Trending: Oil | Gold | BITCOIN | EUR/USD | GBP/USD

Is Chinese technology preparing to trigger a collapse in US AI stocks?

Economies.com
2026-09-23 17:01 UTC

When it comes to artificial intelligence, the attention of governments, businesses and the media is currently focused on the risks it poses to jobs, democracy and even the future of humanity itself. For many people of a certain generation, those fears are embodied by the cold, relentlessly polite voice of HAL, the artificial intelligence system aboard the spacecraft in Stanley Kubrick’s 2001: A Space Odyssey.

 

After one crew member questions HAL’s diagnosis of an antenna malfunction, the system secretly reads the astronauts’ lips, discovers their plan to disconnect it and then kills them one by one. Few people who have seen the film forget HAL’s chilling response when the last surviving crew member, Dave Bowman, orders it to “open the pod bay doors” so he can return to the spacecraft after HAL leaves him stranded outside during a spacewalk: “I’m sorry, Dave. I’m afraid I can’t do that.”

 

For US artificial intelligence companies, however, a far more immediate concern may be the major structural disadvantage they face compared with their Chinese counterparts, and the potentially enormous decline in their stock valuations that could soon result.

 

“The main problem here is that the leading Chinese AI developers are currently achieving around 90% of the performance of their US competitors, but at only about 10% of the cost,” Mehrdad Emadi, head of risk analysis and energy derivatives market advisory at London-based Betamatrix, told OilPrice.com exclusively last week.

 

“As much as half of the costs faced by US AI companies comes from the electricity required to operate data centers, and that is likely to increase from here,” he added. “The significantly lower costs of their Chinese AI competitors are likely to remain around current levels and could even decline slightly.”

 

“At the same time, Chinese AI performance is likely to continue improving toward a level close to 100% of that achieved by its US competitors,” he said. “That looks like a potentially fatal convergence for US AI companies.”

 

Steve Keen, professor emeritus at University College London and an economist who prominently warned about the approaching economic crisis that began in 2007, believes the enormous cost gap between US and Chinese AI companies stems from a fundamental difference in how the two countries’ electricity grids were designed and built.

 

“China benefits from a centralized electricity grid designed from the outset by engineers, unlike the decentralized US grid, which was designed by accountants and similar parties,” Keen told OilPrice.com exclusively.

 

“The US grid may have been cheaper to install initially, but it cannot move electricity across the country on anything like the scale of the Chinese system, and it loses considerably more energy during transmission,” he added.

 

More specifically, China’s electricity grid operates under a unified national strategy managed by the State Grid Corporation of China, allowing electricity to be transmitted seamlessly across thousands of kilometers.

 

By contrast, the US electricity system is divided into three largely separate interconnected grids: the Eastern, Western and Texas/ERCOT systems, making large-scale electricity transmission considerably more difficult and expensive.

 

To increase the amount of electricity transmitted at any given time, grid operators must increase either current or voltage. China’s electricity network relies on an extensive system of ultra-high-voltage direct-current transmission lines operating at between 800 and 1,100 kilovolts.

 

This configuration can transmit as much as 12 gigawatts of electricity, equivalent to the output of 10 nuclear power plants, through a single transmission corridor over distances exceeding 3,000 kilometers, with minimal power losses.

 

By contrast, the US electricity grid cannot approach this level of total power transmission or achieve similarly low losses over long distances, relying primarily on conventional 345-to-500-kilovolt transmission lines within a high-voltage alternating-current system.

 

In short, the fragmented and aging US electricity grid acts as a physical constraint on the scale and speed at which American technology companies can expand their AI computing clusters. China’s system, and consequently Chinese AI companies, does not face the same limitations.

 

Energy constraints threaten US AI expansion

 

There has been widespread speculation that rapidly rising US electricity demand could be met through increased oil and gas supplies or additional generation from other energy sources.

 

On the renewable side, however, solar and wind power still struggle to provide the “always-on” electricity supply required by AI data centers because of the intermittent nature of both energy sources.

 

Nuclear power, including small modular reactors (SMRs), has recently become a major focus, with US technology giants including Microsoft, Amazon and Google moving to invest tens of billions of dollars in such projects.

 

More importantly, however, both conventional nuclear projects and SMRs face enormous development lead times as well as potentially substantial additional costs.

 

Traditional US nuclear projects have repeatedly suffered multibillion-dollar cost overruns alongside development periods extending to 15 years.

 

SMRs, despite their lower upfront costs, still face targeted construction periods of as long as seven years.

 

Moreover, the need to secure regulatory approvals from bodies such as the US Nuclear Regulatory Commission, combined with the physical constraints of construction, means the commercial SMR industry is unlikely to reach significant scale before the late 2030s.

 

Among fossil fuels, oil remains too valuable to burn for baseload electricity generation, leaving natural gas as the most obvious primary fuel source for the power grid needed to support future AI expansion.

 

“The problem here is that everyone is trying to obtain the same equipment at the same time to turn gas into electricity for the grid, and there are very few suppliers manufacturing what is required,” Emadi said.

 

One of the clearest examples is the gas turbine manufacturing market, which is dominated by an oligopoly of the “big three”: US-based GE Vernova, Germany’s Siemens and Japan’s Mitsubishi Heavy Industries.

 

“Order books are completely full, waiting times have reached seven years and are increasing, and prices are on course to triple,” he said.

 

None of these structural obstacles is likely to ease anytime soon, particularly given expectations for a massive increase in US electricity demand from artificial intelligence over the coming decade.

 

According to Goldman Sachs Research, US data center capacity demand is expected to rise from a current baseline of around 42 gigawatts to between 118 and 134 gigawatts by 2030, while longer-term BloombergNEF models suggest it could reach around 194 gigawatts by 2035.

 

“With US companies producing high-quality AI at a substantially higher cost than Chinese companies producing products of broadly comparable quality, US firms need to start generating revenues commensurate with those costs, and to achieve that they need to massively increase their reliable electricity supply,” Keen said.

 

“But doing so appears extremely difficult given the physical constraints on power supply, so the potential for declining valuations among US AI companies looks significant.”

 

“What is happening now in the US AI sector is typical of an industry bubble in a capitalist economy, a phenomenon that invariably produces boom-and-bust periods, as happened, for example, with the US railroad bubble of the 1870s, which represented cutting-edge technology at the time,” Keen told OilPrice.com.

 

“Investment and attention poured into the railroad companies that were first to market, resulting in enormous overinvestment. Ultimately, however, only a handful of railroad companies became profitable, while most suffered losses and were sold for very little during the downturn that followed the boom.”

 

The situation could be even worse in the case of a US AI bubble because of the significant involvement of private credit, Emadi told OilPrice.com.

 

“To circumvent rules governing riskier lending by banks that were introduced after the 2007-2008 financial crisis, several major AI companies have used their enormous cash reserves to finance their own AI projects, but they have also borrowed from non-bank lenders, meaning the private credit sector, to increase leverage,” he said.

 

“So we are talking about hundreds of billions of dollars in debt linked to US AI companies at a time when their costs are running at around five to 10 times their revenues.”

 

“When AI companies begin going bankrupt, private credit lenders will have to sell other assets in an attempt to offset their losses. Given the scale of the exposure, that is likely to trigger a cascading effect of selling, falling prices and further selling throughout the financial system, causing valuations to collapse and triggering bankruptcies,” he added.

 

“This would not simply be a market reset, just as being shot in the head is not simply a physical reset. My estimate is that we will see between 35% and 50% of the entire US AI sector’s market capitalization disappear in a very short period of time.”

 

“That includes the bankruptcy of almost all the major players except perhaps two or three, which could be rescued by the US government because of their close links to the defense sector. Beyond that, I see most of the other AI companies that survive being left worth around 11 cents on the dollar,” he said.

Was this useful?

Thanks for your feedback.

Wall Street falls as oil and bond yields rise ahead of US-China summit

Economies.com
2026-09-23 14:43 UTC

Wall Street’s major indexes fell on Wednesday as crude oil prices resumed their upward trend and bond yields climbed, while investors awaited progress in talks aimed at ending the conflict in the Middle East ahead of the highly anticipated US-China summit.

More

Bitcoin extends gains, outperforming US stocks and gold since mid-August

Economies.com
2026-09-23 12:12 UTC

Bitcoin (BTC) extended its rally on Wednesday, trading above $86,000 at the time of writing after gaining more than 6% so far this week. Strong institutional demand continues to support Bitcoin’s bullish momentum, with US spot Bitcoin exchange-traded funds (ETFs) recording more than $714 million in inflows on Tuesday, following nearly $1 billion in positive flows the previous day.

More

Oil rebounds as diesel refining margins hit record high

Economies.com
2026-09-23 11:57 UTC

Oil prices rose on Wednesday despite improving crude supplies from the Gulf region, while diesel refining margins climbed to a record high amid the possibility of restrictions on US diesel exports.

More