Compute, not data, is the new oil. As the world races to develop ever-larger artificial intelligence (AI) models, mastering the technology to design and manufacture advanced, small-node chips has become a key to winning the global technological race. As a testament to this dynamic, the United States government has introduced a range of restrictive measures, ranging from simple tariffs to strict export controls, in an attempt to limit China’s technological ambitions.
This puts China in a difficult position, as it has to balance between adequately resourcing its own industry and exporting valuable resources elsewhere. The rising power is struggling to reconcile its domestic actions with the professed grand ambition to expand its digital partnerships with the Global South through initiatives like the AI Capacity-Building Action Plan for Good and for All.
Its planning is further complicated by an ever-changing U.S. strategy. For example, in July, the Trump administration reversed a previous de facto ban on the sale of Nvidia’s H20 chips to China imposed in April. The chips themselves were specifically designed for the Chinese market to comply with the export rules implemented during the Biden administration. This shows how unpredictable the next U.S. move is, even if sometimes it brings China short-term relief.
Under such circumstances, despite its enthusiasm for reducing a dependency on an unreliable supplier (one that that has been suspected of installing backdoors into its chips) and for accelerating its progress in domestic chip development, China’s actual capabilities in advanced chipmaking and export viability remain at a bottleneck. But even if supply was not the limiting factor constraining China’s chip exports, demand might be.
Unlike China, with its well-developed AI infrastructure and abundant investments, the Global South has limited demand capacity. If China hopes eventually to export its chips, it has to first ensure that its recipients possess the requisite infrastructure and capacity to make meaningful use of them. Consequently, China may be more tempted to focus on infrastructure, training, and data to help the Global South build the foundation for AI development. This will also facilitate any future expansion of chip exports that ultimately consolidate China’s market position.
The Endless Chip Race
Commentators have often assessed China’s semiconductor capacities by examining its progress in manufacturing cutting-edge chips, which is significantly hindered by the lack of access to the right machinery and technology. Although the Semiconductor Manufacturing International Corporation (SMIC) is already producing and selling 7-nanometer chips and reportedly made some progress in developing 5-nanometer chips, SMIC and other companies like Xiaomi and Lenovo likely rely on foreign technology, including software tools from U.S. electronic design automation companies, which continue to account for 82 percent of China’s market. Export controls also prevent China from acquiring the latest extreme ultraviolet lithography equipment, forcing the use of older deep ultraviolet machines and less efficient methods like “double patterning” that can increase production costs by up to 50 percent.
All these suggest that China has to travel a significant distance to catch up with the West. In 2024, the best machinery China could produce is only capable of making chips that are 28 nanometers wide, whereas its competitors abroad are developing 2- and 3-nanometer nodes. Although Huawei reported breakthroughs in producing 14-nanometer software tools, there remains uncertainty about the veracity of these claims.
Chinese authorities have issued warnings to industry professionals against disclosing technical details to the media and external consultants, thereby avoiding further U.S. attention that could trigger further restrictions. Due to this secrecy, China’s ability to produce cutting-edge chips with its own capabilities is difficult to assess. Although China recently introduced a new AI-driven chip design platform, QiMeng, to close the gap in design capabilities and enhance its resilience against export controls, its actual impact remains to be seen.
While China historically struggled to showcase outstanding achievements in producing cutting-edge chips, it holds a more comfortable position in the realm of mature or legacy node (28-nanometer or larger) manufacturing capabilities. Despite the popular belief that “less is more” in chip fabrication, these chips are far from being obsolete. Instead, they remain important for markets including the industrial, aerospace, and defense sector. They are used in almost all electronic devices, from remote controls to refrigerators. Thanks to a carefully designed subsidy policy, China currently controls about 24 percent of global capacity for chips in the 50–180 nanometer range, a share that is projected to rise to 50 percent by 2030. This pronounced dominance is so strong that the U.S. Department of Commerce conducted a survey of 100 companies last year to assess their dependence on Chinese mature semiconductors.
Between Capacity and Demand
Given China’s uneven production capabilities across different tiers of chips, a key question arises: What types and tiers of chips is China focusing on in its digital export strategy? Undoubtedly, China aspires to be a technological leader, and eventually, a main exporter of technology. This would allow the rising power to enjoy the various economic and geopolitical benefits that come with being able to control the very technology that is used throughout the economy. However, given its relatively inferior position in producing advanced chips, will it be motivated to export larger-node chips instead? Moreover, will any other countries opt to import chips from China given alternative options?
While these questions are significant, they contain a hidden assumption. China would have to actively promote chip exports, just like how it is striving to become a leading exporter of other high-tech products, as other industrial policies like the Made in China 2025 suggest. The reality, however, paints a different picture.
Regardless of the node sizes, China has been cautious in scaling up exports of this critical asset, even for the mature-node ones. Around 80 percent of sales focus on the domestic market, and the value of its exports amounts to just 5.5 percent of its imports. Given the escalation of Sino-U.S. tensions, China is now attempting to push for even more domestic purchases of its own technology.
SMIC exemplifies this transformation, having transitioned from approximately 60 percent foreign customer production five years ago to nearly 80 percent domestic customer focus currently. Similarly, Huahong Group has oriented close to 80 percent of its production capacity toward domestic customers. This inward-facing production model contrasts sharply with other Chinese industrial sectors, such as photovoltaic manufacturing, where export markets constitute the primary target.
Meanwhile, China has launched initiatives like the Digital Silk Road to export its technology, placing a huge emphasis on helping the Global South to build its own AI capacity. However, explicit talk of chip exports is notably absent in the strategy, despite the Global South’s obvious compute deficit. This apparent contradiction raises questions about the commercial viability and prioritization of chip exports in China’s immediate technological strategy.
The success of China’s chip exports hinges on its ability to not just fulfill but also to generate the demand for chips of the countries it seeks to support. Demand is not just about the willingness to possess a given technology, but also about one’s ability to acquire the product. China itself has a strong domestic demand for compute. Over 300 generative AI service models are fully registered in China as of January 2025, and AI is creating a market that could reach $1.4 trillion in China over the next five years.
Perhaps unsurprisingly, none of the Global South countries feature prominently among China’s primary export markets for semiconductors. Several factors explain this absence: It remains unclear whether the Global South has sufficient demand for a meaningful AI capacity given the shortage of talent, infrastructure, capital, and technology needed to develop similar capacity. Even the more developed economies like India, Brazil, and Malaysia remain constrained by energy issues. Thus, even if China displayed greater enthusiasm about exporting its semiconductors, the Global South may not be able to make meaningful use of them unless their existing bottlenecks are alleviated, which could explain why chip exports are not the main emphasis of China’s official AI strategy.
Yet, this does not necessarily imply a devastating blow to China’s AI ambitions, nor does it mean China will remain lukewarm in becoming a chip exporter in the future. Rather it means that the priorities are shifting to a different area. At this moment, China is more incentivized to apply a “walk before you can run” strategy and focus on the first building block of AI capacity, namely infrastructure, training, and data. These aspects can be more aligned with the Global South’s national objectives and ensure their ability to harness the potential of AI. China also has a relative advantage in promoting strategic cooperation in these aspects, given its well-developed AI infrastructure for pooling resources like the National Unified Computing Power Network, idle computing capacity, and experience in catching up with the West. Thus, the export of chips is of secondary importance for now.
Add to this the fact that exports would come at the cost of not being able to supply domestic players with the much-needed resource. Current projections suggest that China’s domestic production capacity will remain insufficient to meet China’s own AI processing demands, thus limiting its ability to promote chip exports aggressively. Although the Made in China 2025 plan set a target for 40 percent (and later raised to 58 percent) of the chips used in China to be domestically produced by 2020, the country fell significantly short of this goal. It remains reliant on advanced chips produced abroad and foreign technologies in 2025, as evidenced by the planned installation of an additional 1.9 million Nvidia H20 or B20 GPUs by the end of this year.
That said, the tides may be turning soon. Projections indicate that by 2030, assuming all announced fabrication facilities in China become operational, domestic capacity could satisfy approximately 90 percent of Chinese demand for mature node semiconductors. This represents a significant increase from the 37 percent domestic demand coverage achieved in 2020, while also signalling growing feasibility for chip exports in the coming years, which are gradually rising in volume over the years (albeit at a slower pace compared to its competitors).
Meanwhile, the capacity-building efforts targeting the Global South also serve Chinese interests as they help to create the conditions for China to seed future demand for its technological exports, thus gaining a stronger foothold in the chip market and compensating for the increasing difficulty in accessing the Western markets.
Combined, these dynamics suggest that despite aggressive export controls targeting China, it continues to possess the potential to shape the global AI infrastructure from the ground up. There is room for China to seek leadership in the aspect of capacity-building and consolidate its relationships with the Global South, where countries are already resentful of the Western dominance of resources and equally eager to make use of AI to foster their development, so as to create the markets needed to fulfill its AI ambitions. While relatively insignificant right now, the importance of chip exports in China’s pursuit of AI leadership is likely to increase over time as the Global South gradually acquires the infrastructure and capacity required for sustainable AI development.
Can China Overcome the Chip Bottleneck?
Ultimately, the question of whether chip design and development represents a meaningful hurdle lies at the heart of any assessment of the feasibility and likelihood of Chinese success regarding their global AI ambitions. While China has a strong foothold in legacy markets and its genuine progress in chipmaking offers it some breathing room, there is a long way to go to secure a leadership position in advanced chip development.
Given the structural limitations facing many nations in the Global South, chip export is unlikely to be the most effective strategy to facilitate China’s pursuit of AI leadership in the short term. Although China will ultimately hope to establish leadership in the chip market, it may be more inclined to adopt a pragmatic strategy now, which emphasizes the export of infrastructure, training, and provision of capital as predecessors for future export plans.
Compute remains the new oil, but for now, China is not trying to provide the liquid gold itself. Instead, it is prioritizing the investment in building the oil fields, technical expertise, and pipelines for its partners.