Lateo.net - Flux RSS en pagaille (pour en ajouter : @ moi)

🔒
❌ À propos de FreshRSS
☷
△ 🔃
Il y a de nouveaux articles disponibles, cliquez pour rafraîchir la page.
À partir d’avant-hierAnalyses, perspectives

Taiwan Shows Resilience in Trying Times

Par : Dennis Yang

Executive Summary:

  • Taiwan is showing resilience in the face of headwinds ranging from domestic politic strife, increased military pressure from the People’s Republic of China (PRC), and uncertainty regarding U.S. political support.
  • Taiwan’s semiconductor sector is benefitting massively from the global artificial intelligence (AI) boom. Its gains contributed to Taiwan posting almost 13 percent quarterly GDP growth and reaching record stock market highs.
  • In a bid to strengthen its relationship with the United States, Taiwan is investing $265 billion into American semiconductor manufacturing through Taiwan Semiconductor Manufacturing Corporation (TSMC). Taiwan has also hosted multiple high profile bipartisan congressional visits since the beginning of 2026.
  • Domestic political strife, now centered around a proposed drone procurement bill, may affect U.S. assessments of Taiwan’s will to fight. Taiwan is also waiting for the Trump administration to approve a delayed $14 billion arms sale, which will be critical to its defense against a potential military campaign by the PRC.

Taiwan today faces an unusually difficult geopolitical environment in which multifaceted pressure from a militarily powerful People’s Republic of China (PRC) is mounting, domestic political strife is undermining defense readiness, and support from the United States is inconsistent. Taiwan has dealt with each of these challenges separately in the past, but this is the first time it must grapple with all three simultaneously.

Despite the difficult geopolitical situation, Taiwan has shown impressive resilience. The island nation has become a leader in producing hardware for artificial intelligence (AI), which has made it the world’s fastest growing advanced economy. At the same time, Taiwanese officials have quietly worked with their American counterparts to deepen ties. The missing piece of the puzzle is its military resilience: legislation to support Taiwan’s drone industry remains stalled and a large U.S. arms sales package is still pending. These delays continue to raise questions about Taiwan’s defense readiness.

AI-Fueled Surge

As the world’s chief producer of advanced semiconductors, Taiwan is benefiting hugely from the AI boom, arguably more than any other country. In the second quarter of the year, the Taiwanese economy grew at a torrid clip of almost 13 percent. This was the third consecutive quarter of double-digit expansion and the briskest quarterly rate since 1988. Taiwan’s Directorate General of Budget Accounting and Statistics (行政院主計總處) forecasts annual GDP growth this year to reach 9.6 percent, which would be the best performance since 2010 (National Statistics, Republic of China (Taiwan), accessed August 5).

Taiwan’s foundry-centric semiconductor manufacturing model, best encapsulated by the Taiwan Semiconductor Manufacturing Company (TSMC), has never been more effective. Fabless designers at the vanguard of AI hardware, including U.S.-based Nvidia, AMD, and Broadcom, all rely on TSMC to fabricate their most advanced chips. Their demand for AI chips led TSMC to post record profits for the fifth consecutive quarter in July, with net income of roughly $22 billion. The company is projecting revenue to grow 40 percent this year to more than $171 billion (Taiwan Semiconductor Manufacturing Corporation [TSMC], accessed August 5).

The semiconductor sector’s robust growth has coincided with Taiwan’s stock exchange reaching a new all-time high. After the TAIEX reached 48,218 points in late June, the market cooled somewhat in the following weeks, before rallying back to roughly 43,000 points in early August (Unique Business News, June 23; Commercial Times, August 4). [1] This has apparently spurred a new wave of retail investors in Taiwan, where the number of people with brokerage accounts rose to a record 14.6 million in July, with investors under 30 accounting for nearly two-thirds of the increase over June (Taiwan Stock Exchange [TWSE], accessed August 5).

Deepening Ties With the United States

In tandem with its robust economic growth, Taiwan has managed to quietly strengthen ties with the United States this year. Taipei has done so by focusing on strategic industries in line with Trump administration priorities and avoiding criticism of President Trump, who has prioritized U.S.–PRC relations.

For the Trump administration, Taiwan’s AI capabilities are among its most attractive assets. Taiwan is leveraging this competitive advantage effectively through both government-to-government initiatives and massive private investment. At the sixth Taiwan–U.S. Economic Prosperity Partnership Dialogue in February, the two countries agreed to a “strategic alignment for AI development” (AI發展戰略對接) (Taiwan Ministry of Digital Affairs, February 2). In July, TSMC announced it would invest an additional $100 billion in its Arizona production facilities, bringing its total investment to $265 billion (UDN, July 17).

Drones are another area in which the United States and Taiwan are enhancing cooperation, though with more modest results than AI. In June, Taiwan’s semi-governmental Industrial Technology Research Institute (ITRI; 工業技術研究院) joined the Green UAS certification program of the Washington-based nonprofit Association for Uncrewed Vehicle Systems International (AUVSI), which verifies supply chain and cybersecurity compliance for commercial drones. The certification will facilitate the entry of more Taiwanese drone makers into the U.S. market (Industrial Technology Research Institute, June 4).

Bipartisan congressional delegations have made three visits to Taiwan so far in 2026, signaling continuity in the bilateral relationship at a crucial time. Drones came to the fore during the most recent visit (July 30–August 1), which was led by Representative Young Kim (R-CA). This time, the context was Taiwan’s strained domestic politics and a stalled $6.6 billion drone procurement bill (Focus Taiwan, June 18). In a meeting with Taiwan President Lai Ching-te (賴清德), Kim noted the intensifying military threat the PRC poses to Taiwan and urged Taipei to pass necessary legislation to support its drone industry, which has been delayed more than six months. [2] “Deterrence takes two to tango, so help us help you,” she said (Office of the President of the Republic of China (Taiwan), July 31). After her meeting with Lai, Kim told reporters, “We hope that the Legislative Yuan can send a strong demand signal to build and buy as many drones as possible, and do it quickly” (Taipei Times, August 1).

The Art of the Delay

Representative Kim’s comments well embody the sentiment among U.S. lawmakers that Taiwan needs to move quicker on drone legislation. While congressional support for Taiwan remains strong, the inability of the ruling and opposition parties to find a modus vivendi on the issue is causing some strain in the bilateral relationship. Skeptics of the U.S. security commitment to Taiwan can point to the drone issue and argue that Taipei does not take its own defense seriously, and therefore is not worth risking American blood and treasure to defend.

At the same time, the Trump administration has so far declined to approve a new arms sale to Taiwan, for undisclosed reasons. The package, estimated at $14 billion, has been delayed since early 2026. President Trump described the package as a “very good negotiating chip for us” with the PRC during a May Fox News interview, heightening speculation that his administration may withhold it as a result of PRC pressure (Taipei Times, May 18).

Republican lawmakers, however, say that the White House will approve the arms sale before long. Representative Michael McCaul (R-TX), said at the Ketagalan Forum in Taipei on August 4 that he was “confident the president will forward soon another package to Congress with an additional $14 billion in critical military defense articles” (Focus Taiwan; Office of the President of the Republic of China (Taiwan), August 4).

Conclusion

Taiwan has shown impressive resilience in a challenging geopolitical environment. Its centrality to the AI industry has helped it avoid the fractiousness that has characterized the Trump administration’s approach to its other allies and partners. Strong congressional support suggests Trump will greenlight the $14 billion arms sale, though likely not before Chinese leader Xi Jinping’s expected September 24 state visit to Washington (Fox News, July 23).

All eyes will be carefully watching the progress of drone legislation through the Legislative Yuan in the meantime.

Unlike the arms sale, this is entirely under Taiwan’s control. Whether it can pass smoothly—and inclusive of sufficient financial support for the industry—will be taken as a signal about Taipei’s commitment to military modernization.

Notes

[1] The TAIEX stands for the Taiwan Stock Exchange Capitalization Weighted Stock Index. It is the primary stock market index for companies listed on the Taiwan Stock Exchange (TWSE).

[2] The “more than 6 months delay” refers to the drone procurement efforts that were cut as part of the original special defense budget proposed in late 2025 (China Brief, May 29). A new special defense budget bill targeting domestic drone procurement was proposed by Taiwan’s Ministry of National Defense in late June and remains stalled in the Legislative Yuan (Focus Taiwan, June 18).

The post Taiwan Shows Resilience in Trying Times appeared first on Jamestown.

Procurement Documents Reveal AI Chip Workarounds

Par : Jonah Reisboard

Executive Summary:

  • Procurement records suggest that U.S. export controls on frontier artificial intelligence (AI) chips imposed meaningful constraints until January 2026, as institutions in the People’s Republic of China’s (PRC) appeared to be adapting procurement practices to preserve access rather than replacing foreign hardware with domestic alternatives.
  • Tender documents from universities and state-linked entities show repeated efforts to obtain Nvidia H200-class computing power. Some mention H200s explicitly, others specify capabilities that can only refer to H200s, and others still obfuscate by appending politically acceptable labels such as “domestic chips” or “H20” to specifications that indicate they actually refer to H200s.
  • The persistence of these workarounds indicates that despite visible progress by domestic firms such as Huawei, PRC alternatives remain insufficient for the most demanding frontier AI workloads, making continued access to U.S. hardware strategically consequential.

At Huawei’s China Partner Conference 2026 (华为中国合作伙伴大会) in March, the company unveiled a new neural processing unit (NPU), the Atlas 350. This latest artificial intelligence (AI) accelerator card is powered by Huawei’s Ascend 950PR chip, which Huawei claims can deliver 2.87 times the computing power of Nvidia’s H20, marking a new stage in the commercialization of domestic AI compute (Guancha, March 21). The launch reinforced an official narrative that the country’s indigenous computing stack is maturing, and that a more self-reliant, high-performance domestic compute ecosystem is beginning to take shape (People’s Daily, March 11).

Public procurement records from before the U.S. government’s decision in early 2026 to lift its restriction on buyers in the People’s Republic of China (PRC) purchasing Nvidia’s H200 chips tell a different story. Despite aggressive promotion of domestic substitution from political leaders, in recent months Chinese institutions have sought to preserve access to Nvidia’s frontier hardware, indicating that U.S. export controls continue to impose meaningful constraints on PRC technology development.

Obfuscation Provides Workarounds to Procure H200s

A review of PRC tender documents reveals repeated efforts to acquire computing hardware with capabilities equal to Nvidia’s highest-performance tier, the H200. Some institutions embedded in the PRC’s broader AI and advanced computing research ecosystem, including Jilin University (吉林大学) and Zhejiang University of Technology (浙江工业大学), identify the chip explicitly in their procurement documents (Zhiliao Tender Information, December 16, 2025, November 11, 2025; Jilin University, accessed April 2; Chinese Association for Artificial Intelligence, April 23, 2025).

In other cases, tendering entities do not name Nvidia directly but instead specify technical requirements that mirror the H200’s performance profile. Specific references to benchmarks of 141 gigabytes (GB) of video random-access memory (VRAM)—the high-speed memory built into graphics processors that is particularly important for training AI models—are notable. This figure stands out because Nvidia’s earlier flagship H100 offered 80GB of VRAM and most PRC accelerators currently deployed remain below that threshold. A requirement for 141GB of VRAM functions as a strong indicator of H200-class hardware, even when the product itself is not named (Zhiliao Tender Information, December 17, 2025; Ascend Developer Community, January 30; Nvidia, accessed April 3; Metax Tech, accessed April 3; Cambricon, accessed April 3; Moore Threads, accessed April 3).

Some institutions employ a more subtle and politically adaptive strategy. Southern University of Science and Technology (SUSTech; 南方科技大学) and Jiangsu University (江苏大学), for instance, describe procurement needs using language more consistent with domestic policy preferences, such as referring to “domestic chips” (国产芯片). Henan University of Economics and Law (河南财经政法大学), meanwhile, specifies export-compliant alternatives in its procurement documents, such as the H20. In all cases, however, the documents list technical benchmarks associated with substantially more powerful systems (Zhiliao Tender Information, November 3, 2025; November 25, 2025; December 18, 2025). [1] Similar patterns have also appeared in tenders linked to China National Nuclear Corporation (中核集团), suggesting that this strategy of obfuscation is not limited to universities alone (Zhiliao Tender Information, October 31, 2025).

Adaptation Involves Legal Gray Areas

These procurement patterns highlight a tension in Beijing’s technology policy. Beijing has spent years promoting self-reliance in core technologies through policies such as “xinchuang” (信创), which pushes public sector users toward indigenous information technology products (Sina, June 20, 2025). When domestic substitutes cannot fully meet the requirements of frontier AI research, institutions appear to adapt by finding workarounds that preserve access to restricted foreign technology.

According to an interview conducted with two industry participants, compliance checks in the PRC often focus primarily on documentary consistency, such as whether contracts, invoices, and customs declarations match (Author interview, February 21). This creates room for ambiguity between politically acceptable descriptions and actual end use, allowing for plausible deniability from compliance teams and authorities.

Delivery timelines in some tenders also raise questions about the nature of the supply chain for high-performance AI hardware. The White House lifted its ban on the export of H200 chips on January 20, but the deadlines specified in some tender documents appear too short for newly sourced high-end chips to move through fully compliant export channels. This suggests that inventory is already circulating through underground networks and ready to be deployed once procurement procedures are completed.

Conclusion

The set of tender documents reviewed in this article suggest that the PRC’s AI ecosystem is still reliant on Nvidia chips. Although domestic firms, and Huawei in particular, have made visible progress in AI accelerator development, foreign chips apparently continue to be preferred for the most demanding workloads (21st Century Economy, September 18, 2025). If this were not the case, procurement practices would have shifted decisively toward cheaper, domestic alternatives where stable supply is assured. If PRC actors are yet to solve the frontier chip problem and instead are attempting to work around it, this suggests that the technological gap remains real and that controls on advanced semiconductors continue to impose meaningful constraints on the PRC’s technological progress, even if those constraints are uneven and imperfectly enforced.

For now, continued access to advanced U.S. chips can help PRC institutions maintain momentum until domestic alternatives mature. This fails to divert the PRC from its long-term push for self-sufficiency, and simply alleviates near-term constraints, buying time for PRC firms to develop without sacrificing progress at the technological frontier. This is critical, because time matter in AI development. More computing power enables larger models, faster iteration, and stronger research ecosystems, all of which can compound over time.

The PRC’s procurement behavior therefore offers a useful empirical window into the state of the technology competition between the two global leaders. It suggests that, notwithstanding official rhetoric about self-sufficiency, U.S. frontier chips remain valuable enough that institutions continue to find workarounds to secure access to them. This evidence reflects a continuing U.S.–PRC capability gap, as well as diverging realities between Chinese indigenization interests and industry needs.

Notes

[1] Southern University of Science and Technology participate in national defense-related research programs under PRC’s military-civil fusion framework (ASPI, accessed April 7).

The post Procurement Documents Reveal AI Chip Workarounds appeared first on Jamestown.

PRC’s Photonic Chip Push Signals Leapfrogging Moment

Par : Dennis Yang

Executive Summary:

  • The People’s Republic of China (PRC) has gone from a single pilot production line to a string of headline breakthroughs in photonics technology since 2024. Beijing has framed progress by researchers at Shanghai Jiao Tong University, Tsinghua University, Fudan University, and the Chinese Academy of Sciences (CAS) as a way around U.S. chip sanctions.
  • Photonic chips, which move information using light rather than electricity, are faster, run cooler, and—crucially for Beijing—do not depend on the cutting-edge factory equipment that the United States has blocked the PRC from buying.
  • PRC labs are at or near the global frontier in several photonic research benchmarks, but the United States and Taiwan still dominate the parts of the photonic supply chain that turn lab demonstrations into viable, scalable products.

On April 3, several major Chinese labs jointly announced “LightIN,” a new type of experimental photonic computer chip (Ying Zhu et al., March 14). [1] Photonic chips, unlike conventional chips, uses light to process information. This enables faster and more efficient data processing, as light travels faster than electricity, generates much less heat, and can carry multiple signals at once via different parts of the spectrum. Photonic chips therefore can move and process information at much higher speeds while consuming a fraction of the energy (Nature, December 23, 2015).

Within weeks of the research being published, state media in the People’s Republic of China (PRC) had spotlit LightIN. One government website declared that the breakthrough “injected strong ‘photonic power’ into the development of the digital economy” (为数字经济发展注入强劲的“光子动力”) (CCTV, April 3; Department of Economy and Information Technology of Hubei Province, April 6). Beyond the hype, the LighIN project is a reminder that PRC labs are shifting toward building general-purpose photonic computing platforms rather than just one-off experiments. Such breakthroughs can complicate U.S. approaches to technology controls, as they rely on technologies not currently addressed by existing policies.

Bypassing Bottlenecks for AI Advances

Photonic chips are particularly well suited to two jobs: shuttling enormous volumes of data between artificial intelligence (AI) processors in data centers and performing the kinds of repetitive mathematical calculations that drain the most power on conventional chips. Importantly, they can be fabricated without the most advanced factory equipment, as performance depends on how cleverly their light pathways are designed, not on cramming switches into ever-smaller spaces. Photonics therefore provides a potential solution to the obsolescence of Moore’s Law, which characterized the exponential growth of processing power for over 60 years (Synopsys, June 30, 2024).

Since 2015, the PRC has made photonic technology a national priority. The 14th Five-Year Plan designated it a strategic technology, directing funding toward national labs from agencies such as the Ministry of Science and Technology and the National Natural Science Foundation and establishing dedicated research and development programs. This allowed for long-term capability-building rather than immediate commercialization (China Brief, July 12, 2024). The possibility of “leapfrog development” in chip design has driven Beijing’s strategic obsession with photonics over the last decade. This has accelerated since U.S. restrictions on advanced chipmaking machines arrived in 2022, as photonic chips offer a way to sidestep the U.S.-imposed bottleneck by providing alternative efficiency gains through technological solutions that obviate the need for advanced equipment.

The technology has garnered top-level industry support. In June 2025, Huawei founder Ren Zhengfei (任正非) sat for a rare front-page interview with the People’s Daily. Ren acknowledged that the PRC is still “one generation behind the United States” (还是落后美国一代) in conventional chips, but argued that the country could close the gap by “compensating for physics with mathematics, for Moore’s Law with non-Moore approaches, and for single-chip limitations with group computing” (用数学补物理、非摩尔补摩尔，用群计算补单芯片) (People’s Daily, June 10, 2025). Photonics is the most concrete “non-Moore” bet on the table. Huawei underscores this through a decisive move, appointing of Dr. Martin Schell, a leading photonic expert from Germany, as Head of R&D at its Prague Research Center (Baijiahao, April 4). The pattern seen across other industries is now repeating in photonics. Government money has followed: a National Natural Science Foundation roadmap, a National Key Research and Development Program, and a new State Key Laboratory dedicated to photonic materials all sit behind the recent push.

Lab Breakthroughs yet to Translate to Applications

The PRC plausibly leads in photonic patents, as well as in certain published benchmarks. An early milestone came in 2024 with the arrival of the country’s first dedicated photonic chip pilot factory in Wuxi: The Chip Hub for Integrated Photonics Xplore (CHIPX; 光子芯片研究院). Run by Shanghai Jiao Tong University, in June 2025 it announced that it had produced the PRC’s first batch of six-inch photonic chip wafers and began mass production of high-speed chips operating at world-class speeds. State media framed the moment as a leap from “technology follower” (技术跟跑) to “industrial leader” (产业领跑) in high-end optical components (IT Home, June 6, 2025; People’s Daily, June 7, 2025).

Other universities have provided additional breakthroughs. A team at Tsinghua University has created two variants of a photonic chip, which they call Taichi (太极) I and II. Taichi II could train AI models using light alone, sidestepping the standard approach used for training Western models (Tsinghua, April 10, 2024; Tencent News, August 8, 2024; The China Academy, August 12, 2024). Chinese media even claimed Taichi II could perform with roughly 1,000 times more energy efficiency than Nvidia’s flagship H100 graphics chip, which should be further scrutinized (Baijiahao, September 2, 2024), More recently, researchers at Fudan University unveiled a photonic chip that they claimed could process data inside an AI cluster at significant speeds (IT Home, March 12, 2025; Science Net, March 13, 2025). A team at the Chinese Academy of Sciences (CAS) Shanghai Institute of Optics and Fine Mechanics (SIOM; 上海光机所), meanwhile, unveiled Meteor-1 (流星一号), advertised as the world’s first “ultra-high parallel” optical chip, capable of running calculations simultaneously on different colors of light (SIOM, June 17, 2025; CAS, June 25, 2025).

These breakthroughs put the PRC near the global frontier, yet laboratory achievements are yet to translate into viable products. They have no domestic customer queue and limited foundry production slots. By contrast, Nvidia has announced that it will start shipping photonic networking switches in 2026 that offers higher bandwidth and data throughput compared to the Fudan University team’s chip. This device is co-developed with TSMC and built into the same product family that already trains most of the world’s frontier AI models, and will likely find cloud buyers like longtime partners Microsoft and Meta (Nvidia, March 18, 2025; TrendForce, March19, 2025). Samsung, meanwhile, has unveiled a roadmap to fabricating photonics AI chips by 2029 (The Elec, March 30).

Conclusion

Photonics developments could complicate Washington’s strategy of denying Beijing access to the most advanced chip manufacturing technology. If Chinese researchers can extract similar performance from photonics chips as from traditional semiconductors, U.S. export controls will lose much of their efficacy. Any future restrictions would have to target a very different set of inputs, such as specialty crystals, lasers, and design software, that have not been studied in depth.

Two years ago, CHIPX was an aspiration; today, it is shipping wafers, and four other PRC institutions have produced credible photonic computing advances. A decade of state-driven efforts have begun to pay off. Beijing has not yet caught up with Nvidia or TSMC where it matters most—inside the world’s working AI data centers—but it is no longer obvious that the United States can keep the PRC permanently behind by controlling lithography and GPU technology alone. The next phase of the technology contest will be fought less in nanometers than in wavelengths of light.

Notes

[1] Zhu Ying, Liu Yifan, Yang Xinyu, Liu Kailai, Hua Xin, Luo Ming, Liu Jia, Chang Siyao, Yan Jie, Zhang Shengxiang, Wu Miao, Wang Zhicheng, Zhang Hongguang, Wang Dong, Chen Daigao, Xiao Xi, and Yu Shaohua. “LightIN: a versatile silicon-integrated photonic field programmable gate array with an intelligent configuration framework for next-generation AI clusters.” Light, science & applications. (2026) 15:165. 10.1038/s41377-026-02209-5.

The post PRC’s Photonic Chip Push Signals Leapfrogging Moment appeared first on Jamestown.

The Chip Packaging IP That Export Controls Cannot Reach

Par : Jonah Reisboard

Executive Summary:

  • Semiconductor packaging giant Tongfu Microelectronics holds a permanent, royalty-free license to intellectual property from U.S. fabless manufacturer AMD. This license was granted in 2016 with PRC state backing from the National Integrated Circuit Industry Investment Fund.
  • U.S. export controls on semiconductor packaging, introduced in 2025, cannot reach technology permanently transferred by contract nine years earlier.
  • The license permits the use of AMD’s technology for any customer at facilities in Suzhou, Zhejiang, and Penang, Malaysia, with no expiration, and survives termination of all other agreements.

During the Two Sessions on March 5, Science and Technology Minister Yin Hejun (阴和俊) repeated a claim made by both General Secretary Xi Jinping in his new year’s speech and by Premier Li Qiang (李强) in his government work report that the country’s chip development “has achieved new breakthroughs” (取得新突破) (Xinhua, December 31, 2025, March 5; S&T Daily, March 5). None stated where the breakthroughs had come, but a securities industry report published the same week identified packaging capacity, not fabrication, as the binding constraint on the deployment of artificial intelligence chips, arguing that “advanced packaging and testing will determine the upper limit of computing power supply” (先进封测将决定算力供给上限) (Fxbaogao, March 5).

Days later, Tongfu Microelectronics (通富微电子), one of the largest semiconductor packaging companies in the People’s Republic of China (PRC), confirmed on the Shenzhen Stock Exchange investor platform that it is “actively deploying Chiplet, 2D+, and other top-tier packaging technologies” (积极布局Chiplet、2D+等顶尖封装技术) (Weibo/Stockstar, March 12). The foundation for that deployment is a permanent, irrevocable, royalty-free license to access intellectual property from AMD, one of the largest U.S. fabless semiconductor manufacturers. Tongfu heralded the deal as one that was not just good for the firm itself but was “even more so an embodiment of national strategy and will” (更是国家战略和意志的体现) (Tongfu Micoelectronics, May 3, 2016). The license permits the use of AMD’s technology for any customer, at two facilities in Suzhou and Penang, with no expiration and no equity threshold.

As a result, Tongfu is beyond the reach of the current U.S. regulatory regime. The U.S. Bureau of Industry and Security (BIS) only released a whitelist of outsourced semiconductor assembly and test (OSAT) companies approved for handling advanced chips in early 2025 (Federal Register, accessed April 6). [1] As these controls are designed for the flow of items, they cannot address technology that has already been permanently transferred by contract.

Transaction Preceded Regulatory Focus on OSAT

In April 2016, AMD sold 85 percent of its only two semiconductor packaging facilities to affiliates of Tongfu for approximately $371 million. The buyer was backed by the PRC’s National Integrated Circuit Industry Investment Fund (国家集成电路产业投资基金), also known as the “Big Fund” (大基金), which provided roughly three quarters of the total acquisition capital through two purpose-built investment vehicles (Cninfo, December 28, 2016). [2] AMD retained 15 percent. The Committee on Foreign Investment in the United States (CFIUS) did not conduct a review of the transaction, as it was classified as a routine restructuring from integrated device manufacturer to fabless model (SEC, October 15, 2015).

At the time, semiconductor packaging was merely a mature, high-volume manufacturing process. This is no longer the case. Advanced packaging techniques now enable chiplet architectures, where multiple dies are integrated into a single package, and 2.5D/3D stacking, where chips and memory are layered vertically. AMD’s own MI300X accelerator, packaged by Taiwanese semiconductor foundry TSMC using advanced 2.5D techniques, illustrates the strategic value now embedded in the packaging layer. The IP that AMD licensed to Tongfu covers foundational processes upon which advanced packaging capabilities are built. Mastering these substrate-level processes is a prerequisite for scaling to the architectures that define current artificial intelligence (AI) chip packaging. The classification of this work as commodity labor was consistent with how Washington understood semiconductor packaging in 2016, but Beijing’s state fund did not share that assessment. As far back as 2011, the State Council had called for extending tax benefits to packaging to “complete the integrated circuit industry chain” (完善集成电路产业链) (State Council, January 28, 2011).

Licensing Deal Set Stage for Ongoing Entanglement

AMD’s filing with the Securities and Exchange Commission (SEC) omitted the IP license agreement from its exhibits, disclosing only that the company had granted the joint ventures a non-exclusive, perpetual, and irrevocable license to the technology owned and used by the company in its back-end processing business at existing facilities (SEC, October 15, 2015).

Tongfu’s response to the China Securities Regulatory Commission (CSRC), prepared by the law firm King & Wood Mallesons and reviewed by China Merchants Securities (招商证券), provided the operative detail. It describes the license as “non-exclusive, global, fully paid, free, non-transferable, permanent, and irrevocable” (非独占性的、全球性的、全额付清的、免费的、不可转让的、永久性的且不可撤销的). Its scope is facility-based—the joint ventures may use AMD’s technology solely for conducting business at the TF-AMD Suzhou and TF-AMD Penang facilities—but has no restrictions on customers, simply stating that the joint ventures will use AMD’s licensed technology “to provide services to AMD and other customers” (为AMD及其他客户提供服务) (Cninfo, December 28, 2016). [3]

Tongfu therefore can continue to license AMD’s IP, despite not appearing on the BIS whitelist. This is because U.S. regulations do not address whether CFIUS retrospective review authority extends to IP licenses granted before the 2018 Foreign Investment Risk Review Modernization Act (FIRRMA) reform or whether BIS should condition future OSAT approvals on the absence of irrevocable IP licenses from U.S. companies. As the Tongfu–AMD CSRC filing confirms, the license “continues in effect after the agreement expires or terminates” (在该等协议到期或终止后继续有效), even if AMD sells its entire remaining stake (Cninfo, December 28, 2016).

This could lead to U.S. technology being used to directly support the PRC’s national technology strategy. The CSRC filing names Huawei’s chip design subsidiary HiSilicon (华为海思) as an important customer of Tongfu, in a passage directly referencing the former AMD facilities. Whether Tongfu has used AMD’s licensed technology specifically in work for non-AMD customers has not been documented in any subsequent public filing reviewed for this analysis, and whether TF-AMD Penang directly packages Ascend chips is a question Tongfu has declined to answer on investor platforms; but the co-location of AMD-licensed packaging capacity and Huawei’s AI computing infrastructure in Malaysia raises questions about the relationship. TF-AMD Penang announced a Ringgit 2 billion ($454 million) new manufacturing site at Batu Kawan in 2022, bringing total Malaysian capacity past 2.3 million square feet (Invest Penang, June 2022). In May 2025, Malaysia also announced plans to host a Huawei Ascend AI deployment, although the announcement was retracted under U.S. pressure within 24 hours (Malay Mail, May 21, 2025). Malaysia is a useful third-country for PRC firms looking to avoid regulatory scrutiny, as it has no foreign direct investment security review mechanism—a fact documented in the PRC Ministry of Commerce’s investment guide and confirmed in the filing.

Conclusion

No provision in the BIS rule or in existing technology controls restricts how a non-approved OSAT firm uses previously licensed American packaging technology for non-AMD customers. The nine-year lag between the 2016 AMD–Tongfu license and Washington’s decision to recognize packaging as a strategic technology whose export was worth regulating speaks to a broader problem of failing to predict the future strategic significance of emerging technologies.

Any efforts to reverse AMD’s entanglement—such as by sanctioning Tongfu—can no longer be achieved without causing significant disruption to the U.S. firm’s product pipeline. Its purchases from the joint venture reached $1.5 billion in the first nine months of 2025, up 25 percent year-over-year (AMD, November 5, 2025). No alternative supplier can absorb this volume for AMD’s core processor lines.

Notes

[1] An authorization deadline, which went into effect on April 13, 2025, requires all shipments of advanced computing integrated circuits through non-approved packagers to obtain a license. No PRC OSAT firm appears on the whitelist. It should be noted that Tongfu does not appear on any U.S. export control list.

[2] Tongfu Microelectronics [通富微电子股份有限公司], Response to the “Notice of First-Round Feedback on the Review of Administrative Licensing Projects by the China Securities Regulatory Commission”
(No. [163440]) (Revised Draft) [通富微电子股份有限公司 关于《中国证监会行政许可项目审查一次反馈意见通知书》（[163440]号）之反馈意见回复（修订稿）], December 28, 2016.

[3] TF-AMD Penang is also ineligible to appear on the whitelist, as the framework applies an ultimate-parent test that bars entities whose parent is headquartered a “Country Group D:5 destination”, which includes the PRC (see 15 CFR §748.16(a)(1)).

The post The Chip Packaging IP That Export Controls Cannot Reach appeared first on Jamestown.

Strategic Snapshot: U.S.–PRC Tech Rivalry

Par : Dennis Yang

Under General Secretary Xi Jinping, the Chinese Communist Party (CCP) has made technology competition the focus of its long-term struggle against the West. The 15th Five-Year Plan (2026–2030) has codified the fusion of national planning, security strategy, and technological control under Xi’s direct command. The plan, whose early chapters are dedicated to building a modern industrial system, accelerating tech self-reliance and cultivating new productive forces, and deepening informatization and intelligentization, highlights a range of priority high-tech sectors. These include artificial intelligence (AI), quantum, biotechnology, connected devices and the Internet of Things (IoT), robotics, and open-source technologies, among others. The Party believes that seizing the “commanding heights” in these areas is critical to ensuring long-term dominance (Xinhua, March 13).

Xi has directed the Party-state system to advance a whole-of-nation approach to promoting strategic sectors. This is seen in the mechanisms embedded in the Party’s seven national development strategies, of which the military–civil fusion development strategy the most important, coordinating the system to diffuse innovation throughout the pillars of the Party’s power. It also involves boosting competitiveness through securing supply chains, promoting innovation, and ensuring that critical technologies remain in reliable hands. “Reverse constrainment,” or the selective weaponization of trade, has emerged as one tactic to adapt to the competitive international environment; but the Party deploys a variety of other tools in pursuit of supply chain sovereignty, technological dominance, and ultimately victory in international technological competition.

Looking forward to 2030, Xi is preparing the People’s Republic of China (PRC) for sustained confrontation. State investment, flexible regulatory environments, and courting foreign capital support reliable supply chains. “Command innovation,” or state investment in targeted sectors, serves the Party’s goal of global technological dominance. The CCP hopes that these strategies, combined with its weaponization of trade dependencies, will create the conditions necessary for victory in U.S.–PRC competition.

Selected Jamestown Analysis:

A humanoid robot, developed by Shenzhen-based UBTech, operating in a BYD factory. (Source: CnEVPost)

Strategic Industries

  • PRC robotics firms use Nvidia’s AI platforms to build their own products. This reduces innovation time as Chinese developers do not need to build foundational infrastructure from scratch. Leading Chinese AI models, such as DeepSeek’s, may have also been trained using Nvidia processors, as reflected by increases in shipments to Singapore following U.S. export controls. PRC robotics firms also import 90 percent of the machine tools they use. Most come from Japan, with Germany also featuring prominently. Rather than cultivating reliance, usage of Nvidia and other external materials feeds the PRC’s technological advancements, accelerating development.
  • Robotics and AI developments are intended to serve as revolutionary economic forces. The Robotics+ Action Plan incentivizes adoption in manufacturing, healthcare, logistics, and education, and the AI+ Action Plan encourages AI use in similar priority areas, including “global cooperation,” which alludes to the Party’s system of tightening its control both at home and abroad. Likewise, Party promotion of artificial general intelligence (AGI) is seen as creating “dynamism” for “new-type industrialization” (新型工业化).
  • In certain industries, such as smart devices, the PRC’s development directives have prioritized standards-setting. Setting technical standards gives Chinese technical ecosystems an advantage in the long-term, as changing existing standards is expensive, and this transition would impose higher costs on consumers. Digital Silk Road diplomacy supplements state-driven investment to export products and establish technical norms around the world.

Selected Jamestown Analysis: 

Members of the People’s Liberation Army in Tiananmen Square after the opening session of the Chinese People’s Political Consultative Conference (CPPCC). (Source: Kevin Frayer/Getty Images)

PLA and Military–Civil Fusion

  • Under the Party’s military-civil fusion development strategy, technological development is inseparable from military advancement. With Party guidance, the People’s Liberation Army (PLA) has started to adopt DeepSeek’s AI models, which are well-suited for use in unmanned systems. The PLA’s interest in AI applications also covers decision-making, intelligence, and cognitive and psychological warfare.
  • The robotics ecosystem is overseen by a new technical committee composed of military-industrial representatives and those affiliated with PLA-linked research centers. This committee is best understood as a strategic coordination mechanism which will promote industrial development while ensuring alignment with military interests.
  • Xi has promoted officials with military-industrial backgrounds to key roles in state economic planning and execution, ensuring that technological competitiveness serves military goals. In recent years, rhetoric around military-civil fusion has been extended from “innovation” toward “industry.” Xi’s personnel management formalized this shift.

Selected Jamestown Analysis: 

Customs officials inspect 306 CPUs that a traveler had attempted to smuggle into the People’s Republic of China. (Source: Gongbei Customs)

Technology Transfer and Global Partnerships

  • The PRC continues to exploit technology transfers to fuel its development but has diversified its approach in the wake of U.S. export controls. These tactics include a resurgence in smuggling, as well as a rise in trade diversion, talent attraction, and the use of offshore innovation bases.
  • Offshore innovation bases are hubs for science and technology innovation located across the PRC, which act as recruiting pools for international talent and as incubators for start-ups. They serve as a means to connect PRC researchers with global counterparts, which facilitates technology transfer. Similar risks emerge from PRC partnerships with overseas research institutions, where Chinese researchers have access to embargoed technology and resources, such as advanced semiconductors.
  • Regulated dual-use technology from the United States has been purchased by Chinese companies acting as intermediaries for military end-users. The transfer of U.S. military and dual-use technology to the PRC is also executed by attracting U.S. military-linked talent, which can be achieved legally and transparently.

Selected Jamestown Analysis: 

The opening ceremony of the 2024 Central Asia Digital Security Expo. (Source: CIAAE)

Censorship, Surveillance, and Information Manipulation

  • Consumer smart TVs provide a platform for consistent access to millions of homes in the United States. Hisense, a state-owned enterprise (SOE), controls the entire software stack of approximately 30 million TVs across 180 countries. At this scale, aggregated behavioral data provides a huge intelligence, surveillance, and manipulation asset.
  • Chinese private firms enhance surveillance and censorship capabilities for authoritarian countries abroad. The same company that built the architecture for Beijing’s “Great Firewall” is now operating on behalf of government clients across the Digital Silk Road, such as in Ethiopia, Pakistan, Myanmar, and Kazakhstan. The PRC also monitors these partners’ networks through hacking contractors, such as iS00n (安洵资讯).
  • Technological advancements also enable surveillance within the PRC’s borders. DeepSeek is seen as an effective tool to integrate and analyze data for public security functions, and the platform censors topics sensitive to the CCP. Beyond AI, technology firms have met in the Xinjiang Uyghur Autonomous Region (XUAR) for security industry expos to promote their products amid the CCP’s campaign of repression.

Selected Jamestown Analysis: 

The post Strategic Snapshot: U.S.–PRC Tech Rivalry appeared first on Jamestown.

❌