CALIFORNIA, USA — In May 2024, former Google CEO Eric Schmidt told Bloomberg that the United States held a “solid two-to-three-year advantage over China” in Artificial Intelligence development. He attributed this edge to factors like export bans on semiconductors, and supercomputing equipments.
Six months later, Schmidt’s perspective shifted dramatically. At a November 2024 forum hosted by the Harvard Institute of Politics, he expressed concern that the U.S. was “falling behind China in the race to develop more powerful artificial intelligence.”
This marked a stark reversal of his earlier position, and underscored growing apprehension about China’s accelerating progress in AI.
Last Christmas, barely a month after Schmidt’s notable shift in tone, Deepseek, China’s free open-source AI model, rocked the Silicon Valley tech industry.
The AI model outperformed the most powerful open-source AI in the market across a wide range of analytic fields—math, physics, chemistry, biology, coding, data analysis, creative writing, medical diagnostics, legal document retrieval, proprietary datasets, and more.
Deepseek has now become the most widely-used AI model in U.S. universities—and it’s poised to spread across the globe.
What’s truly shocking is that Deepseek was developed with just $5.6 million—compared to the massive investments made by U.S. companies.
OpenAI spends $5 billion annually on AI development, while Google allocates $51 billion for AI capital expenditures in 2024, Microsoft invests $13 billion for its stake in OpenAI, and Meta plans to invest $65 billion for its AI capabilities this year.
So what happened? Didn’t the U.S. ban exports of high-performance computing components like graphics processing units (GPUs), and AI accelerators to China? Didn’t the U.S. impose strict controls on dual-use technologies—those with both civilian and military applications—out of national security, geopolitical, and intellectual property concerns?
I offer two explanations on this issue:
First, there’s the timeless principle: Necessity is the mother of invention. Challenges often spur innovation, leading to breakthroughs that redefine our way of life. Whether driven by competition, crises, limited resources, or a thirst for progress, necessity remains one of the most powerful forces fueling technological advancement.
Second, there’s the belief that nothing is original, and people steal from anywhere that resonates with inspiration or fuels their imagination.
China, in particular, has demonstrated remarkable resourcefulness. Before the export ban, China had secured NVIDIA’s H100 and H800 GPUs.
For instance, Tsinghua University acquired 80 H100 chips in 2022, followed by two additional H100 chips in December 2023. Similarly, a lab within China’s Ministry of Industry & Information Technology obtained one H100 chip in the same period. Additionally, an unnamed entity within the People’s Liberation Army in Wuxi sought three A100 chips in October 2023, and one H100 chip in January 2024.
As for the NVIDIA H800—a modified version of the H100 designed to comply with U.S. export restrictions—while precise figures are unclear, it’s reported that ByteDance alone ordered approximately $1 billion worth of H800 GPUs in 2023.
Given this, it’s reasonable to assume that thousands of H800 GPUs have made their way to China. However, without official data, the exact numbers remain uncertain.
These instances point to China’s continued effort to acquire high-performance GPUs—often through underground markets and international intermediaries—highlighting a persistent drive to secure the resources necessary for technological advancement.
The new arms race has begun.
_____________________________________
Author’s email: [email protected]
