Germany nd
Artificial Intelligence | The AI industry and its value
The boom in artificial intelligence is generating billions in profits. Is this all hot air? Or real added value? Or technological feudalism?
Deutsch (original) ·Auto-translated to English
This is what it looks like when semiconductors are produced, like here in Infineon's Smart Power Fab in Dresden. But is this also how value is created? Photo: IMAGO/Sylvio Dittrich One of the largest AI companies, Anthropic, is going public. Speculation on the success of artificial intelligence has been driving not only the stock markets for months, but also the so-called real economy: billions upon billions are flowing into the construction of data centers and their energy supply. In addition to the question of whether all this effort brings capitalist profits, it is worth taking a closer look at where the profits actually come from?
According to its own information, Anthropic is aiming for a valuation of over one trillion dollars for its IPO. On the one hand, this seems almost modest, as the first investor estimates were up to two trillion dollars. On the other hand, the company recorded a net loss of almost $42 billion in 2025. According to its own reports, it was most recently profitable for the first time in the third quarter of 2026 - and a lot of creative accounting was also involved in this calculation.
The final proof of the extent to which artificial intelligence can be used to organize a viable capitalist business is still missing. That's one thing. The other thing is that the first publicly traded pure-play AI company is going public at a time when AI-related stocks have seen a huge price slide. The expectation that the US Democrats, who won the midterm elections, could restrict the construction of data centers had destroyed more than a trillion dollars in stock market value in the USA. However, the often predicted bursting of the “AI bubble” does not seem to have happened yet, as stock market values have recovered again.
A trillion-dollar stock market valuation of Anthropic on the one hand, a trillion-dollar share price losses on the other: the extent to which the speculation on artificial intelligence will ultimately come true remains as open as the question of how these valuations can actually be explained economically. Is this all “hot air”? Or does it have something to do with value-creating work? Let's try an orthodox explanation in the Marxian sense. Let's start with the empirical.
At the center of the AI economy are a number of mega-corporations, including Nvidia, the second largest company in the world with around four trillion dollars in market value. As a result of the price decline, its market value fell by $400 billion the week before last. Nvidia is traded as a “chip manufacturer”. However, this is a somewhat misleading label when you look at the business and financing model.
The former manufacturer of graphics cards has become big because its computer chips are the material basis of artificial intelligence. Doubts are growing as to whether AI will achieve the hoped-for productivity gains and thus justify its boom. Despite this, billions are flowing into AI development. Nvidia has helped a lot. So that the AI development for which Nvidia supplies the chips does not stall, the company itself ensures that its customers remain liquid: 30 billion dollars went to OpenAI, up to ten billion to Anthropic, sometimes two billion here and there to CoreWeave and Nebius. These companies, in turn, invest the loans granted to them back to Nvidia, which can be sure that demand for its semiconductors will not decline. A billion dollar carousel.
So much for phenomenology, which shows that speculation on future wealth in capitalism already has the status of wealth itself. This is only fitting in an economy where everything is geared towards growth and where companies are valued solely on how much future profit they promise to bring. The extent to which this increase in the AI industry will actually take place has not yet been determined. But speculation on this pumps trillions into the industry and makes those investors rich whose securities represent a promise of a share in the expected AI boom, for example in the form of dividends.
If AI actually becomes a bomb business and Anthropic and Nvidia actually make the profits they hope for, this would only have very little to do with the work of the employees of these companies. First of all, Nvidia is not the company that produces the chips. Rather, it is the company that has the intellectual property rights to the semiconductor technology. On his behalf, companies like the Taiwanese TSMC produce the chips that Nvidia counts as the producer - simply because it is their owner. Anyone who criticizes this equation of owner and manufacturer should remember: The company TSMC is actually not the manufacturer of the chips, but rather its employees. They produce the goods, which then belong to the company, which is why – legally correct – it operates as a “manufacturer”.
Where Nvidia's wealth is not predetermined wealth, the source of the company's profit is the added value that does not arise at Nvidia at all, but at the chip factories in Korea or Taiwan. Thanks to its ownership rights to the chips, Nvidia can benefit from this added value quite successfully and receive a pension accordingly. What this means is that Nvidia realizes value that was not actually produced by it.
However, some Marxist critics do not accept this definition of Nvidia profits as a pension, including Michael Roberts recently in the “Yearbook for Marxist Social Theory”. After all, rent is “the form of income associated with goods that have no value.” But even at Nvidia, according to Roberts, work is not manual, but rather intellectual work, which is just as valuable as manual work. The fact that Nvidia also generates added value suggests that the company does not just sell the chips produced by others, but rather, for example, with its CUDA software, provides an entire architecture of a digital nature that also contributes to the utility value of the AI chips. And why should one negate the creation of value in this work just because it is of a spiritual and not a material nature?
But this argument doesn't hold water. Because the economic difference does not lie in the oft-quoted dichotomy of material versus intellectual activity. Rather, it is about the question of the extent to which socially necessary, average work is contained in a product, which makes it a thing of value in the first place. And this is where a peculiarity of the software comes into play that is often neglected: Although a lot of “brain, muscle and nerve” goes into its development, this work only occurs once - no further work is required to reproduce your product. And Marx rightly points out that it is not the work that is necessary for the one-off production that creates value; but that which is needed for constant reproduction.
The fact that Nvidia and Co. can make money with software is solely due to the fact that this software is their intellectual property.
The fact that Nvidia and Co. can make money with software is solely due to the fact that this software is their intellectual property. And this in turn is due to the fact that the bourgeois state uses its law to guarantee that even information such as the design of a semiconductor and a piece of software may not simply be used by producers. Chip manufacturers like TSMC have long known the specifics of producing chips for Nvidia. In order to work with them, however, they have to give part of their added value as a rent to the owners of these ideas. The state, with its copyright and patent law, ensures that even ideas whose use - once conceived - no longer requires any work, become a bubbling source of private property.
The fact that pensions are accruing on a massive scale in the tech sector should not lead one to diagnose the approach of “rentier capitalism” or “tech feudalism,” as economist and former Greek finance minister Yanis Varoufakis has been doing for years. He assumes that rents will gradually replace surplus value and thus “kill” capitalism. But this theory is - among other theoretical shortcomings - at least as recursive as Nvidia's financing model. “Rent” cannot “replace” “value” at all and therefore cannot “replace” a particular type of capitalism – or establish a “neo-” or “digital feudalism”.
Companies like Nvidia can only build their profits on a pension because other companies produce corresponding added value, which is divided between both companies. Where profit neither comes from the production of surplus value nor is based on rent, it arises from speculation about future wealth. And that is no substitute for the fact that this has to come. If he doesn't come, the loan that previously replaced the value will also be devalued. And should this moment come and the “AI bubble” burst, the consequences will not be limited to the stock market.
This article cost you nothing. Us: research, production and technology. We saved on office rent as well as printing and distribution. And we gave up salary. Nevertheless, the calculation doesn't work without you. We need 300,000 euros by the end of the year. You pay what you can, in return we keep writing.
No paywall, no small print, just one goal: that nd stays.