In July, free software from China set off the worst week for U.S. chip stocks in over a year and briefly cost Nvidia its place as the world’s most valuable company. Wall Street panicked because Chinese firms had matched the leading U.S. artificial intelligence systems. But the selloff exposed something deeper. The AI boom rests on the promise that a handful of U.S. corporations will own the technology and charge the world for every use.
China is breaking that monopoly before it can be built.
On July 16, the Beijing firm Moonshot AI released Kimi K3. DeepSeek’s V4 model was already on the market. Both compete with the strongest systems from Anthropic, OpenAI and Google. Both are “open weight.”
Breaking open the toll gate
An AI model costs enormous sums to build. Once it has been trained, however, the model consists of a vast set of numbers called weights.
Those numbers can be copied.
When a company releases (opens) the weights, anyone with sufficient computing power can download the model and run it without paying the company for each use. U.S. firms generally keep their weights locked inside corporate data centers. Customers must pay every time they use the model.
Chinese firms are releasing some of the strongest models for others to use directly.
That threatens the toll gate Wall Street has already built into the price of every major AI stock.
The U.S. AI boom is not based on profits already being made. It is based on profits promised for the future. The companies at its center lose money on the services they sell. What customers pay to use the models does not cover the cost of running them.
Technology writer Ed Zitron estimates that the data-center capacity needed to support Nvidia’s projected chip sales through 2027 would have to bring in roughly $380 billion a year from customers buying AI computing power. That is revenue, not profit. Current annual spending on AI computing does not reach $70 billion, and OpenAI and Anthropic account for most of it.
Even that limited demand is partly manufactured. Nvidia invests in AI and data-center companies that use the money to buy Nvidia chips. Those purchases increase Nvidia’s sales, which support its stock price and attract still more investment.
The entire structure assumes that a few corporations will eventually own and control the technology and collect monopoly rents from everyone who uses it.
Open models attack that assumption directly.
Fictitious prosperity
Marx had a name for wealth of this kind: fictitious capital. A thin layer of real income supports a towering mass of debt and stock valuations — paper claims on profits that have not yet been made.
Production races ahead of the market. Across the industry, corporations are building data centers, installing vast numbers of chips and securing enormous supplies of electricity on a scale far beyond current demand. Investment continues because each capitalist expects future profits, even when the industry as a whole is creating far more computing capacity than can be sold profitably.
The tower stands only as long as investors continue to believe the promise beneath it.
The technology is real. The paper wealth piled on top of it is not.
Capital has already cashed in profits that the technology may never produce. The expected surge in productivity has not appeared in the figures. A Federal Reserve adviser reported in mid-July that businesses see little effect on productivity or employment so far, and most studies of the wider economy have found no clear productivity gain from AI.
Wall Street is trading a claim on future rents, not present income from a profitable product.
Those rents depend on monopoly control. Open-weight models threaten to make that monopoly impossible.
When an advanced model can be downloaded and operated outside a U.S. corporation’s servers, no company can force every user through its toll gate. The income assumed in the stock valuations never arrives.
China destroyed nothing profitable. It exposed how little profit was ever beneath the boom.
The July selloff showed how quickly belief can reverse. Once investors stop expecting the promised rents, they rush out of paper claims and back toward cash and safer assets.
That flight ends every fictitious prosperity.
The column holding up Wall Street
Capitalism regularly produces gluts. Too much money enters an industry. Too much is produced. Weaker companies collapse, capital pulls out and excess capacity is destroyed until supply falls closer to demand and profitability returns.
What makes this glut capable of shaking the entire U.S. economy is its scale.
AI-linked companies account for around 40% of the value of the S&P 500. They supplied the overwhelming majority of the stock market’s growth over the past year. Without them, the U.S. market barely rose.
Artificial intelligence has become the column holding up the market — no longer one speculative industry among many, but the one the others lean on.
A collapse in AI stocks would spread through the credit system, corporate borrowing and workers’ retirement funds. The same paper valuations that enriched billionaires have been used as collateral throughout the economy.
The Federal Reserve cannot create the missing value.
It can pump more dollars into the financial system to support asset prices. It cannot create the value those dollars are supposed to represent. It cannot make an unprofitable investment profitable. Flooding the system with more money to protect paper wealth transfers the loss through inflation. Refusing to intervene allows the paper wealth itself to collapse.
The ruling class faces a choice between devaluing the dollar and allowing its financial assets to fall. No monetary policy can erase the underlying loss.
Washington turns to coercion
Washington is therefore using the government to defend the threatened monopoly.
Security officials have proposed federal restrictions, official warnings and an executive order aimed at open-weight models. The stated reason is national security. The economic purpose is to wall off the Chinese systems, keep U.S. businesses on U.S.-owned models and preserve the toll gate through government power.
The government already intervenes directly in the market. In June, a Commerce Department order barred Anthropic from offering its most powerful model for 19 days by placing it under export controls. Washington will take even a leading U.S. firm’s flagship model off the market when its tech war with China demands it.
This is the chip war repeated in another form. U.S. sanctions and export controls failed to stop China from developing advanced semiconductors and AI systems. Now coercion is being used to defend a monopoly that competition has already broken.
When the market cannot preserve monopoly profits, the capitalist state steps in.
A different direction
The political answer came from Shanghai.
At the World Artificial Intelligence Conference on July 17, Xi Jinping called for AI to be developed openly as a shared asset of humanity rather than as the weapon of a single country. Dozens of countries joined a new cooperation body headquartered in Shanghai.
That policy rests on a material foundation created by the 1949 Chinese Revolution: state banks, public enterprises and national planning capable of directing investment toward production and public use.
Chinese technology companies compete and seek profits. But they operate inside a system in which the state can direct credit, infrastructure and research toward national development instead of allowing private monopoly rent to determine every decision.
That is the source of Wall Street’s panic.
The U.S. capitalist class staked the economy on enclosing a technology created through the collective labor and accumulated knowledge of the whole world. It planned to sell that technology back to the world at a monopoly price.
The enclosure is failing. The fictitious wealth built upon it is as fragile as it is enormous.
When the crisis breaks, capital will be destroyed on a vast scale. The capitalist class will try to preserve its own wealth by forcing the consequences onto workers through layoffs, wage cuts, lost pensions, inflation and austerity.
Workers created the technology. Workers produced the wealth that financed it. The decisive question is who will own it and whose needs it will serve.
Wall Street’s answer is a toll gate. The workers’ answer must be social ownership and public use.
Join the Struggle-La Lucha Telegram channel