China has made an astonishing leap in artificial intelligence (AI), announcing the development of a large language model (LLM) with an astounding one trillion parameters—trained entirely on Chinese-made chips. This breakthrough by China Telecom, a state-backed telecommunications giant, signifies a crucial milestone in China’s efforts to achieve chip autonomy and develop AI technologies independently, without relying on Western technology. But beneath the surface of this impressive achievement lies a story not only of domestic innovation but also of stolen technology and growing threats to global stability.
China Telecom’s unnamed trillion-parameter model, along with another large language model called TeleChat2-115B, was trained on tens of thousands of chips produced domestically. This marks a turning point as China seeks to overcome U.S. restrictions on access to advanced semiconductors, such as Nvidia’s high-performance AI chips. As one of China’s largest state-owned carriers proudly stated, these models indicate that China has reached “total self-sufficiency in domestic LLM training.” This push toward independence comes in response to U.S. export restrictions, which have aimed to curtail China’s access to advanced technology by banning the sale of chips like Nvidia’s A100 and H100 to Chinese firms.
China’s progress in AI development, however, is not solely the result of independent innovation. There have been numerous instances of intellectual property theft from American companies, which have bolstered China’s technological capabilities. This ongoing theft, described by FBI Director Christopher Wray as a “shadow war,” has raised alarms among Western intelligence agencies and corporate executives alike. In recent years, China has been linked to a series of high-profile thefts involving cutting-edge AI technology, which experts fear could be used to further China’s geopolitical ambitions.
Innovation and Cooperation Amid Restrictions
To circumvent restrictions, Chinese tech firms are turning inwards for chip supply. ByteDance, the parent company of TikTok, has announced plans to train new AI models using Huawei’s Ascend 910B chips, signaling a move away from reliance on foreign hardware. This shift toward domestic resources is critical as the United States continues to tighten control over chip exports to China. ByteDance has been accelerating development using these Huawei chips, and insiders have confirmed that these models are becoming increasingly reliant on homegrown technology due to the lack of access to Nvidia’s premium graphics processing units.
Huawei, another major Chinese tech player, has been at the center of this development, supplying its Ascend chips to major Chinese firms for AI training purposes. Huawei’s AI training solutions are emerging as a popular alternative to Nvidia chips for Chinese companies trying to stay competitive in the global AI race. In addition to Huawei, other firms like Cambricon are also contributing to China’s goal of building a robust and independent AI ecosystem.
China’s Supercomputing Strength: Independent and Advanced
The unveiling of a fully Chinese-made supercomputer for AI training in Wuhan further underscores China’s determination to develop its technological capabilities independently. Built at the Central Intelligent Computing Center, this supercomputer reportedly has the capability to train AI models with trillions of parameters. With advanced liquid-cooling systems and a Power Usage Effectiveness (PUE) below 1.15, it reflects China’s remarkable progress in supercomputing efficiency and capability. While not officially benchmarked against the fastest supercomputers in the world, the capabilities of China’s systems rival those of the U.S. Frontier supercomputer, highlighting China’s push to compete on the global stage.
China’s ambitious supercomputing efforts, including the development of multiple computing centers with 5,000 petaflops of computing power, show a serious commitment to becoming a global leader in AI. In 2022, China’s OceanLight exascale supercomputer was even considered for the prestigious Gordon Bell Prize, demonstrating its growing status as a contender in the global supercomputing arena.
The Shadow of Stolen Technology
Despite the impressive progress China has made on its own, allegations of technology theft have consistently shadowed the country’s advancements in AI. The U.S. has repeatedly accused China of stealing intellectual property, seeing it as a deliberate strategy to leapfrog the West in the tech race. One example occurred in 2018, when former Apple engineer Xiaolang Zhang was arrested while attempting to board a flight to Beijing. He was caught with trade secrets related to Apple’s autonomous-driving program. This incident was part of what Wray described as a “continuing shadow war between the U.S. and China for supremacy in artificial intelligence.”
In another recent case, former Google engineer Linwei Ding was charged with stealing hundreds of confidential files related to AI while secretly working for two Chinese tech companies. Ding’s actions were seen as part of an ongoing campaign by China to acquire critical technologies to fuel its rapid growth in the AI sector. The stolen information included key details about the infrastructure and software needed for AI supercomputing—a crucial aspect of training large AI models.
The FBI and U.S. intelligence agencies are deeply concerned about the implications of such thefts. As Wray explained, “They [China] are working to use AI to improve their already-massive hacking operations using our own technology against us.” This sentiment was echoed by Brad Smith, president of Microsoft, who stated that China’s use of AI to analyze massive data sets stolen from U.S. companies and government institutions is already underway. “We should assume that AI will be used to continue to refine and improve targeting, among other things,” Smith warned.
The Broader Threat: AI as a Military Tool
The implications of China’s AI developments go far beyond the commercial and economic sectors—they present a potential national security threat. The integration of AI into military applications, combined with China’s growing ability to gather and analyze data using advanced AI models, poses a direct challenge to Western military and technological dominance. The U.S. intelligence community is concerned that China could use AI not just to boost its own military capabilities but also to undermine the security of other nations.
One of the most concerning aspects of China’s AI growth is its ability to use AI for mass surveillance and data analysis. Over the past decade, Chinese hackers have been linked to numerous major data breaches, including the theft of records from Marriott International, Equifax, and the U.S. Office of Personnel Management. With the power of AI, China could potentially cross-reference and analyze these vast data sets to identify U.S. spies, track government officials, and create detailed dossiers on American citizens. Glenn Gerstell, a former general counsel at the National Security Agency, warned that “China can harness AI to build a dossier on virtually every American, with details ranging from their health records to credit cards and from passport numbers to the names and addresses of their parents and children.”
Moving Forward: A Technological Tug-of-War
China’s rapid development in AI represents a major challenge for Western nations, not just in terms of economic competition but also in terms of national security and technological dominance. As China continues to leverage a combination of stolen intellectual property and domestic innovation, Western countries must respond by not only advancing their own AI capabilities but also bolstering security around critical technologies. FBI Director Wray and other leaders in the intelligence community have emphasized the need for vigilance to prevent further intellectual property theft and to protect critical American innovations.
The U.S. and its allies must also consider the broader implications of China’s AI capabilities. AI has the potential to change the balance of power in global politics, economics, and military strategy. As China moves forward with its AI ambitions, including trillion-parameter models and powerful supercomputers, the West faces an urgent need to develop a strategic response. This response must include stronger cybersecurity measures, technological advancements, and international cooperation to counterbalance the rise of China’s AI influence.
Ultimately, the race for AI supremacy is far from over. China’s trillion-parameter model, built entirely with domestic chips, is a testament to the country’s rapid progress and ambition. But this progress has come at a cost—both in terms of global trust and the heightened risk of international conflict.
At ACZ we have no doubt that the motives are sinister, likely both advanced military applications, predatory applications to dominate the market, and the ultimate control of information through propaganda and fake news, and surveillance of its customers and their productivity.