Did you know that the United States has held its position as a leader in artificial intelligence over China for nearly ten years? However, Alexandr Wang, the 28-year-old CEO of Scale AI—a startup valued at $13 billion—says this supremacy is now facing significant challenges due to a groundbreaking A.I. model introduced by the Chinese startup DeepSeek, which was launched just before Christmas. Wang shared his apprehensions regarding the diminishing gap between the two nations during an interview with CNBC on January 23.

This week, DeepSeek introduced its second A.I. model, demonstrating reasoning skills that rival those of top U.S. companies such as OpenAI. The swift progress made by this Chinese firm has not only captivated researchers but has also ignited discussions about the effectiveness of A.I. chip export controls intended to limit China’s access to the advanced GPUs essential for A.I. development.

Wang underscored the importance of the A.I. competition between the U.S. and China, calling it one of the most pressing issues of our era. In a notable full-page ad in The Washington Post, he urged the Trump administration to protect America’s technological advantages in light of China’s swift advancements in A.I. He advocated for increased investment in computing resources and data infrastructure, alongside an energy strategy to foster the A.I. revolution.

A self-made billionaire, Wang hails from Los Alamos, New Mexico, where his parents worked as physicists at the prestigious Los Alamos National Laboratory. After brief experiences at Addepar and Quora, he transitioned into the tech sector. His path led him to explore machine learning at MIT before founding Scale AI, a company focused on delivering expertly labeled data for A.I. training through contractual partnerships.

With a valuation of $13.8 billion and a clientele that includes the U.S. Department of Defense and OpenAI, Scale AI has firmly established itself in the A.I. ecosystem. Wang’s connections with OpenAI CEO Sam Altman, who he roomed with during the Covid-19 pandemic, significantly contributed to Scale AI’s expansion. The company was incubated within Y Combinator, the startup accelerator previously headed by Altman.

How does DeepSeek’s new model compare to leading U.S. A.I. systems?

Recently, Scale AI partnered with the Center for A.I. Safety to launch “Humanity’s Last Exam,” described as the most challenging benchmark test for A.I. systems. While current models have struggled to exceed a 10% success rate on this test, DeepSeek’s latest reasoning model, DeepSeek-R1, has emerged as a frontrunner. Wang disclosed to CNBC that this model’s performance is comparable to the leading American A.I. models.

Despite stringent regulations on GPU exports to China, DeepSeek claims to have achieved impressive outcomes using far fewer resources than their American counterparts. For example, the DeepSeek-V3 model, released in December, was trained on about 2,000 Nvidia A.I. chips, while Meta’s Llama 3.1 model required 16,000 GPUs for training.

Wang remains cautious, noting that Chinese labs might have access to more H100s, a type of Nvidia GPU not legally available in the country, than is commonly assumed. He hinted that DeepSeek may have access to around 50,000 H100s, a detail they cannot publicly confirm due to U.S. export regulations.

To drive the progress of cutting-edge A.I. models, Wang emphasized the necessity for enhanced computational resources and infrastructure in the U.S. He forecasts that the A.I. market could soar to $1 trillion once artificial general intelligence (A.G.I.) is achieved—a milestone he believes could be reached within the next two to four years. A.G.I. refers to systems that can operate as highly effective remote workers, a goal Wang envisions is on the horizon.