If you follow artificial intelligence at all — and it’s virtually impossible not to if you pay attention to the news — you’ve undoubtedly heard of China’s DeepSeek and how it’s a “Sputnik moment” for America.
Not only does DeepSeek perform as well as its competitors, its creators claim, the AI model was developed using less-advanced processors and far less capital resources than OpenAI’s ChatGPT, Google’s Gemini, or X’s Grok.
It’s enough that President Donald Trump’s White House eliminated a set of executive orders by former President Joe Biden on AI which “established unnecessarily burdensome requirements for companies developing and deploying AI” and called “for departments and agencies to revise or rescind all policies, directives, regulations, orders, and other actions taken under the Biden AI order that are inconsistent with enhancing America’s leadership in AI.”
Given that AI is seen as the next big technological revolution, the comparison to Sputnik — the first Earth-orbiting satellite, launched by the Soviet Union in 1957, which started the space race in earnest — may not seem inappropriate, given the importance of which major industrial giant holds a technological edge in the sector.
However, the situation is a bit more complicated than it appears on the surface.
First, the facts: DeepSeek, a little-known Chinese startup, announced in late January that it had built its own AI model, DeepSeek-V3, using “only about $6 million in raw computing power,” according to The New York Times.
Furthermore, this was allegedly done with less chips than most AI systems use. “The world’s leading A.I. companies train their chatbots using supercomputers that use as many as 16,000 chips, if not more. DeepSeek’s engineers, on the other hand, said they needed only about 2,000 specialized computer chips from Nvidia,” the Times reported. The U.S. is working to limit how many chips, like those made by Nvidia, can be exported to China.
And, on its surface, DeepSeek-V3 seemed to be, indeed, a “Sputnik moment.”
“The DeepSeek chatbot answered questions, solved logic problems and wrote its own computer programs as capably as anything already on the market, according to the benchmark tests that American A.I. companies have been using,” the Times reported.
The Times declared DeepSeek “a win for China in the A.I. race,” although it noted that the Chinese Communist Party might end up stifling it. (For instance, don’t try asking it about Tiananmen Square.)
#DeepSeek is the Chinese state 🇨🇳Downloaded the app and put it to the #Tiananmen test and the #1962. Check the responses: pic.twitter.com/0CSaIFFYzR
— Sandeep (@SandeepUnnithan) January 28, 2025
There are deeper questions, however, beyond whether the CCP will censor DeepSeek’s wealth of knowledge. Other journalists, academics, and tech experts questioned both whether the $6 million budget was the only money DeepSeek received, whether it was truly independent from the CCP in terms of funding (as is claimed), and whether it had access to higher-level chips that aren’t supposed to be exported to China but made their way there anyway.
“It’s very much an open question whether DeepSeek’s claims can be taken at face value. The AI community will be digging into them, and we’ll find out,” said Pedro Domingos, professor emeritus of computer science and engineering at the University of Washington, in an interview with Al Jazeera.
He went on to say that while it was “plausible to me that they can train a model with $6 [million] … it’s also quite possible that that’s just the cost of fine-tuning and post-processing models that cost more, that DeepSeek couldn’t have done it without building on more expensive models by others.”
That raises the question of IP theft, something not unknown when dealing with China’s tech sector.
And some were more dismissive, including Palmer Luckey, founder of Oculus VR — who said the claims were clearly “bogus” and the result of “useful idiots” buying into “Chinese propaganda.”
DeepSeek is legitimately impressive, but the level of hysteria is an indictment of so many.
The $5M number is bogus. It is pushed by a Chinese hedge fund to slow investment in American AI startups, service their own shorts against American titans like Nvidia, and hide sanction…
— Palmer Luckey (@PalmerLuckey) January 28, 2025
“It is pushed by a Chinese hedge fund to slow investment in American AI startups, service their own shorts against American titans like Nvidia, and hide sanction evasion,” he said on social media.
“America is a fertile bed for psyops like this because our media apparatus hates our technology companies and wants to see President Trump fail.”
Beyond the funding numbers and the chips used, too, is how well DeepSeek performs. In a blog post, Pennsylvania-based Dolphin Studios said that even if the model operates efficiently on lower-level hardware, “this efficiency doesn’t necessarily translate into superior performance or unique capabilities.”
“This aspect is particularly appealing in regions where resources are limited or expensive. However, when it comes to innovation and pushing the boundaries of AI capabilities, DeepSeek falls short compared to giants like OpenAI and Google,” the post noted, adding that “this focus on affordability often means compromises in terms of research depth and technological advancement.”
Demis Hassabis, CEO of Google’s DeepMind AI subsidiary, was a little kinder but offered a similar assessment, saying that it “is probably the best work” he’s seen out of China but that it offered nothing new and that the United States was still ahead when it came to AI.
“Despite the hype, there’s no actual new scientific advance … it’s using known techniques [in AI],” he said, calling the hype “exaggerated a little bit,” according to CNBC.
Furthermore, DeepSeek — like most Chinese large-language model AI models — isn’t meant for higher-end tasks and doesn’t outperform U.S.-developed AI models at even more modest tasks.
As several researchers noted in the Harvard Business Review, “Chinese LLMs rely on less advanced hardware and initially focus on lower end — more specific, less general-purpose — applications that require less computational power.”
There’s a threat in this, mind you — there’s a disruptive advantage in having a cheaper model — but this all depends, again, on how much you trust DeepSeek’s numbers.
And while the big selling point on a global scale is efficiency — keep in mind that the high-end Nvidia chips that power OpenAI’s models and others are very power-hungry, meaning that AI could hit a very big wall when it comes to the energy capacity needed to scale it up further — this is all dependent on DeepSeek’s claims of using lower-end hardware and not using other models developed on higher-end systems.
Finally, it can’t be overstated that DeepSeek’s claims are just that: claims, and from a company without much of a history in the industry.
As The Times of London noted: “DeepSeek shocked the U.S., but intriguingly, it seems to have shocked China as well. ‘Before December 2024,’ noted ChinaTalk, an influential podcast on Chinese affairs, ‘DeepSeek was rarely mentioned in China’s AI community.’ It did not come from one of China’s tech giants — Baidu, Alibaba, Tencent, and Xiaomi — but from a side-project in a modestly successful financial institution.
“But just as Soviet advances in the space race often flattered to deceive, we should perhaps remain skeptical about developments in China. As the dust settles on DeepSeek, it may turn out that its achievements are not all that they seem.”
That said, President Trump is right to treat any advancement by our most serious geopolitical adversary as a major threat and cut Biden-era red tape on AI development. As for a “Sputnik moment” or a win for the Chinese. A wake-up call, yes. But one hopes America is already sufficiently woken up enough regarding the threat China poses — and unproven “Sputnik moments” don’t guide intelligent policy, either in AI or elsewhere.
Advertise with The Western Journal and reach millions of highly engaged readers, while supporting our work. Advertise Today.
