The AI Circular Economy

Big tech companies and wealthy investors are driving an AI bubble - but they won't be the ones feeling the pain when it bursts.

Anyone keeping up with the news will be familiar with the idea that we’re living through an AI bubble. Labs like OpenAI have seen their valuations skyrocket, often without a commensurate increase in profits. There’s been a massive rollout of AI infrastructure, akin to overenthusiastic investments in fibre optic cables during the dot-com boom. And pretty boring companies that slap ‘AI’ on the end of their logo seem to be attracting vast amounts of investment.

Now, there’s even more evidence that we’re in bubble territory. The AI labs (creators of LLMs like ChatGPT, such as OpenAI), infrastructure providers (which provide hardware used to build and train these models), and big tech companies (which are buying from these companies, while trying to develop their own models and invest in their own infrastructure), have become enmeshed in a web of ‘circular financing’ that looks suspiciously similar to the dot-com era.

Circular Financing

Three companies have become particularly embedded in these financing loops: Nvidia, Oracle, and OpenAI. Nvidia makes the chips that power AI models. Oracle provides the cloud infrastructure. And OpenAI builds the models. On paper, each company plays a distinct role in the AI value chain. In reality, the lines between them are blurring.

Nvidia has announced plans to invest as much as $100 billion in OpenAI over the next few years, in a deal that I wrote about a few weeks ago. And the cash Nvidia is investing is going to flow straight back into its own coffers. OpenAI will use Nvidia’s investment to buy Nvidia’s own chips to power its new ‘AI factories’: data centres designed to train and deploy its models at scale.

Oracle is part of the same loop. OpenAI has agreed to spend $300bn to purchase data centre capacity from Oracle, which is now rushing to build new data centres packed with (you guessed it) Nvidia chips. So, Nvidia invests in OpenAI, OpenAI rents compute from Oracle, Oracle buys Nvidia’s hardware, and Nvidia records huge sales. Everyone in the circle reports rising revenue, rising valuations, and surging share prices, but the cash is largely moving in circles.

Behind all of this is Nvidia, the company that has benefitted the most from the bubble so far, and which is now using its profits and market power to make sure the boom continues. One observer commented that Nvidia has become the “central bank of AI”. In other words, Nvidia is the lender of last resort for the tech industry: it will keep lending to other companies to ensure the boom continues. But Nvidia can’t prop up the industry forever. At some point, OpenAI will have to show that there is real demand for their models – that customers are willing to pay very large sums of money to use them.

The OpenAI-Nvidia-Oracle loop is just one part of a broader system. OpenAI has made another massive deal with chipmaker AMD, which could ultimately result in the company purchasing enough chips to require “three times the [electric] capacity of the Hoover dam”. Microsoft has also been backing OpenAI for years, and Google and Amazon are also involved in another circle of deals with OpenAI competitor, Anthropic. Underpinning these deals are circular flows of financing that keep the whole system going.

If all this sounds a bit abstract, let’s take a more concrete example. Imagine if a car manufacturer invested in a taxi firm, which used the money to buy only that car manufacturer’s vehicles. The carmaker could claim rising sales, the taxi firm could say it was growing, and investors might conclude the transport sector was booming. Everyone would start to throw money at these firms for fear of missing out on the boom, creating a self-fulfilling cycle of rising valuations. But there’s no one at the end of the chain buying the product – taxi rides. Instead, much of the activity is being financed by the carmaker’s own money. Unless real customers start paying lots of money for taxi rides, the flow of money between these companies will eventually reach an abrupt halt. But why would anyone pay so much money for a taxi ride when, thanks to the investment boom, there are so many options available?

This kind of loop is known as circular financing: when a company’s investment or lending ends up returning as revenue through an interconnected web of partners or subsidiaries. There’s nothing illegal about it – and it’s not like investors can’t see what’s happening. But it is yet more evidence that the hype surrounding AI is driving a huge bubble.

Inflating the Bubble

Financial markets’ enthusiasm for AI has already reached astonishing heights. AI companies have accounted for 80 per cent of the gains in US stocks so far in 2025. Nvidia, already the world’s most valuable chipmaker, is trading at valuations that assume it will be able to keep growing at the same rate for decades. Oracle’s stock soared on the promise of cloud contracts that depend heavily on OpenAI’s still-unproven business model. And OpenAI is pumping hundreds of billions of dollars into deals that assume the bubble will continue forever, even as Sam Altman says it won’t.

The last time we saw anything like this level of circularity in the tech sector was the dot-com bubble. The FT explains:

“Back then, an enterprise software company might have paid to advertise with a new internet media company, in return for the media company buying its software. That artificial arrangement would have created the illusion of stronger demand for both companies’ services…

In the closest parallel to today’s AI infrastructure boom, telecom equipment companies such as Lucent and Nortel advanced money to customers in the 1990s to buy equipment, only to face write-offs when a wave of bankruptcies hit the industry.”

The big difference this time is scale. The amounts being pumped into AI infrastructure dwarf the amounts that were spent laying down fibre optic cables. All in all, according to the same article, OpenAI has put itself at the centre of a web of deals worth at least $1trn. And that’s just OpenAI. Google, Amazon, and Anthropic are part of their own web of circular deals too.

Another difference is debt levels, which were much higher among the telecoms companies at the peak of the boom than they are among the infrastructure firms now. But the direction of travel is heading very much in the wrong direction. Elon Musk is trying to raise over $12bn in debt for xAI, and Oracle has also been drawing much more heavily on bond markets to raise money. As I wrote last week, many other data centre builders are also relying on backroom deals with ‘shadow banks’ to raise money.

This time is different?

Some bubbles drive cash into largely useless technologies like cryptocurrency. But the another lesson from the dot-com bubble is that bubbles can lead to speedy rollouts of genuinely useful technology. The infrastructure laid out during the tech bubble - and, for that matter, the railway bubble of the 1840s - remains in use to this day. The tech bubble only burst because the companies investing in that infrastructure didn’t manage to make their money back in time.

In fact, it was the very speculative fervour underpinning their investments that ensured their profits remained depressed. With several companies competing to lay down as much fibre optic cable as possible, no one firm was able to enclose the market. Instead, they ended up all working together to deliver what became a public good: cheap, plentiful technology that allows people to connect with each other quickly over vast distances. These bubbles were classic Marxist crises of ‘overproduction’ - and in both cases, governments stepped in to help clean up the mess.

Eventually, investors realised that fibre optic cables weren’t going to generate vast profits for the now heavily indebted telecoms companies. In the mania of the dot-com boom, fraud had become rife. And many completely useless companies had been built on nothing more than the promise of the ‘internet revolution’. The end of the bubble delivered a shake out for the industry. The weak firms went under, and the strong ones remained – along with a lot of very useful new technology.

AI models have already achieved some incredible things. Google DeepMind’s AlphaFold, for example, has been able to solve a long-standing problem in biology by predicting a protein’s 3D structure from its amino acid sequence. There will be many more such breakthroughs as a result of the application of AI to tough problems. But that doesn’t mean that all the investment currently being undertaken in AI tech is going to pay off. Far from it.

First, there’s not much evidence that AI will deliver the massive productivity gains required to justify current levels of investment. If companies are going to pay thousands of dollars a month for models like ChatGPT (which is what would be required to deliver long-term profits for OpenAI), it’s going to have to generate thousands of dollars a month in cost savings or new revenue. And, right now, that’s not happening. As I wrote in a piece a few weeks ago, the AI boom is everywhere except in the productivity statistics. There’s good reason to believe that the AI labs will not be able to sell their product at a high enough price to warrant the sums being spent today.

To be fair, this was also true of the internet. It wasn’t obvious how new ICT technologies would increase productivity, and nor was it obvious how providers would make money from them. The big tech companies adjusted their business models, enclosing all our data and selling it on until they became some of the largest monopolies in history. If you’d invested in Google or Amazon at the height of the tech bubble, you’d be very rich today. It’s highly likely that companies like OpenAI and Anthropic will hold similar hegemonic positions in financial markets in another 25 years time.

But there’s reason to believe that the short-term losses will be bigger than those seen during the dot-com bubble. AI infrastructure becomes obsolete very quickly. Constant innovation in hardware is needed to meet the AI labs’ endless desire for compute. There’s every reason to expect that, by the time many of the promised data centres have been constructed, they’ll already be obsolete. That’s hundreds of billions of dollars effectively thrown away. In the short- to medium-term, someone is going to have to pick up the tab.

Who is going to pay?

It’s not like those involved in the AI boom don’t know that markets are getting pretty manic. Dozens of op-eds have been written laying out the case for why AI is a bubble. Many investors involved in this frenzy will have been involved in the dot-com boom years ago. But, for now, while there’s still money on the table, there’s absolutely no incentive for any of them to walk away. Each of them believes that they’ll be able to exit at the peak of the cycle – and some of them will do just this.

The AI companies, meanwhile, will be betting on the fact that they’re big enough to avoid going under regardless of which way the way the market goes. Amazon, Google, Microsoft, and the other tech companies that have come to define our generation survived the dot-com bubble. They’ll all survive this one too – joined by the new hegemons, like OpenAI and Anthropic. So, who is going to be left picking up the pieces?

Some professional investors who have made big bets will lose money. Retail investors who have pumped money into stock markets based on the expectation of infinite returns will lose even more money. And the big asset managers who manage our pensions, and are all exposed to some extent, will likely lose out too. Meanwhile, more nimble investors will be protected. And, when the bubble finally does burst, the state will step in to rescue their friends in the tech and finance sectors.

There is, however, a big open question about how much hidden leverage is building up in the system. As I outlined earlier, the big tech companies aren’t nearly as indebted as the telecoms companies were during the dot com boom, but debt levels are rising. As I wrote last week, it’s not entirely clear where all that debt is coming from. Much of the lending is being undertaken in private credit markets, which are notoriously opaque. In some senses, this diminishes risk, as the banks in which all of us hold our cash aren’t so exposed. But it also suggests that a few big non-bank financial institutions - ‘shadow banks’ - are very exposed. Some will get into hot water when valuations drop.

Even those of us who will not personally lose money (because we’re not wealthy enough to have savings to invest) will suffer when the AI boom bursts, because right now it’s one of the only things driving economic growth. As I wrote a few weeks ago, US GDP right now is pretty much entirely dependent upon AI spending and the consumption of the very rich, which is, in turn, based on ballooning tech valuations. If the bubble bursts, we’ll find ourselves in recession.

Big companies will use the downturn as an excuse to lay people off for good – replacing jobs with AI. Those who remain will spend much of their time managing the roll out of these new technologies, aware that their jobs are also on the line. All this creates the perfect conditions for another shift of power away from workers, who are too scared of losing their jobs to organise, and towards bosses, who will exploit workers’ fear to drive down wages.

For now, there’s still money to be made, so the cycle will continue. Nvidia’s cash will continue to fund the circular flows of finance until the money runs out. And when that money does run out, these companies will tap big financial institutions for yet more cash. Greedy investors will be only too happy to oblige. But the moment one link in the chain breaks, the whole thing will come crashing down. The tech companies will survive, and the governments will bail out the financial institutions that get into trouble. Once again, it will be ordinary people who are left picking up the pieces.

I’m not worried that the AI bubble will hearald the end of capitalism as we know it. I don’t even think it’s going to lead to the end of the AI boom. I am worried that it is going to result in a massive upwards transfer of wealth from ordinary people to some of the wealthiest and most powerful human beings on the planet - and, in the process, further endanger our democracies.