What does Deepseek tell us about the relationship between the tech oligarchy and market competition?
Yanis Varoufakis is a Greek economist and politician. A former academic, he served as the Greek Minister of Finance from January to July 2015 under Prime Minister Alexis Tsipras. He has been Secretary-General of MeRA25, the political party he founded in 2018.
Cross-posted from ScheerPost
Hello, this is Yanis Varoufakis on behalf of DiEM25 with an end of month review, the month being January 2025. You have heard all about DeepSeek, the Chinese artificial intelligence AI company, which, this week, changed the global AI landscape, not to mention the one trillion dollar losses it caused in the New York Stock Exchange and NASDAQ. DeepSeek is China’s response to OpenAI’s ChatGPT-4. Its models perform as efficiently as their American counterparts. The difference is that DeepSeek is offered to you, to everyone, for free. How do they make money? Only by selling services to developers. And that, compared to OpenAI’s services, GPT-4, 5, at a fraction of OpenAI’s prices.
The gist of DeepSeek’s arrival on the AI scene and the carnage in the American stock exchanges is a sudden transition from proprietary to open source technology. It is therefore no great wonder that the moment DeepSeek became the most downloaded app on the Apple Store, it pulverized the market capitalization of the hitherto overinflated US Big Tech stocks. But how did this happen exactly? How is it that a private commodified service is suddenly offered for free? And does this mean that technofeudalism is in trouble?
To begin with, it’s important to note that AI was never a proprietary technology in itself. The underlying code of all AI companies was always open source. What made American AI a quasi-private commodity was the way in which these models were trained using huge amounts of privatized data. Where I say privatized, you should translate stolen data, your data, my data. There was a Google memo that was leaked in 2017 that was widely discussed and refuted, but it was a harbinger for what happened with DeepSeek.
In that memo, we read the following words: If an open source large language module, it said, trained for a few million dollars, comes to outperform a proprietary model, then there’s going to be trouble. There will be no firewall, the memo continues, even to safeguard open AI. That’s what happened. DeepSeek pierced the United States’ AI company’s bubble by decommodifying the results of the model’s training and doing it at a tiny, tiny cost to itself, shifting the results of AI-trained models from behind a paywall to the public realm. Within days since the release of the latest version of DeepSeek, developers around the world started building their own models on top of DeepSeek’s.
This was the nightmare of American Big Tech AI service providers who have been offering the results of prompts as a commodity in the form of subscriptions. You see, DeepSeek-type applications can now produce high quality translations for free. That’s just an example. And in so doing, they undermine the business model of companies like DeepL, the German company. In the broader scheme of things, this means that the morsels of cloud capital that Europe owned, like DeepL, essentially have lost their market value.
Nevertheless, and this is a huge nevertheless, it is only AI as a commodity that has lost its grossly exaggerated market price or value. In sharp contrast, cloud capital utilized as Amazon, Meta, Google and so on have been utilizing it, that is not as a commodity producing piece of tech but as a produced means of behavioral modification, that business model is not at all threatened by companies like DeepSeek. And since technofeudalism is powered by cloud capital working that way, rather than commodity-like AI services of the ChatGPT-4 or 5 type, our technofeudal order is not threatened by competitors such as DeepSeek. To help understand the difference between cloud capital and AI-based commodified services. It helps to compare and contrast Alexa, take Amazon’s Alexa, and OpenAI’s ChatGPT. Now, Alexa is not offering you a commodified service. It is your free, pretend slave. Unlike GPT-4 or 5, you do not pay a subscription to Amazon for the right to order Alexa, to order your milk, or to switch off your lights.
Rather, you train Alexa to train you, to train it, to know you, so that it wins you over, it wins your trust, with good recommendations. So that it can ultimately modify your behavior, so that it can encourage you to buy a commodity from Amazon.com with Jeff Bezos, the owner of Amazon, retaining up to 40% of the price you pay for a book or an electric bicycle, money that will be retained as cloud rent by the owner of Amazon, Jeff Bezos. In short, and this is very important, the work that Alexa performs for you is not a commodity that you buy, unlike ChatGPT, which works to sell you a commodity, even in a subscription form.
To put it in different words once more, ChatGPT is subject to market competition and therefore vulnerable to companies like DeepSeek. But Alexa is not. This is why OpenAI, ChatGPT’s maker, is seriously damaged by the emergence of DeepSeek, but Amazon is not. That’s my basic point. Cloud capital is in a league of its own, beyond market competition from DeepSeek-like upstarts, because its power lies in its capacity to modify our behavior and remove us from any market for example, to shift us from real markets to cloud fiefs like Amazon or Alibaba.
To wrap this up, in conclusion, cloud’s capital capacity to drive technofeudalism is not challenged by companies like DeepSeek. Only companies like OpenAI, which invested so much, and so foolishly I would add, in providing a commodified service, these companies stand to lose enormously. This, I believe, is yet another sign that capitalism is dead at the hand of cloud capital, while technofeudalism is going from strength to strength. And as it does so, it fuels even further the new Cold War between the United States and China, which in my book, “Technofeudalism: What Killed Capitalism,” I have explained away, I have explained this new Cold War as the almighty clash between these two huge concentrations of cloud capital, the American dollar denominated super cloud list power and the Chinese Yuan denominated super cloud list power. Now, speaking of this new Cold War, which I have argued is mostly fueled by the clash between American and Chinese cloud capital, I wonder what impact DeepSeek’s success will have on the United States government. Not just Trump, but the whole gamut of the American state and its government.
Silicon Valley and Washington, D.C., until very recently, until DeepSeek’s arrival on the scene, they had convinced themselves that America had a huge AI lead over China. Now that a tiny Chinese company has destroyed that confidence by producing on a shoestring better AI tech and services than Silicon Valley had imagined possible, I don’t know about you, but I can almost hear the wearing of the cogs and wheels inside the heads of people in authority both on the East Coast and the West Coast of the United States as they are thinking, trying to understand, to predict if the Chinese can do this out of the blue—as DeepSeek did only two days ago—what else can the Chinese do tomorrow? It is reminiscent, isn’t it, of the Sputnik moment. So it will be interesting to see how Donald Trump reacts to this threat to companies like OpenAI, especially since Elon Musk understood some time ago, quite presciently I should say, and has spoken out against companies like OpenAI. He seems to have understood the folly of commodifying AI services rather than going full-on technofeudal. Goodness only knows what happens in a White House containing both the thoughts of Elon Musk and someone like Donald Trump. These are indeed turning out to be interesting times, of course, in the traditional Chinese proverbial sense of the phrase. For more, watch this space.

Be the first to comment