In the run up to the early 2000s stock market crash which saw the Nasdaq go from 5,000 to below 1,000, there was a massive over-construction of optical fibre linkages. This led to a colossal level of over-capacity and reams of unused “dark” fibre. Are we seeing the same mass corporate groupthink again with the colossal investments by the likes of OpenAI, Google, Meta, Apple, Microsoft and xAI? Will Nvidia become the new Cisco as an incredible growth in sales disappears into thin air as the buyers realize that there is simply not enough profitable demand to utilize anywhere near the compute power installed?

This all goes to the US model of AI that is a result of the monopolistic and extractive minded nature of the US corporate leaders; real innovation left long ago to be replaced with monopolistic and oligopolistic profiteering. The US model is to build a huge amount of AI processing power behind a big paywall, supported by intellectual property lawyers, through which the providers can charge monopoly/oligopoly profits as a limited set of firms capable of funding such gargantuan computing facilities. Such an approach will hold back the general societal adoption of AI due to the high prices being charged and restrictions placed upon use.

The problem for this model is that the Chinese competitors have set out to fundamentally undermine it with a strategy of low compute requirement AI models, low cost access to those models, and open source. This strategy sidesteps US restrictions on the export of high-end AI chips (and even with these Huawei is very competitive in the AI chip space), while also facilitating a rapid uptake of AI models by software application providers and business and government users. For society as a whole, the greatest benefit is through the widespread adoption of new AI technologies, not through a select few charging exorbitant fees and restricting usage. The Chinese strategy could be likened to “let a thousand flowers bloom” while the US way is “let a few flowers suck the nutrients from the rest”.

If things were left to their own devices the outcome would massively favour the Chinese strategy, with the massive investments of the US monopolists being shown to be huge errors of competitive judgement. But US monopolists will act like monopolists and are already pushing to have Chinese AI models and services, such as Deepseek, banned from the US on “national security” excuses. Given the proliferation of such low cost models in China, the US would end up having to ban all Chinese AI models. This of course would be good for the monopolists but awful for US competitiveness in general, as the rest of the world leave the US behind; a US that would become the East Germany of AI models (as with EVs etc.). There is also the question of how the US would be able to stop the usage of AI-derived tools from other nations that embedded Chinese open source.

On May 13th 2025, the US attempted to close the barn door after the Chinese AI horse had already bolted into the next farmer’s field. The US Commerce Department’s Bureau of Industry and Security asserted that the use of Huawei’s Ascend AI chips anywhere in the world violates US export controls, and warned about “potential consequences” for anyone using US AI chips for the training and inference of Chinese AI models. The former statement is an implicit acceptance that Huawei has sidestepped US technology controls by engineering a competitive chip for AI. The latter is an acceptance that Chinese AI models are at least on a par with US AI models. This is an act of desperation, not one of strength, and will simply drive the Chinese state to accelerate even more the effort to remove all dependency on US high technology as well as further damage US chip manufacturers. The glaringly obvious attempt to force the world into a US AI rentier monopoly will also make other countries question their dependence on US high tech goods and intellectual property.

At the same time, the wheels are starting to wobble on the US AI train as some companies have started to cut back on data centre investments, and an over supply of access to Nvidia high-end chips is being reported in China. Once the growth story fades and the Dark Computing phenomena becomes more and more visible, the previous investment and stock market bubble will rapidly becoming a self-feeding bust. A company such as Nvidia may have a p/e of “only” 36, but if the buyers go on strike the “e” will rapidly fall pushing up that ratio very fast. Capital spending by publicly traded telecomm companies went from about US$120 billion in 2000 to less than half that amount in 2002; that’s how fast revenues can crash and earnings crash a lot faster than that. Cisco was found to be famously funding much of its own growth through loans to companies that bought their hardware, a parallel may be between the very close relationship between Nvidia and Coreweave; with the former rescuing the latter’s IPO with a US$250 million investment. Then there is the dependency of OpenAI on funding from Softbank, a perennial picker of technology investment losers, to fund the continuance of its massive buildout; with OpenAI showing no reasonable path to profitability and Softbank already stretched in its ability to borrow to fund its first set of commitments to OpenAI.

With the inability to produce monopoly profits from their huge compute power investments, the attempted monopolists will be forced to write down those investments; creating large financial write offs. For the pure play AI players, such as OpenAI and xAI, their business models will be shown to be utterly invalid and their perceived financial value will rapidly dissipate.

Below is a good documentary on the 2000 “dot com” crash. If anything, the players involved are much more leveraged and the financial intermediaries much less trustworthy than in the earlier period. Since then we have had the 2008 GFC and the hidden-under-COVID 2019 GFC, with the US financial markets currently supported by very large US fiscal deficits and a colossal Federal Reserve balance sheet.

One more bubble in the long list of bubbles created by the mass deregulation of financial markets of the past few decades will have burst. The only way to redirect US corporate leaders back toward building real competitive advantage instead of brittle monopolies and oligopolies that suck the creative life out of the US economy, as well as crushing living standards for the majority, is to put the financial system back in its New Deal straight-jacket, ban both share buybacks and executive stock options, and very aggressively enforce anti-trust legislation. Of course, none of that will happen given the tight oligarch control of the US state. Instead, the US will continue to lag as a China that tightly controls the financial sector and disciplines corporate leaders when required goes from strength to strength.