By Hugo East
February 10, 2025

The Chinese company DeepSeek released its artificial intelligence (AI) chatbot to the U.S. market on Jan. 20. By the following week, it was the most downloaded app on the iOS App Store, surpassing Open AI’s ChatGPT.

The rapid rise of DeepSeek caused an unprecedented crash in the valuation of multiple U.S. tech companies, wiping out close to $1 trillion in combined market value from chip giant Nvidia Corp. and other peers. The loss to Nvidia was by far the largest, fastest devaluation of a U.S. company in history.

Socialist economic planning behind DeepSeek’s success

DeepSeek owes its efficacy to the socialist character of the People’s Republic of China (PRC), in which it was developed. The PRC’s economic central planning, through which it seeks to combine the advantages of strictly regulated capitalistic markets with state-owned enterprises designed for the benefit of the Chinese people, conforms to socialist methods of planning initiated by its first leader Mao Zedong. Socialist planning has enabled the PRC’s meteoric rise as a world power rivaling the U.S., as evidenced by the success of DeepSeek.

The latest iteration of that socialist planning is a ten-year initiative that began in 2015 called “Made in China 2025” (MIC 2025). In a report issued in 2017, the U.S. Chamber of Commerce said of MIC 2025: “Contrary to key elements of the Third Plenum Decision [PRC’s previous central economic plan], in which the Chinese leadership called for markets to play a decisive role in the allocation of resources across the economy, MIC 2025 instead appears to reaffirm the government’s central role in economic planning.”

MIC 2025 has in the final year of its implementation, through DeepSeek, seen the seeds it has planted bear fruit. A September 2024 study called the Critical Technology Tracker — published by the think tank Australian Strategic Policy Institute (ASPI) prior to the rise of China’s now-world-class AI sector — found that from 2003 to 2007 China led the world in just three out of the 64 critical technologies that the study tracked. From 2019 to 2023, however, China was the leading country in 57 of those 64 technologies, with the U.S. leading in the remaining seven — one of which in 2023 was AI.

The ASPI report states: “Some observers might argue that China’s ascendance into a research power — indeed the research power — doesn’t matter, because other countries, the U.S. in particular, remain ahead in commercialisation, design and manufacturing. That might be true for some technologies, but it represents a very short term attitude. China, too, is making enormous investments in its manufacturing capabilities, subsidising key industries and achieving technological breakthroughs that are catching the world by surprise.”

MIC 2025 fits squarely within China’s socialist economic development as first initiated by the Communist Party of China (CPC) in 1949 under the leadership of its then-leader Chairman Mao Zedong. In an interview with U.S. journalist Edgar Snow excerpted in “Red Star Over China,” Mao said, “When China really wins her independence, then legitimate foreign trading interests will enjoy more opportunities than ever before.

“The power of production and consumption of 450 million people is not a matter that can remain the exclusive interest of the Chinese, but one that must engage the many nations. Our millions of people, once really emancipated, with their great latent productive possibilities freed for creative activity in every field, can help improve the economy as well as raise the cultural level of the whole world.”

The PRC’s current leader, Xi Jinping, has carried on Mao’s plans by reigning in the “reform and opening-up” policies of the previous Deng Xiaoping government. Bloomberg reported in February 2024, in an article headlined “Xi Crackdown on ‘Hedonistic’ Bankers Fuels Industry Brain Drain,” that, “Indications are growing that Xi is shifting away from four decades of market-oriented reforms and financial innovation.”

Bloomberg further quoted University of Hong Kong professor, Zhiwu Chen, as saying, “I see a new Chinese economy coming into shape soon in which the financial sector will have only two types of players: government-run banks and government-run insurance companies. … While it will not totally go back to its pre-1978 planned-economy mode, it will be close.”

It is significant, given these reforms to incentivize workers to move away from finance jobs, that the owners of DeepSeek’s parent company, High-Flyer, came from the world of finance, as hedge fund managers. Their recent forays into AI perfectly align with the CPC’s goals for the Chinese economy, to guide China’s top minds into productive areas of the economy and away from the useless gambling and speculation markets that constitute so much of the financial sector.

These regulations imposed by the CPC come from its conformance with the long-term socialist project of the PRC and contrast sharply with the way U.S. companies overwhelmingly are operated. In a July 2024 interview, DeepSeek’s founder Liang Wenfeng made this fact plain: “Our principle is neither to sell at a loss nor to seek excessive profits. … This time, our goal isn’t quick profits but advancing the technological frontier to drive ecosystem growth.”

Chinese tech again innovates around U.S. trade embargo

Just like the PRC’s recent ascendency in automotive manufacturing, DeepSeek has found success despite the U.S.’s attempts to starve China’s AI industry of supposedly vital resources through a targeted trade embargo. The Biden administration created trade controls that the U.S. used to keep the hardware — mostly computer chips manufactured by the U.S. company Nvidia — out of China.

But, as the New York Times columnist Cade Metz said in a Jan. 27 article, “The impressive performance of the DeepSeek model raised questions about the unintended consequences of the American government’s trade restrictions. The controls have forced researchers in China to get creative with a wide range of tools that are freely available on the internet.”

A recent article in the British scientific journal Nature summarized the details of DeepSeek’s innovation. Wenda Li, an AI researcher at the University of Edinburgh, [Scotland,]  said: “Having limited computing power drove the firm to ‘innovate algorithmically.’ During reinforcement learning the team estimated the model’s progress at each stage, rather than evaluating it using a separate network.”

Mateja Jamnik, a computer scientist at the University of Cambridge, [England,] said: “This helped to reduce training and running costs. The researchers also used a ‘mixture-of-experts’ architecture, which allows the model to activate only the parts of itself that are relevant for each task.” (Jan. 23)

These innovations have put DeepSeek’s performance on a variety of tasks on par with the U.S.’s current leader, OpenAI’s ChatGPT — even outperforming it on some.

These algorithmic innovations also had the effect of making DeepSeek much less expensive, both in direct financial cost and in energy consumption. The Nature article reports: “DeepSeek hasn’t released the full cost of training R1, but it is charging people using its interface around one-thirtieth of what [OpenAI’s] o1 costs to run.

“The firm has also created mini ‘distilled’ versions of R1 to allow researchers with limited computing power to play with the model. An ‘experiment that cost more than £300 [US $370] with o1, cost less than $10 with R1,’ says Krenn. This is a dramatic difference which will certainly play a role in its future adoption.”

DeepSeek’s R1 and OpenAI’s o1 are types of language models used to solve complex problems and perform scientific reasoning.

U.S. market response symptomatic of capitalist flaws

When DeepSeek’s release became evident to the U.S. financial markets, the effects were devastating. In one day Nvidia lost $600 billion in its stock’s value, a record for a single company. The S&P market lost 1.5% of its value, and tech-heavy Nasdaq dropped a whopping 3.1%. The combined loss was nearly $1 trillion.

The scale of these losses is attributable to the central role that speculation plays in U.S. valuations. Venture capital firms have pumped billions of dollars into U.S. AI companies based on the assumption that U.S. AI would remain dominant in the world, and the companies’ valuations, determined by the stock market, were extremely high. OpenAI was valued at $157 billion prior to the release of DeepSeek.

The speculative value of the U.S. energy sector also fell precipitously due to DeepSeek. U.S. AI requires a tremendous amount of energy to develop and operate. DeepSeek requires far less because of the algorithmic innovation it employs to compensate for its relative lack of computer chips. Fighting Words, the journal of the Communist Workers League, said in a Jan. 29 article: “Since the DeepSeek app requires far less electricity than its U.S. counterparts, power companies, which had been preparing new projects for nuclear, fossil-fuel and some ‘green’ power plants took a beating as well. Power companies like Constellation Energy, GE Vernova and Vistra each plunged more than 20 percent.”

U.S. AI requires so much energy that, in October 2024, Microsoft entered into an agreement to rehabilitate the Three Mile Island nuclear plant, 100% of whose power would flow directly to Microsoft in order to support its AI projects. They are seeking a $1.6 billion federal loan for the project. Prior to the release of DeepSeek, it was estimated that by 2030, 17% of the U.S. energy output could be going to data centers used by tech companies to power AI.

DeepSeek is just one of several technological and scientific innovations developed under a socialist economy that challenges capitalist profits while benefiting the whole world.

To read the interview with the founder of DeepSeek, Liang Wenfeng, go to thechinaacademy.org/interview-with-deepseek-founder-were-done-following-its-time-to-lead.