BIS "may revoke the qualification of any third-party testing lab at any time and for any reason," and revocation suspends the case-by-case policy for every exporter using that lab until a new one is named.^[6] BIS also modified section 744.23 so that the same case-by-case standard applies under the supercomputer and advanced-node end-use controls.^[6]

The rule contains no revenue-share provision of any kind. The 25 percent figure comes from the President's announcement and subsequent reporting, not from the regulation. The published analysis that best reconciles the two is that the payment is not collected as a payment at all. Because the Section 232 windows bracket exactly the H200 and MI325X, and because every exemption in note 39 is defined by a United States end use, a chip imported into the United States for onward export to China can claim no exemption and bears the 25 percent duty. CNAS analysts, writing the day after both instruments took effect, said the combination "effectively imposes a 25 percent export fee on these chips to China" alongside the licensing rule.^[16] On that reading the announced 25 percent is a condition of the H200 opening whose legal vehicle is the import tariff, while the earlier 15 percent on the H20 and MI308 remains a reported but never codified arrangement.

What has actually shipped

Movement was slow. In January 2026 NVIDIA was reported to be demanding advance payment for Chinese H200 orders, which the company denied, saying "we do not require upfront payment, and would never require customers to pay for products that they do not receive."^[14] On March 17, 2026 Huang said "we've been licensed for many customers in China for H200," that purchase orders had arrived, and that NVIDIA was restarting manufacturing of the part, adding that the situation was "different from what it was two weeks ago or three weeks ago."^[15]

Financial results told a flatter story. NVIDIA's first quarter of fiscal 2027, ended April 26, 2026 and reported on May 20, 2026, produced record revenue of $81.6 billion, up 85 percent year on year, with data center revenue of $75.2 billion. The outlook for the following quarter was $91.0 billion plus or minus 2 percent, and the company stated that it "is not assuming any Data Center compute revenue from China in its outlook."^[9]

In May 2026 Commerce cleared roughly ten Chinese firms to buy the H200, reportedly including Alibaba, Tencent and ByteDance, with no deliveries at that point.^[17] The first official confirmation of movement came on July 14, 2026, when Under Secretary of Commerce for Industry and Security Jeffrey Kessler told the House Foreign Affairs Committee that shipments had begun but the number was "very few," and that "there have been minimal exports of any H200s to China so far." Reuters reported at the same time that ZTE and two other Chinese firms had received the most recent approvals to buy advanced AI chips from NVIDIA and AMD.^[17]

How the two instruments fit together

Read side by side, the tariff and the export rule are calibrated to each other.

• The export rule sets a ceiling: TPP below 21,000 and total DRAM bandwidth below 6,500 GB/s may be licensed case by case for China and Macau.^[6]
• The tariff sets two narrow windows, the first well inside that ceiling and the second straddling it at 20,800 to 21,100, each bracketing a single shipping product.^[3]
• The tariff's product coverage is 8471.50, 8471.80 and 8473.30, which are assembled units and boards, not the bare integrated circuits of heading 8542.^[3] By contrast, the Section 301 action on Chinese chips reaches 8541 and 8542 goods.^[18]
• Every tariff exemption is defined by a United States end use, so a chip routed through the United States for export to China qualifies for none of them.^[3]

The consequence is that the same 25 percent number appears on both tracks, and conflating them is easy. They remain legally distinct: the tariff is collected by CBP from the importer of record under 19 U.S.C. 1862, while the license conditions are enforced by BIS against the exporter under the Export Administration Regulations, with different penalties, different appeal routes and different beneficiaries.

Adjacent measures

Several other instruments run alongside and are frequently mistaken for parts of the same action.

Section 301 on Chinese semiconductors. USTR initiated a Section 301 investigation into China's targeting of the semiconductor industry for dominance on December 23, 2024. On December 23, 2025 the Trade Representative determined the conduct actionable and imposed a tariff action, published at 90 FR 60848 on December 29, 2025, at an initial rate of 0 percent, "increasing in 18 months on June 23, 2027, to a rate to be announced not fewer than 30 days prior to that date."^[18] The covered lines include high-purity silicon (2804.61), doped chemical elements and compounds (3818.00), and headings 8541 and 8542. The new rate will be additional to the existing 50 percent Section 301 tariff on semiconductors from China imposed under the earlier forced-technology-transfer investigation.^[18] The action is therefore a scheduled future tariff rather than a present one, and it targets China's semiconductor industry rather than AI accelerators.

Federal procurement. Section 5949 of the James M. Inhofe National Defense Authorization Act for Fiscal Year 2023 (Public Law 117-263, enacted December 23, 2022) prohibits executive agencies from procuring or obtaining products or services containing semiconductors designed, produced or provided by SMIC, ChangXin Memory Technologies or Yangtze Memory Technologies, or their subsidiaries and affiliates, along with entities determined to be owned or controlled by a foreign country of concern. The prohibition takes effect December 23, 2027, and the Federal Acquisition Regulation Council has been developing implementing rules since 2024.^[20]

Entity listing and end-user programs. BIS revoked Validated End-User authorizations for Intel Semiconductor (Dalian), Samsung China Semiconductor and SK hynix Semiconductor (China) effective December 31, 2025, ending general authorization for those China fabs and returning them to individual licensing.^[19] The Entity List remains the principal instrument for named-party restrictions and is separate from both the tariff and the general licensing policy.

Enforcement. In testimony on July 14, 2026 Kessler reported that BIS monetary penalties rose from $16 million in 2024 to $324 million in 2025, an increase of more than eighteen times, and that 2026 penalties had already doubled the 2025 figure. He cited a $252 million settlement with Applied Materials over transshipment of semiconductor manufacturing equipment, a $95 million penalty against Cadence Design Systems over export of semiconductor design software, and $36 million in penalties in a Bosch matter, along with an indictment of three individuals over a scheme to divert servers containing advanced semiconductors through Southeast Asia.^[17]

Industry impact

On the import side the tariff's practical reach in its first phase is narrow by design. The exemption headings are written around the end uses that account for most of the demand: hyperscale AI data centers, research, consumer electronics, industrial equipment and public-sector systems. What is left inside heading 9903.79.01 is chiefly imports that serve none of those, which includes chips brought in for onward export. The binding limit on the largest exemption is the definition of a qualifying data center, which requires more than 100 MW of new load dedicated to AI work, so smaller colocation and enterprise deployments cannot use heading 9903.79.03 and must find another exemption or pay.^[3]

The larger commercial question is phase two. The Information Technology and Innovation Foundation estimated in June 2026 that a 25 percent tariff applied broadly to semiconductors would reduce US GDP growth by 0.2 percent in the first year, about $58 billion, and cost $1.6 trillion cumulatively over ten years, with a 50 percent tariff costing $122 billion in the first year and $4.4 trillion over a decade. The authors, Meghan Ostertag and Trelysa Long, argued that higher chip costs would raise prices for vehicles and electronics and impede AI infrastructure investment, and recommended removing the existing tariffs rather than expanding them.^[21] Those are model-based projections from a technology policy think tank, not measured outcomes.

On the export side the commercial result through mid-2026 was near zero. NVIDIA recorded no data center compute revenue from China in the quarter ended April 26, 2026 and excluded China from its next quarter's guidance.^[9] Licenses existed, purchase orders existed, manufacturing had restarted, and shipments had begun by July 2026, but at volumes the responsible official called minimal.^[15][17]

The policy backdrop on the Chinese side is a domestic substitution effort that predates both instruments and that the export controls accelerated. The principal domestic alternatives are Huawei's Ascend accelerator line, foundry capacity at SMIC and Hua Hong Grace Semiconductor, memory from CXMT and YMTC, and equipment from NAURA, AMEC and SiCarrier. Every US measure described here is calibrated against the assumption that restricting supply slows Chinese AI capability faster than it accelerates Chinese self-sufficiency, and that assumption is the central contested question in the debate.

Reception

Reception split along predictable lines and was sharpest on the revenue share, which had no clear precedent: the federal government does not ordinarily take a percentage of a company's export revenue. NVIDIA treated the arrangement as a legal exposure rather than defending it, telling investors that any such request "may subject us to litigation, increase our costs, and harm our competitive position," and noting that no regulation codified the requirement.^[10] The constitutional objection is the Export Clause; the statutory one is that the Export Control Reform Act contains no authority to charge an exporter for a license. Neither has been tested in court.

National-security analysts questioned the H200 opening on its merits. Janet Egan and James Sanders of the Center for a New American Security wrote on January 16, 2026 that the policy could transfer up to 890,000 H200-equivalent chips to China, roughly twice what Chinese fabs were expected to produce that year, and comparable to the compute OpenAI had deployed as of October 2025. They argued the safeguards were weak in practice, that certifications against military end users are difficult to enforce against military-civil fusion and shell companies, and that the requirement not to delay US orders was "impossible to credibly meet" given advanced-node capacity constraints. They also warned the rule set a precedent for further relaxations.^[16]

Supporters inside the administration framed the package as leverage rather than concession: tariffs and licenses together would pull manufacturing onshore while extracting value from a market that would otherwise be served by smuggling or by domestic Chinese accelerators. Kessler's July 2026 testimony made the industrial-policy case explicitly, describing Section 232 as producing "a manufacturing renaissance," noting more than a dozen BIS investigations and tariffs already imposed on steel, aluminum, copper, pharmaceuticals, semiconductors, autos, trucks and lumber, and requesting an additional $20 million for the Section 232 program and a 147 percent increase for enforcement, from $122 million to $301 million, taking the enforcement unit from 299 to 669 positions.^[17]

Beijing's own position was the mirror image, discouraging purchases on security grounds while promoting domestic parts, which left both governments arguing against the same transaction for opposite reasons.^[25][26] As of August 2026 the combined picture was a tariff in force but narrow, a second phase announced but not imposed, an export channel open but barely used, and a payment condition announced by the President that appears nowhere in the rule implementing the policy. See also AI regulation and the CHIPS and Science Act.

Timeline of actions

Scroll sideways for more →

References

1. ^Bureau of Industry and Security, "Notice of Request for Public Comments on Section 232 National Security Investigation of Imports of Semiconductors and Semiconductor Manufacturing Equipment," 90 FR 15950, April 16, 2025. govinfo.gov/...2025-06591
2. ^Proclamation 11002 of January 14, 2026, "Adjusting Imports of Semiconductors, Semiconductor Manufacturing Equipment, and Their Derivative Products Into the United States," 91 FR 2443, January 20, 2026. govinfo.gov/...2026-01052
3. ^The White House, Annex to Proclamation 11002 (new U.S. note 39 to subchapter III of chapter 99 and headings 9903.79.01 through 9903.79.09), January 2026. whitehouse.gov/...SemiConductor.prc_.rel-ANNEX.pdf
4. ^U.S. International Trade Commission, Harmonized Tariff Schedule, headings 9903.79.01 through 9903.79.09. hts.usitc.gov/...search
5. ^U.S. Customs and Border Protection, CSMS #67400472, "GUIDANCE: Section 232 Import Duties on Semiconductors," January 14, 2026. content.govdelivery.com/...4047318
6. ^Bureau of Industry and Security, "Revision to License Review Policy for Advanced Computing Commodities," final rule, 91 FR 1684, January 15, 2026 (RIN 0694-AK43, docket 260112-0028). govinfo.gov/...2026-00789
7. ^Bureau of Industry and Security, "Implementation of Additional Export Controls: Certain Advanced Computing Items; Supercomputer and Semiconductor End Use; Updates and Corrections," 88 FR 73458, October 25, 2023. govinfo.gov/...2023-23055
8. ^NVIDIA, "NVIDIA Announces Financial Results for First Quarter Fiscal 2026," May 28, 2025. investor.nvidia.com/...default
9. ^NVIDIA, "NVIDIA Announces Financial Results for First Quarter Fiscal 2027," May 20, 2026. investor.nvidia.com/...default
10. ^Laura Bratton, Yahoo Finance, "Nvidia still hasn't finalized deal to kick 15% of H20 China chip sales back to the US government," August 28, 2025. finance.yahoo.com/...o-the-us-government-230229161
11. ^The Register, "Trump teases 'approximately' 100 percent chip tariffs," August 7, 2025. theregister.com/...1519054
12. ^The Register, "EU-US deal confirms 15% tariff cap on semiconductors, more," August 21, 2025. theregister.com/...678749
13. ^Simon Sharwood, The Register, "Trump says Nvidia can sell H200s to China," December 9, 2025. theregister.com/...2020308
14. ^The Register, "Nvidia reportedly wants up front payments for Chinese H200s," January 8, 2026. theregister.com/...4767852
15. ^The Register, "Nvidia H200 back on in China, production ramping: Huang," March 17, 2026. theregister.com/...5223086
16. ^Janet Egan and James Sanders, Center for a New American Security, "CNAS Insights: Unpacking the H200 Export Policy," January 16, 2026. cnas.org/...ights-unpacking-the-h200-export-policy
17. ^David Shepardson and Karen Freifeld, Reuters, "US official says Nvidia has begun shipping powerful H200 AI chips to China," July 14, 2026. yahoo.com/...fficial-says-shipments-h200-145319415 ; Jeffrey I. Kessler, prepared statement, House Foreign Affairs Committee hearing "FY27 BIS Budget: the AI Arms Race and the ICTS Office," July 14, 2026. docs.house.gov/...A00-Wstate-KesslerJ-20260714.pdf
18. ^Office of the United States Trade Representative, "Notice of Action: China's Acts, Policies, and Practices Related to Targeting of the Semiconductor Industry for Dominance," 90 FR 60848, December 29, 2025. govinfo.gov/...2025-23912
19. ^Bureau of Industry and Security, "Revocation of Validated End-User Authorizations in the People's Republic of China," 90 FR 42321, September 2, 2025. govinfo.gov/...2025-16735
20. ^Federal Acquisition Regulatory Council, "Federal Acquisition Regulation: Prohibition on Certain Semiconductor Products and Services," May 3, 2024. govinfo.gov/...2024-08735
21. ^Meghan Ostertag and Trelysa Long, Information Technology and Innovation Foundation, "Section 232 Semiconductor Tariffs Could Undermine US Economic Growth," June 4, 2026. itif.org/...ffs-could-undermine-us-economic-growth
22. ^Office of the United States Trade Representative, "Implementing Certain Tariff-Related Elements of a Trade and Security Agreement Between the American Institute in Taiwan and the Taipei Economic and Cultural Representative Office in the United States," 91 FR 31818, May 28, 2026. govinfo.gov/...2026-10571
23. ^NVIDIA, "NVIDIA H200 Tensor Core GPU" product specifications. nvidia.com/...h200
24. ^Simon Sharwood, The Register, "Report says US to charge Nvidia, AMD, 15 percent China export tax," August 11, 2025. theregister.com/...719821
25. ^Tobias Mann, The Register, "China warns against using Nvidia H20s for government jobs," August 12, 2025. theregister.com/...1513897
26. ^Tobias Mann, The Register, "Beijing demands Chinese tech terminate Nvidia orders," September 18, 2025. theregister.com/...1128121