The Quiet Arms rACE
In the summer of 2017, China released a document outlining its ambition to develop the next generation of artificial intelligence. Eight years later, those ambitions are coming to fruition.
The first weeks of 2025 have seen China’s AI efforts dominate global headlines— from the 12-hour TikTok ban to the market-disrupting release of Deepseek-v3, its meteoric rise to the top of the Apple charts, and the ensuing fallout over allegations that it was trained using OpenAI’s proprietary data.
These high-profile events are merely the surface of a much deeper and more consequential reality: a silent but intense arms race between the West and China to define the technological landscape of the modern era.
Rebounding from the Chip Ban
In October 2022, the U.S. Commerce Department imposed restrictions on exporting advanced microchips with military and AI applications, aiming to curb China’s technological advancements. At the time, China was already several years into its stated plan to dominate the AI industry.
The effectiveness of this move is now under scrutiny. Breakthroughs like Deepseek-v3 and Tencent’s Hunyuan-Large have prompted industry leaders, including former Google CEO Eric Schmidt, to express shock at China’s progress.
“This is shocking to me,” Schmidt remarked. “I thought the restrictions we placed on chips would keep them back.”
Stockpiling and smuggling may explain why the chip ban has failed to stall China’s AI development. However, the broader implication is more significant: U.S. policymakers appear reactive and disorganized, while China’s tech sector advances with methodical precision.
The TikTok Factor
A key directive from China’s 2017 New Generation Artificial Intelligence Development Plan emphasizes AI’s role in understanding and shaping public cognition. It aims to enable authorities to “grasp group cognition and psychological changes in a timely manner; [ ] take the initiative in decision-making and reactions;[ ] and significantly enhance social governance.”
Just one year before this plan was unveiled, ByteDance launched TikTok, a social media juggernaut that by 2022 had been downloaded 3 billion times and, by 2024, boasted 170 million U.S. users spending an average of 51 minutes per day on the app.
TikTok’s core strength lies in its algorithm, which delivers a precisely tailored content stream to each user. A 2021 New York Times investigation into an internal ByteDance document, “TikTok Algo 101,” revealed how the platform optimizes engagement.
The document describes TikTok’s ultimate goal as maximizing user retention and time spent on the app. It provides an equation for how videos are scored, factoring in user interactions such as likes, comments, and playtime:
Plike X Vlike + Pcomment X Vcomment + Eplaytime X Vplaytime + Pplay X Vplay
In essence, the app creates a mathematical response to human boredom, ensuring users remain engaged. While this design is effective for entertainment, it aligns seamlessly with the goals outlined in China’s AI development strategy—leveraging cognitive insights to reinforce social stability.
Another Failed Ban
On January 19, 2025, TikTok preemptively shut down ahead of a scheduled U.S. ban. However, just 12 hours later, as former President Donald Trump prepared for his inauguration, the app was abruptly reinstated. Both the removal and reinstatement where punctuated by notifications sent to all users which praised the incoming President for the ongoing availability of the app.
Despite being removed from major U.S. app stores, TikTok remains widely available, with pre-installed phones fetching nearly $1 million on eBay. The continued ability to use the app hinges on ByteDance either selling or partnering with a U.S. company to maintain domestic operations. “Every rich person has called me,” Trump quipped, referencing the flood of interest in acquiring the platform.
Among the most intriguing potential buyers is Perplexity AI, a company specializing in generative AI. The prospect of merging TikTok’s behavioral analytics with a GPT-powered AI model carries profound implications, potentially reshaping the internet into an even more personalized, yet unregulated, information ecosystem.
Deepseek Upends the AI Market
Amid the TikTok drama, Chinese AI firm Deepseek AI launched Deepseek-v3, an ultra-efficient large language model (LLM) trained using a fraction of the computing power required by its U.S. competitors.
The release sent shockwaves through both the AI industry and financial markets. While OpenAI accused Deepseek of illegally distilling its training data, legal scholars quickly raised another red flag: privacy concerns.
Deepseek’s terms of service state:
“DeepSeek’s privacy policy gives DeepSeek broad rights to exploit user data collected through prompts or user devices, including by monitoring interactions, analyzing usage patterns, and training its technology. All personal data is stored on servers in China.”
These terms effectively render the model unsuitable for industrial use. No major corporation would entrust its proprietary data to a service with such open-ended data-sharing policies. Even individual consumers have begun to reconsider their use of Deepseek, though the app remains atop the App Store rankings.
Despite the privacy controversy, Deepseek’s efficiency raises another crucial issue: the role of microchip scarcity in shaping AI innovation.
Arvind Ravikumar, co-director of the Energy Emissions Modeling and Data Lab at the University of Texas, explains:
“There was not a need to think about efficiency; you could just add more chips. What this means is that there are significant unrealized efficiency gains in AI computing that could reduce energy consumption.”
If access to hardware defined the first front of the AI arms race, energy access will define the next.
Microsoft Goes Nuclear
In 2028, the Three Mile Island nuclear site—home to the worst nuclear disaster in U.S. history—will reopen to power Microsoft’s AI data centers. This project will create thousands of jobs and inject billions into Pennsylvania’s economy, a key swing state.
Microsoft is not alone. In late 2024, Amazon and Google secured similar deals, laying the groundwork for Project Stargate, a sweeping initiative to develop nuclear-powered AI infrastructure.
Mere days into his presidency, Trump announced a $500 billion federal investment in AI infrastructure. Flanked by tech titans—including SoftBank’s Masayoshi Son, OpenAI’s Sam Altman, and Oracle’s Larry Ellison—he framed the initiative as a historic leap forward.
“This will be the most important project of this era,” declared Altman.
Ellison added that Oracle had already begun construction on ten hyperscale data centers, which could transform digital healthcare by enabling rapid disease treatment and vaccine development.
In tandem, Trump rescinded Biden-era AI regulations, paving the way for an industry-led approach. Project Stargate envisions a vast, interconnected network of nuclear-powered AI data centers—an unprecedented technological shift, unrestrained by regulation.
This investment comes nearly a decade after China’s own AI development strategy was set in motion.
Key Takeaways
2025 is the year the quiet AI arms race will go public. The contrast is stark: U.S. policymakers appear fragmented and reactionary, while China advances with centralized efficiency.
China’s head start in AI investment—combined with its ability to bypass ineffective U.S. restrictions—has given it a competitive edge. Meanwhile, the lack of AI regulation in the U.S. raises critical questions, particularly as nuclear power plants are handed over to private tech companies.
Will the growing dominance of Chinese AI products force the U.S. to impose stricter domestic and international regulations around the use of AI technology? TikTok has already dodged one ban despite concerns over its societal impact. Deepseek’s privacy issues have hurt consumer trust, but will that scrutiny extend beyond individual users and spark broader legislative action? Watch this space.
As the world watches this unfolding battle for AI supremacy, one thing is clear: the stakes have never been higher.
About this Article
As a graduate of the University of Missouri School of Journalism, I understand the value of strong editorial oversight. While I crafted the initial draft of this article, I recognize that refining complex narratives benefits from a meticulous editing process.
To enhance clarity, cohesion, and overall readability, I collaborated with The Editorial Eye, a ChatGPT-based AI designed to function as a newspaper editor. According to the tool, its refinements aimed to “enhance readability, strengthen argument flow, and polish phrasing while preserving the original intent.”
However, the editing did not stop there. After reviewing the AI-assisted revisions, I conducted a final pass to ensure the article accurately reflected my voice and intent. The AI did not generate new ideas or content; rather, it helped refine my original work.
What you see here is the product of a thoughtful collaboration between human insight and AI-driven editorial support.