Just a week ago Donald Trump gathered some of his young new tech friends to announce the hubristically named ‘Stargate’: a colossal government-backed plan to spend $500 billion on AI and associated infrastructure.
This, Trump confidently announced, was “the largest AI infrastructure project by far in history”, the purpose of which was to keep ”the future of technology” in the US.
Pride comes before a fall, as they say. After his resounding victory, with capitalists lining up to ‘kiss the ring’, Trump felt he could walk on water. He quickly sought to set the world to rights in a show of force to China. Little did he or his tech bro chums know that China could so quickly change the game.
DeepSeek
DeepSeek, a Chinese tech startup that no one in the West had heard of until about two days ago, suddenly dropped an AI ‘large language model’ that instantaneously exploded all the received truths about AI and the US’ dominance of it.
Not only has China seemingly caught up with the US in this cutting-edge field already, but it has done so despite crushing technological restrictions that were based on the assumption that to train the most advanced AI, one needed to have the largest amount of the fastest chips in the world. It is reported that DeepSeek may have achieved its results on the basis of a mere $6 million investment, compared with the billions being poured into the US giants.
This is therefore a twin blow to the US tech sector and Trump’s expensive plans for it. China has caught up and overtaken the US in many important fields of technology, but there was one field, the most crucial one, in which the US maintained dominance: microprocessors and the AI that is based on them.
Trump and Biden have spent huge amounts of money and thrown all kinds of impediments at their competitors in the form of trade restrictions and tariffs, in an attempt to ‘build a moat’ around their technological advantage and hold China down.
If China can build equally good AI with fewer and older chips, then the US’ technological advantage, not only in AI but also in microprocessors, threatens to vanish.
What is interesting about the Stargate plan is that the US government has placed all of their eggs in the basket of OpenAI, who are the only partner of the project on the software side. Google, Meta, and all others were excluded. The government has placed a bet that this one company will continue to be the best in the field.
And yet OpenAI makes no profits. It has admitted that it loses money on every subscription to ‘ChatGPT Pro’, even though it costs $200 per month. It has yet to show how its services will ever make any money. They are popular, yes, but mainly because they are free and entertaining. It remains unclear how this technology is actually to be used to make money.
AI bubble bursts
The Financial Times points out that “The largest 10 stocks account for almost two-fifths of the S&P 500. Such concentration is unprecedented in modern times.” Eight of these, and all of the top seven, are tech companies (Nvidia, Apple, Google, etc). Although many of these companies are genuinely very profitable, the scale of the concentration of capital around them reflects the fact there is a huge AI bubble.
For this reason, the ‘China AI shock’ caused an enormous fall on the US stock market. Around $1 trillion was wiped out, although some of this has been recovered. Nvidia alone fell by around $600 billion, the biggest single loss in a day in US stock market history.
Nvidia’s CEO, Jensen Huang, saw his net worth drop by nearly $21 billion. The head of Oracle, also a participant in Stargate, lost $27.6 billion. Less well known tech infrastructure firms actually lost more, as a proportion of their value.
The question everyone is asking is: are the hundreds of billions of expenditures planned for AI in 2025 and beyond worth it? Have Chinese companies just shown that the entire model of this vast industry is bunk?
This sentiment is reinforced by the fact that it is not only DeepSeek that has suddenly achieved the unthinkable. Bytedance (the company behind TikTok), Alibaba, and the startups Moonshot and Zhipu have all recently released models that, in one way or another, challenge the superiority of American AI.
All of this expedites the already looming questions about the economic viability of ‘large language model’ AI, such as chatbots like ChatGPT.
There is no doubt that generative AI (the broader category of technology behind large language models) has enormous potential: it can be used to automate translation, to create new medicines, and make production of all kinds of things far more efficient.
But the money is being thrown after the large language models that generate so much attention. It is not clear how or even if these models can be improved much further, nor what economic use they really have.
These questions were already present. DeepSeek and others have posed it with far more urgency. What are these vast investments going to deliver, and are they even necessary?
It appears that the US has already lost its technological advantage. In such a scenario, it is entirely possible that OpenAI, whom Trump has hitched his wagon to so publicly and at such cost, could go bust.
DeepSeek has already become the number one free app on the Apple and Google app stores, ahead of ChatGPT. It can afford to offer much cheaper ‘pro’ subscriptions to businesses. What is to stop vast numbers of people and companies from deleting ChatGPT in favour of DeepSeek?
Protectionism backfires
Perhaps Trump is already cooking up a ban on DeepSeek’s app. But Biden’s CHIPS Act of 2022 was designed to prevent precisely this eventuality, by banning sales to China of high-end chips used for AI.
It would appear to have spectacularly backfired by forcing Chinese companies to innovate with cheaper and fewer chips. Thus, not only did it fail to prevent advanced Chinese AI, but it gave China an advantage in the form of cheaper, more efficient AI.
What this development shows above all else, is that it is impossible for the US to isolate and hold back Chinese capitalism. Chinese capitalism’s resources are too great. The world economy is too integrated, the technology it is based on is too complex and interdependent – the knowledge is developed and shared by too many people for an economy the size of China’s to be isolated and hobbled. Indeed, DeepSeek has made the technology behind its app open source, meaning anyone can have access to it.
It also reveals the colossal waste of capitalism in today’s epoch. Trillions of dollars are being spent on these AI ventures, and it may not even be necessary for achieving their goals, as DeepSeek has shown.
These goals (of super intelligent AI) may not even be realisable or desirable. The energy used in the data centres for this colossal effort are a significant setback to the goal of sustainability in energy. What we have is a clique of AI salesmen and con-artists hyping up a bubble for their own purposes.
The rapid development of Chinese AI is also an omen of how hard going, counterproductive and crisis-ridden Trump’s protectionism will be, both for the world and for the US.
Sooner rather than later, his aura will evaporate. There will be a lot of angry Americans who have come to realise that Trump cannot and will not ‘Make America Great Again’.
