{"id":29158,"date":"2026-05-21T19:29:16","date_gmt":"2026-05-21T16:29:16","guid":{"rendered":"https:\/\/www.liberalglobe.com\/?p=29158"},"modified":"2026-05-21T19:29:16","modified_gmt":"2026-05-21T16:29:16","slug":"the-us-lost-the-%ce%b1%ce%b9-race-when-you-restrict-a-country-that-has-factories-it-develops-what-you-deprive-it-of","status":"publish","type":"post","link":"https:\/\/www.liberalglobe.com\/?p=29158","title":{"rendered":"The US lost the “\u0391\u0399 race”: When you restrict a country that has factories, it develops what you deprive it of"},"content":{"rendered":"\n

In recent years, technology has not only evolved; it has evolved exponentially. And in the midst of this evolution, the US-China relationship has ceased to be a simple competition and has become a historic power shift.<\/p>\n\n\n\n

DeepSeek was not just an AI model. It was the bell that signaled that the era of America designing and China building was over, and not in the way Washington expected!<\/p>\n\n\n\n

Just last year, in January 2025, a Chinese company that few outside of China knew about presented an open-source artificial intelligence model, trained with less than six million dollars (!!!). As a direct result, in a single stock market session, Nvidia lost 600 billion dollars in capitalization!!!<\/p>\n\n\n\n

The model was called DeepSeek V3, and its reasoning version, R1, matched or outperformed the best American systems. Sam Altman had said that training such a model cost \u201cfar more\u201d than $100 million. But DeepSeek published a confirmation in Nature that R1 cost $294,000. The difference wasn\u2019t just impressive; it was insulting to the American narrative.<\/p>\n\n\n\n

Most panicked people at the time were talking about the \u201cAI race\u201d and whether American companies had burned through billions and whether export controls were working.<\/p>\n\n\n\n

No one saw the big picture: DeepSeek\u2019s engineers didn\u2019t have access to Nvidia\u2019s top-of-the-line chips. American restrictions had locked China out of the H100 and A100 since 2022. What emerged was not a delay, but a generation of engineers who learned to do wonders with second-rate hardware.<\/p>\n\n\n\n

Jensen Huang admitted as much when DeepSeek V4, trained in part on Huawei\u2019s Ascend processors, appeared: \u201cThe best AI researchers in the world, because they are limited in computing power, are developing extremely smart algorithms.\u201d The day he had feared, the day when a top-of-the-line model would run first on Chinese hardware, had arrived.<\/p>\n\n\n\n

Huawei\u2019s chip was made by SMIC, with minerals that had been refined in China. Export controls that were supposed to slow Chinese progress ended up accelerating the vertical integration of the Chinese technology chain. Washington believed it held the levers of pressure. It was, after all, pushing China toward self-sufficiency.<\/p>\n\n\n\n

American manufacturing had peaked in 1979 with nearly 20 million jobs. By 2023, there were about 13 million left, while its share of GDP had fallen from 25% to just over 10%.<\/p>\n\n\n\n

For decades, the idea was that the real value was in design, not production. The factory was seen as a burden to be moved elsewhere.<\/p>\n\n\n\n

NAFTA sent some to Mexico; China\u2019s entry into the WTO in 2001 sent the rest. Between 1991 and 2007, Chinese imports to the United States increased by 1,156%. Economist David Autor has calculated that Chinese competition explains about 55% of the decline in American manufacturing employment. What was not accounted for was that a country that designs but does not produce depends on those that do produce. And when those countries learn to design for themselves, the advantage disappears.<\/p>\n\n\n\n

China\u2019s industrial rise was not limited to AI.<\/p>\n\n\n\n