(Credit: Alex Wong/Getty)

A few days before the Lunar New Year, and barely a week into the new Trump administration, China dropped a bombshell. With most of the West completely transfixed by the promise of AI, and stock market valuations now completely tethered to the technology’s hype, a Chinese company launched a product that revolutionises the industry’s entire cost structure.
Known as Deepseek, this open-source model was released to the entire world for free, with pretty much anyone able to run their own server. That’s surely a positive development in the tech landscape, yet, as the news hit, it caused shock, then consternation, then a growing sense of panic. The certainties of the tech oligarchy were being shaken. The floor had fallen out of the bottom of their new economic model. There’s no such thing as a free lunch, so the saying goes — but in a supreme bit of irony, the Chinese, by releasing a product that was in fact free, caused one of the most expensive stock market disasters in living memory.
Nvidia, which up until recently was mostly known for making graphics cards for computer gaming enthusiasts, lost a whopping $600 billion dollars in market capitalisation when the market opened on Monday. That’s the single biggest loss in human history. Nor was it the only company to suffer: stock market value far in excess of the entire Canadian economy was swiftly obliterated, all on the news that someone in China had managed to invent a slightly better chatbot. No wonder Sam Altman, CEO of OpenAI, was miffed that China had copied his homework.
In truth, though, the tech market’s travails are only one piece in a grander geopolitical chess game. One of the reasons that AI has come to dominate so much of Western discussion is that, in the minds of many, it represented the last, best field where the West (and America especially) maintained a big lead over China. Insecurity over China’s rising industrial might can be found in every corner of Washington. With America slipping in so many other areas of this new cold war, AI was supposed to be their one definite trump card.
Several factors helped build this narrative of Western AI dominance. To begin with, much of the technology and IP involved has traditionally been found inside the West, with China having to play catchup. In this regard, AI actually represents a sort of technological rearguard. After attempts to limit Chinese semiconductor manufacturing, and prevent them from dominating the telecom and 5G markets, production of graphics processing units (GPUs) remained an area where the US and its allies had a significant lead. In the final days of the Biden administration, then, America imposed severe export restrictions on these GPUs, conscious too that they’re the backbone of AI processing.
The problem, as it turns out, was that these limitations only made China stronger. Deepseek’s main innovation is that it’s made incredible strides in economising on GPUs and processing power, producing equivalent results to Western AI models at a fraction of the cost. This innovation, while in theory good for everyone, not only serves to level the playing field, but also undermines the West’s idea of how the future of AI would look: massive, hugely expensive data centres, soaking up more power than a small city, costing so much that only a few giants could afford to enter the market.
The other advantage the West was supposed to have here was innovation. In the common think-tank tale, exemplified in articles such as Harvard Business Review’s “Why China Can’t Innovate”, China was just too rigid, too controlled, too lacking in free thinking to ever invent something truly revolutionary. The Chinese could copy us, the story went, but true innovation was supposedly beyond their grasp.
With DeepSeek, that narrative is dead. Though it’s possible that it could have used some elements of OpenAI’s data or model in the process of developing its own AI, what is beyond dispute is the stunning innovation inherent in Deepseek’s ability to save massively on costs and hardware. Even if OpenAI were to simply copy Deepseek (which is quite easy, given that it is literally open source), the damage is already done: the viability of all these planned future infrastructure investments is now in doubt.
The Western habit of talking about China as some sort of problem child or basket case in terms of industrial innovation has become a form of self-harm at this point. In recent days, China has revolutionised the cost structure of AI, successfully carried out experiments in fusion power, launched a test flight for a new 6th-generation fighter aircraft, and launched new experimental satellites. By all accounts, the Chinese car industry is now overtaking the West; the time in which “made in China” was synonymous with cheap plastic junk is now long past. It should be controversial to say that a nation of 1.4 billion people, with a much higher proportion of Srem graduates than western countries, should eventually find various ways to put all that talent to work. The evidence of that is now all around us — what good does it do anyone to keep denying this reality?
For Trump, it isn’t only in the realm of geopolitics that China’s recent actions are a cause for worry. As the President enters his second week in office, the launch of Deepseek comes at an important time domestically. With the silicon valley billionaires now inside his tent, Trump has made a colossal bet on this de facto oligarchy, in the hope that they represented the economic and technological future of the country.
More to the point, Trump has made a point of embracing the promise that AI can save America, recently launching, with Altman, a new artificial intelligence firm. Stargate was meant to unify America’s biggest tech giants behind a $500 billion AI infrastructure investment programme — precisely the kind of infrastructure that Deepseek now threatens to make completely redundant. Trump’s alliance with other tech giants, notably Elon Musk, is also well known, even as the chief executive has filled important bureaucrat posts with figures from SpaceX and Palantir.
While that bet now looks far less safe, it also exposes another risky economic gamble. For decades, America’s industrialists have been increasingly loath to invest in factories, job training, or expanded production: it is simply more lucrative to plough revenues back into the stock market, buying up stocks and bolstering share prices. This has been thoroughly documented by heterodox intellectuals inside the American right, such as Julius Krein and former Mitt Romney campaign manager, Oren Cass.
Certainly, that’s clear from the numbers. All the way back in 2018, Goldman Sachs estimated that stock buybacks inside the US economy was at a whopping $1 trillion dollars annually. The situation has not improved since then. And right now, a small number of tech companies — called the Magnificent Seven — make up a massive share of the entire US stock market, with valuations that are clearly indicative of a very serious stock market bubble. If that bubble bursts, the fallout risks derailing the entirety of Trump’s second term in office.
To understand America’s challenge here, compare it to one of its rivals. Russia’s Gazprom has half a million workers and countless facilities across the world. Nvidia, by contrast, has less than 30,000 employees. In theory, Nvidia is more valuable than Gazprom — not because of the value of its factories or staff, but because everyone believed it would become valuable sometime in the future. Looking to China, meanwhile, Nvidia alone is worth over half the entire Shanghai stock exchange. Again, this “value” is only down to future expectations, derived from little more than hopes and dreams.
What can Trump do? Probably not much. The nature of stock-market bubbles is that they eventually pop. More generally, though, AI represents only the latest in a failed defensive line held against China. America tried to go after Huawei, and though sanctions and export controls did serious damage to the company initially, the telecoms giant adapted. Rather than crushing the Chinese telecom sector, indeed, America’s attempts to destroy the free trade system has only triggered more technological innovation. Semiconductors tell a similar tale, with China developing its own domestic chip industry at the expense of Taiwan. America keeps on getting beaten at its own game, only to stamp its foot in the unfairness of it.
But attempting to hold off the threat posed by a nation of 1.4 billion people, with an ancient culture and millions of STEM graduates, was always likely to fail. The problem facing the West today is not that it lacks the tools to “win” — it clearly does, given the ambition of its innovation — but that it doesn’t really know when the game should end. So all Washington’s political and economic energy is being channeled into increasingly expensive and desperate moonshots. AI was meant to be the trump card — until one day it was cloned.
In the end, of course, none of this is actually about AI itself. No matter how useful the technology, that’s distinct from the irrefutable logic of America’s financial bubble, a logic now more stark than before the Great Recession. If, or rather when, this bubble bursts, the political and social fallout will be immense. Obviously, the tech oligarchs are lashing out petulantly: China steals technology; China can’t really innovate; China is somehow “cheating” in the tech race — failing to see the irony that they too harvest data to build and improve their products. Essentially, though, the West’s problems have very little to do with Beijing, and more to do with our own systemic imbalances.
In tying his fate so tightly to the tech sector, the President, has made himself and the country he leads, hostage to a literally huge fortune. Consider Nvidia. Even after its recent troubles, the firm sits on a market capitalisation about the same size as the entire UK. Is this situation sustainable? No, it isn’t. The question then is what comes next; what happens if the AI bubble collapses under Trump’s watch? Rather than ushering in a new American renaissance, Trump may be the herald of a new American great depression. And just as the great depression ended up devouring Herbert Hoover, so could the AI bubble potentially take down Donald Trump.
Join the discussion
Join like minded readers that support our journalism by becoming a paid subscriber
To join the discussion in the comments, become a paid subscriber.
Join like minded readers that support our journalism, read unlimited articles and enjoy other subscriber-only benefits.
SubscribeI find it telling that so many Western news outlets are calling this a “Sputnik moment”. During the Great Leap Forward, Chinese newspapers were filled with astounding tales of tremendous industrial and scientific achievements on the part of the Chinese people, often far in excess of what the West during the 1960s was accomplishing and usually touted as having been performed at a fraction of the cost. This was credited to the inherent superiority of the Communist system and the efficiency and genius of the Chinese people. Such accomplishments ranged from constructing hydroelectric dams in only a few weeks to successfully crossbreeding tomatoes and cotton to get red cotton, and were referred to as “launching a Sputnik.”
All of these accomplishments, of course, were totally fraudulent, the purpose being to maintain China’s standing in the eyes of the outside world–and in the eyes of its own people–during a period of intense economic upheaval and mass famine caused by incompetent ideological government policies.
Now, I’m not saying that the DeepSeek breakthrough is necessarily fraudulent, but given China’s lack of transparency, the utter penetration of every aspect of economic life by the Communist government, its penchant for lying to foreigners for reasons of national aggrandizement, and its long history of misrepresenting (and in some cases, outright falsifying) scientific and technological “breakthroughs”, I will say that I am highly skeptical that DeepSeek is what its propagandists are claiming it to be.
In Silicon Valley, the demographic breakdown is as follows:
– Asian 38.9%
– White (Non-Hispanic) 28.7%
-Hispanic or Latino 25.2%
Within the Asian population, the major subgroups are:
– Chinese 30%
-Indian 16%
If the narrative insists that China is always copying, one must ask: who are they copying from?The answer—themselves.
Do not mistake intelligence for media figures, who are often financiers rather than true innovators.
We can’t avoid the embarrassment!
You seem to see people as entirely defined by their ethnicity. Presumably those people of Chinese ethnicity who live and work in the US mostly do so because they prefer US society and politics to the CCP alternative.
Yes, I’m looking at this through ethnicity rather than politics because innovation isn’t political—it’s part of human nature. Chinese people in America may be driving innovation and competing with Chinese people in China, but they remain Chinese by heritage, even if they are American by choice. When we say, “The Chinese in China are lying or cheating,” that’s political language. But when we say, “The Chinese in America are innovating,” we’re simply stating reality.
Politics often shapes interpretation for convenience, but reality speaks for itself. Many people dislike DEI (Diversity, Equity, and Inclusion), yet this situation ironically proves its value. Without Chinese talent in Silicon Valley, would we even have this level of competition? It’s similar to how people once dismissed Japanese cars, yet today, many drive them without hesitation.
The political narrative provides comfort when it suits people, but we have to acknowledge the facts. The Chinese aren’t just copying—they’re innovating. Credit should be given where it’s due.
By the way, ChatGPT has improved significantly since DeepSeek’s emergence. I pay for ChatGPT, so I can tell. They used to hold back certain capabilities, likely to encourage more spending over time. But now, with DeepSeek and China capable of releasing powerful open-source AI, OpenAI has been forced to step up. Competition is driving improvement.
As a country, we’re losing our competitive edge because of poor policies not because we are stupid, and that’s okay. China was fine when it wasn’t as powerful as we were, but now, it’s time to change the language and accept reality. We can’t keep closing our eyes to what’s happening.
Once more you bang on the money
You make many great points, but I am confused as to why you claim DEI somehow is helpful?
I read articles on here, Spectator etc showing how DEI discriminates against Asian students.
If recruitment is fair and based on set criteria, why would talented Chinese and Indians need DEI?
I only know uk IT market where jobs were mainly exported to India.
That was based purely on price.
Me and my colleagues were training Indians and not other way round.
I guess all really talented Indians go to USA?
Are there no Blacks in the Silicon Valley demographic?
If the software works as claimed it isn’t fraudulent. What may be fraudulent is the claim that it was produced so quickly by a small firm for only $6 million.
In the end it doesn’t really matter if they lied about the development costs if it can actually produce transactional costs st 50% of western models. That’s the important bit.
I am not claiming great knowledge of LLM, but question remains, as in the case of other technologies, whether Chinese really invented new version of it, enhanced it or stolen it from the West?
West is too placid in dealing with China.
Covid, under assertive West leadership, would be considered act of war by China.
West did nothing.
If West is to have any future, we need to disengage from China and reshore.
It would be painful.
Reality is that apart from the West there is not enough demand for Chinese goods and services because prospective buyers have no money.
Current West policy only suits top 1% and maybe another 15% of population as enablers of the top 1%.
The rest are loosing out.
I am old and childless, so not effected.
More tantrum – they nasty, they lie.
I’m tempted to agree with you – I suspect the holes in Deepseek will reveal themselves in due course.
What does worry me is that China is very obviously going for the same economic attack on the West that the West succeeded with against the USSR: destruction of the economic model through global competition both fair and foul. The USA finally succeeded in the late 1980s by managing to drive global oil prices down so much that the USSR ran out of money.
In a certain sense it doesn’t matter if China hasn’t really beaten the established Western AI firms. What matters is the perception that China can knock out something 90% as good for 0.1% of the cost. If that turns out to be generally true then the western economic model is in a lot of trouble.
You made me laugh.
There was no attack on Soviet economic model, unless you include sell of USA grain to Russia, so they would not starve due to idiotic nationalised agriculture.
Communist model collapsed because of its internal contradictions.
Low oil price was not a major factor.
Laugh away, be my guest.
I forgive because you know not of what thy Speak
…
Deepseek is open source. If (and it’s a big if), it was really created cheaply it actually opens up the AI field to a host of smaller businesses who otherwise couldn’t afford the $600m+ for model training. Shorts cuts and efficiencies will also get borrowed by all the other AI companies which, in principle, should reduce their development time and potentially allow them to increase the size of their models.
All this creates more competition, and more AI models, in a market where investors had baked in that a handful of American ‘Big Guys’ were in an unassailable monopoly/cartel position due to the high entry costs, with huge, unstoppable demand. Competition means tougher profits and more investment needed to keep on top of the market and make the winners more difficult to predict, reducing company value and so shareprice – thought I can’t see how Nvidia would be negatively affected here, as it all means more AI, not less.
More AI, but less need for high energy consuming GPU s that Nvidia makes. The current US model is very energy intensive, needs huge amounts of electricity to power data centers, and energy infrastructure to develop. Deepseek is only the beginning of that process , if AI is going to succeed it needs to become much more energy efficient, there are more huge surprises coming which will upend everything. The Chinese are just getting going,
Utter nonsense
It was achieved very cheaply
And Important with regards seeking Knowledge it’s vital that to obtain such it’s essential you know exactly what ,where , when and How you ask any important questions of China
Fail to do so and the Answer given
Shall mean little to you
Therefore knowing such
I posed the simplest of Question to Deep Seek as to how was this constructed
And being open sourced you shall be suprised at the Tremendous open highly detailed honest and
Complete answer
The details of which I’ll shall not go into hear
Other than I was given a detailed response that can only be described as a Construction Manual as to how to do so
Ah but like I said unless you know what and how to ask but more important Who
Fail to do so with China then thou
Shall be left in a state of Ignorance
Go Study as opposed to Go compare
The magnificent 7 is about to become the nightmare 7. Nvidia stock price target 15$ , back where it was 2 years ago. Of course it’s a bubble and it’s going to crash. Like Tesla, target price 100$, where it should be.
“If that bubble bursts…“.
I think you mean “When that bubble bursts…”
I remember the 1960s, when Japanese products were regarded as cheap inferior copies – British motorcyle devotees talked about “Jap Crap”.
That didn’t last long…
Do we really want to continue to support and promote the economical model we are currently pursuing for the long run?? has its served the people in the ‘ ‘western world’. (life expectancy is reducing more chronic illness, more mental illness, less people benefiting from the wealth we have built up and resources becoming scarcer (water, energy, metals etc etc) . Maybe there is a difference in what we want for ourselves and what we would like to see in society? Resilience = more actors involved (more small business), = also more people benefitting, and poss less people making money from money which only benefits a few, what would you choose…?
I’m using DeepSeek alongside ChatGPT and a few other models for a project right now, and it seems to me that DeepSeek and ChatGPT answers are uncannily similar, almost like a copy ;-). Maybe DS did job the ChatGPT code and that’s bad for their shareholders but the rest of us are going to benefit from better, cheaper, energy-efficient AI.
The simple ugly truth is that mother is the necessity of invention and hunger is a great driving force. If the US continues to try to legislate the problem away it will run into issues, the solution is instead make it easier to be a producer to make things, and produce, and that is done not by paying people more, that is done by making goods cheaper. It’s also why the EU is never going to be able to catch up to the US or China and will eventually be overtaken by India, because in the guarantee of safety the EU has made it almost impossible to produce.
At last a article by a Author who’s mind is open and albeit a simple basic comprehension of China but still lacking the Basics of Fundamental understanding of Who and what China is
I’m sick to the back teeth of telling all the considerable work and study
Required to get to grips with being able to Comprehend China
So what this article does not inform you of and it’ll take a wee while
Before any MSM Western based source
Shall tell you
And what you are about to learn has even more serious amplyfing
Consequences for Western Neo Liberal Capitalism
So fasten your seatbelt
Deep Seek not only used Nvidia chips but also a completely new
Huwaei Chip
That was absolutely essential in achieving what Deep Seek created
Especially so in LLM ( Large Language Model )
As Norm Macdonald said in a chat show with Bill Maher sometime in the 90s: The Chinese are much cleverer than we are
I wish someone would explain AI to me; my BS detector is going off in “Danger, Danger…” mode.
So far, my own experience is that AI is a flattener and simplifier of information (and art, too. Why do so many images of young women suddenly have such large breasts?)
As far as information goes, one gets a summary of the standard received wisdoms from the previous few decades. That’s no help to me. That is NOT the road to innovation; it’s more like sitting in a parking lot.
But what if I needed to fold some proteins or scale some quantum networks? Or even just get a workable definition of ‘quantum’ as it’s used these days? This is what I really want to know.
We don’t need any more flattening (except maybe some of those breasts). And we’re already drowning in received wisdom. We could use more subtlety. Maybe even an glimpse at something sublime, even if it can’t be put into words in a tiny fraction of a second.
On one side, this is bigger than any administration. It’s about decadence in the West where people try to get ahead by becoming the financial investment fad, by regulatory capture, by pop culture commercialism, but the deeper foundations of prosperity are fading: cultural confidence, serious education, a personal commitment to the future.
On the other side, weren’t we all supposed to be overtaken and bought out by the Japanese 25 years ago??