From Geopolitics to Chips: China's Mission for Technological Self-Sufficiency in the 15th Five-Year Plan

How US Restrictions Catalyzed Huawei and the CCP's Circumvention Strategy

Chinese chip manufacturers are implementing a strategy of ingenious innovation to circumvent American restrictions. In fact, the "microchip war," triggered by US restrictions starting in 2018, did not block the Chinese semiconductor industry but catalyzed its strategic response: innovation by circumvention. This approach is now officially enshrined as a national mission in the Communique of the Fourth Plenary Session of the 20^Central Committee of the CCP (October 2025), which set as a primary objective for the 5^Five-Year Plan (2026-2030) the "significant improvement of self-reliance and high-level scientific and technological self-strengthening" and the construction of a modern industrial system with advanced manufacturing as its backbone.

China aims to build an entire self-sufficient Artificial Intelligence (AI) technological stack, a vision perfectly aligned with Huawei's "Intelligent World 2035" roadmap, which forecasts a 100,000-fold increase in computing power and the evolution of AI into ubiquitous systems capable of controlling critical infrastructure and cognitive warfare. The strategy compensates for the gap in high-end chip quality with quantity, systemic efficiency (CloudMatrix 384 and optical networking), and hardware-software optimization. Although EUV lithography (ASML) remains the "Achilles' heel," the indigenization of equipment (SME) is advancing (NAURA, Hwatsing, Silan Microelectronics), with the goal of "tackling core key technologies."

China's ambitions to achieve 70% self-sufficiency by 2028 are accelerated by Huawei's role as a strategic protagonist in Beijing's "civil-military fusion" policy, with the company directly involved in drafting state regulatory frameworks like the "AI Safety Governance Framework 2.0." Statements by Nvidia CEO, Jensen Huang, who called the restrictions a "mistake" after seeing the company’s market share in China zero out, highlight the tension between US security policies and global market pressures—a tension reflected in the continuous search for alliances (Nvidia with Korean HBM suppliers) and China's determination to build a complete technological alternative.

GettyImages

The Geoeconomic Context: Embargo as a Strategic Catalyst and the CCP’s New Direction

The trade and technological tension between the United States and China has crystallized into the "microchip war," which began with the first export restrictions during the Trump administration and intensified under Biden. This high-tech embargo is not merely a trade conflict but an attempt to limit China's ambitions to build an advanced and self-sufficient chip industry, with clear implications for national security and global technological dominance.

The main effect of these restrictions, however, has been to spur China rather than slow it down. The Chinese leadership hopes that its companies can replicate in hardware what has been achieved in software: innovating by circumventing imposed limits. The example of DeepSeek, which released a competitive AI model using only a fraction of its Western rivals' computing power, is proof that engineering and optimization can compensate for brute force.

Huawei’s Strategic Vision: Intelligent World 2035

This forced innovation strategy finds its ultimate expression in Huawei’s long-term vision, formalized in the "Intelligent World 2035" roadmap. Huawei's evolution from a telecommunications giant to a global protagonist in Artificial Intelligence and advanced computing is not a purely commercial phenomenon, but a strategic component of Beijing's "civil-military fusion" policy.

The "Intelligent World 2035" reports are not mere technological forecasts but a genuine geoeconomic doctrine aimed at creating a global technological superstructure. The technological trends identified by Huawei are not just commercial goals but potential tools for strengthening state power. These include:

1. General Artificial Intelligence (AGI): AGI that "enters the physical world" is conceived as an entity capable of interacting with and controlling critical infrastructure, including autonomous command systems and military robotics.

2. AI Agents: The evolution of simple tools into "decision-making partners" translates into intelligence systems and potential autonomous weapon systems capable of autonomously analyzing massive data streams.

3. New Computing Paradigms: The forecast of a 100,000-fold increase in computing power by 2035 is the foundation for modeling conflict scenarios, cryptography, decryption, and operating large-scale geosystems.

4. Internet of Intelligent Entities: The transition to an omnipresent and omniscient network, if built on Huawei's technology, would grant strategic control over the global digital infrastructure.

Huawei’s Structural Role in State Governance

Huawei's participation is structural. The company is directly included among the entities that contributed to the drafting of the "AI Safety Governance Framework 2.0," a document prepared under the guidance of the Cyberspace Administration of China. This framework is not limited to technical issues but explicitly outlines risks related to a dual-use agenda, both military and commercial, recognizing AI's potential for the proliferation of nuclear, biological, and chemical weapons and for cognitive warfare, including interference in other countries' internal affairs via social bots.

This model, which blurs the line between corporate profit and state objectives, is cemented by China’s National Intelligence Law (NIL), which obliges every organization to cooperate with state intelligence apparatuses. This structural threat justifies US sanctions and global bans on 5G networks, as the competition for technological dominance translates into a direct confrontation for control of future digital infrastructure. The Framework 2.0 itself openly denounces that "certain countries" are using "unilateral coercive measures, such as technological barriers and export controls," to "maliciously harm the global AI supply chain."

The Mandate of the 15^ Five-Year Plan

The innovation strategy is now officially sanctioned by the Communique of the Fourth Plenary Session of the $20^{\text{th}}$ Central Committee of the Communist Party of China, held in Beijing on October 23, 2025. The Communique not only fully confirmed the line of the Political Bureau and General Secretary Xi Jinping but also approved the "Proposal of the CCP Central Committee on the formulation of the $15^{\text{th}}$ Five-Year Plan (2026-2030)."

The 15^ Five-Year Plan is defined as a "critical period for consolidating foundations and fully realizing socialist modernization." Among the main objectives outlined is the commitment to "achieve significant results in high-quality development" and to "significantly improve self-reliance in the scientific and technological field," strengthening "original innovation and tackling core key technologies." This mission is central to accelerating the construction of a new development model and promoting the great rejuvenation of the Chinese nation through Chinese-style modernization.

The Chip Quality Gap and Forced "Multi-Patterning"

The starting point of the competition lies in the chips themselves. Data indicate that the average performance of Chinese AI chips stands around 114 teraflops, significantly lower than American competitors. Huawei's flagship chip, the Ascend 910C, offers 800 teraflops, a value dwarfed by the 2,500 teraflops of the B200, Nvidia's high-end product.

The main cause of this gap lies in the difficulty of production of chips with minuscule transistors, a process that has represented the most reliable way to increase speed for the last fifty years. The ability to produce chips with state-of-the-art transistors is limited to a few global companies: Samsung Semiconductor, TSMC, and, in part, Intel. Due to US pressure, the most advanced TSMC foundries are precluded to Chinese customers, forcing them to turn to domestic manufacturers like SMIC (partly state-owned) and production plants managed directly by Huawei.

Lithography: The EUV Wall and the DUV Ruse

Chip factories rely on advanced machine tools for lithography, which use light to etch circuits. The global industry relies on extreme ultraviolet light (EUV) with a 13.5-nanometer wavelength, a technology exclusively produced by the Dutch company ASML. Due to restrictions, ASML cannot sell these machines to Chinese manufacturers.

The only option for manufacturers like SMIC is to push the old "deep ultraviolet" (DUV) systems, which use 193 nm light, to their limits. The adopted tactic is "multi-patterning": instead of exposing a wafer to the light source only once, the process is repeated multiple times to etch finer details.

While multi-patterning is a brilliant engineering ruse, it comes with costs: it increases costs, drastically slows down production, and reduces yield (the percentage of defect-free chips). For China, self-sufficiency prevails over efficiency, but most analysts agree that DUV has insurmountable physical limits for the large-scale production of the most advanced chips. Consequently, dominance in lithography remains China's Achilles' heel, as highlighted by CSET, with the national champion SMEE holding only a modest 4% of the i-line lithography market for legacy chips.

The System Enhancement Strategy: Quantity over Quality

Being behind on individual chip quality, China is implementing a strategy based on quantity and system architecture. AI arithmetic is ideal for "parallelization," where a task is divided and handled simultaneously by multiple processors. This strategy aligns perfectly with the CCP's mandate to accelerate high-level scientific and technological self-reliance and self-strengthening to guide the development of new quality productive forces.

Processor Clusters and CloudMatrix

The most striking example of this strategy is Huawei’s CloudMatrix 384, an artificial intelligence system designed for data centers. This system connects 384 Ascend 910C chips and aims to compete with Nvidia’s GB200 NVL72, which uses 72 B200 chips.

Analysis indicates that an Ascend chip offers about one-third of a B200's performance. However, by using five times the number of chips, the Huawei system is capable of offering slightly less than double the performance of the Nvidia system.

The crucial trade-off is energy consumption: the Huawei system consumes 600 kW of electricity, over four times more than the Nvidia machine. However, as analysts point out, energy is not considered a significant constraint in China, making this an economically reasonable compromise for the goal of self-sufficiency.

Optical Network and Integration

The success of this strategy depends on the ability to effectively connect all these chips. This is a historical strength for Huawei, given its experience in computer networks. The CloudMatrix 384 leverages optical networking, transferring data via light pulses rather than electricity. This approach produces less waste heat and consumes less energy, and it is increasingly common in data centers. This focus on system integration and optical networking is "fundamentally changing" AI infrastructure, allowing China to overcome the limitations of the single component.

Innovation in Hardware-Software and the Adaptive Numerical Format

The final and perhaps most sophisticated stage of the Chinese strategy is the tight adaptation between hardware and software. This co-optimization aims to extract every ounce of performance from the available hardware and falls under the effort to promote the deep integration between scientific and technological innovation and industrial innovation.

Numerical Precision and Energy Optimization

Traditionally, general-purpose processors use 32 or 64 bits to represent numbers, ensuring high precision but increasing energy consumption. However, AI models can tolerate some imprecision in calculations. For this reason, many modern AI chips use 16, 8, or even 4 bits, a "simple but very effective" way to optimize hardware, as every bit less reduces the number of transistors and electricity consumption.

DeepSeek has become a benchmark for Chinese AI in this direction. The company released a new eight-bit numbering format optimized for AI: it stores numbers without distinguishing between positive and negative and without fractional components. Although this scheme limits range and precision, it promises much higher energy efficiency.

The industrial impact is immediate: shares of Cambricon Technologies, a Chinese chip company, rose after the announcement, as its processors already support this format. While Huawei hardware does not yet support it, it is likely to do so soon, consolidating the native integration between the Chinese AI software ecosystem and national hardware.

Equipment Indigenization and Production Expansion

Parallel to the development of chips and software, China is conducting an offensive for the indigenization of semiconductor equipment (SME), a fundamental step to ensure the resilience of its supply chain, in line with the need to consolidate and strengthen the foundations of the real economy.

Sectoral Progress and National Champions

The CSET report documented China's notable gains in several key segments between 2019 and 2024:

• Chemical-Mechanical Planarization (CMP): Hwatsing Technology led a significant increase, consolidating China's position in this critical process.

• Dry Etch and Deposition: Companies like NAURA and AMEC have strengthened their market share. In particular, China gained ground in dry etch, a crucial technique for more advanced process nodes. In deposition, expansion was driven by NAURA, although a gap remains in the most sophisticated segment of atomic layer deposition (ALD).

• Advanced Packaging (ATP): China is also gaining ground in the final assembly and testing phases, crucial for the era of chiplets and AI. Advances in advanced packaging are particularly important for the future integration of heterogeneous components. Topej (TopKing) and Maivei are developing machines for Hybrid Bonding, a very high-density interconnection technology, marking an important step in the indigenization of key equipment.

• EDA and Design: EDA providers like Huada Empyrean (with the Storm platform for Organic RDL) and Xpeedic (SI/PI simulation) are developing platforms essential for 3D design.

Expansion in Analog Manufacturing and Industrial Strategy

The goal of self-sufficiency concerns not only advanced logic chips but the entire value chain. Proof of this is the recent move by Silan Microelectronics (October 18-19), which, through its subsidiary Silan Jihua and in partnership with the Xiamen government, announced a 20 billion yuan project for a 12-inch analog integrated circuit line. This massive investment aims to achieve a production capacity of 45,000 wafers per month, strengthening domestic production in mature but high-value sectors, essential for the automotive, energy, and consumer electronics industries. This expansion fits perfectly into the Plenary's mandate to accelerate the construction of a strong manufacturing nation and a nation of quality, maintaining a reasonable proportion of manufacturing output.

The Race for AI Self-Sufficiency and Geopolitical Dynamics

The race for AI autonomy is led by Huawei, which coordinates over 2,000 companies for 70% self-sufficiency by 2028. The optimism expressed by founder Ren Zhengfei reflects confidence in "recent breakthroughs" and the integrated system being built.

The Contradiction of Restrictions and the Nvidia Case

Geoeconomic dynamics are complex and often contradictory. Statements by Nvidia CEO, Jensen Huang, illustrate the paradox of the restrictions. Huang called the US technology blockade a "mistake," revealing that the company's market share in high-end chips in China has plummeted from 95% to 0% due to export controls. Although Nvidia obtained the green light to sell the underpowered H20 chip, the Chinese government responded by urging domestic companies to completely abandon Nvidia products and use domestic alternatives, demonstrating that even minimal US flexibility is not seen as relief but as an incentive to redouble efforts.

Nvidia’s Global Alignment

While Nvidia seeks to navigate the restrictions by selling underpowered versions of its chips in China, the company is actively committed to strengthening the global ecosystem on which it depends. Its participation in the APEC summit in South Korea (October 28-31) and scheduled meetings with the heads of Samsung Electronics and SK Hynix underscore its strategic dependence on the supply of high-bandwidth memory (HBM), an essential component for AI.

Conclusions: The Ecosystem Challenge and the Objectives of the 15^ Plan

The Chinese attempt to build a national AI technological stack has made undeniable progress, shifting the focus from individual chip efficiency to system power and software optimization. This is now fully aligned with the CCP's mandate to strengthen original innovation and tackle core key technologies and to coordinate development and security.

However, the road is still long:

1. EUV Lithography: This remains the critical "bottleneck." Without access to ASML machines, China is limited to costly, low-yield techniques like DUV multi-patterning.

2. Software Ecosystem: Despite DeepSeek and others creating new numerical formats, the global industry still largely relies on Nvidia’s CUDA AI programming tools. The Chinese software ecosystem is still weak and not globally standardized.

3. Design Tools (EDA): Chinese designers still heavily rely on American EDA programs (Synopsys, Cadence).

In summary, the Chinese industry, under the threat of sanctions, has responded with ingenuity and unprecedented investment. The strategy is clear: build a comprehensively competitive system by the end of the 15^ Five-Year Plan (2030) and achieve a "significant leap forward in economic, scientific, and technological strength" by 2035. For Beijing, the goal is not only to beat the United States but to ensure national security and technological autonomy, a goal that "simply competitive" hardware is already capable of serving. The Fourth Plenary Session transformed necessity into a binding political mission, making technological decoupling a pillar of Chinese-style modernization.