Jim O’Neill on Burnham, the US and the dangers of following the AI herd

“It doesn’t surprise me in the slightest, really,” O’Neill told me, when I asked him about the new sense of alarm over AI. “Over the past year, there’s been a very important shift that there’s not so much free cash flow for these big hyperscalers to spend, so they’ve all been borrowing.” In other words, the AI companies are not making enough money. 

Which is a problem, because the amount of money the AI industry needs to generate is enormous. PwC, the advisory company, projects that between here and 2050, AI companies will need to spend $31 trillion to build the infrastructure they need. But if they spend that money, at some point they will need to make it all back, meaning they will need customers – and those customers will have to be loaded. That’s where the problems start.

“This is just me being maybe a bit of an old curmudgeon, but also someone that’s been around in markets for 40-odd years,” says O’Neill. “What exactly is this stuff all supposed to do for us all? I mean, yes, it’s going to make a lot of these guys that own all the equity in principle very wealthy, and it could help corporations have yet another area where they can keep their costs under control by having less staff. But is it really going to generate economy-wide productivity gains?”

“If there’s no natural demand for this stuff, then there’s not going to be the demand for the products. In fact, I haven’t seen any details, but I noticed people are saying that Anthropic’s latest quarterly performance shows their sales slowing.” 

Anthropic operates the Claude AI platform, and is currently gearing up to offer shares for sale at some point in November. That initial public offering will be a huge test for both the AI industry and for the company, which estimates its own value at $2 trillion, although that self-valuation has been called into question by its weak sales – that customer problem again. Anthropic is one of the biggest, most successful and high-profile AI companies in the world. It has never made a profit. 

To make back the money spent on all those data centres, the AI companies will have to generate revenues on a colossal scale and giving away an AI platform to individuals in the form of an app obviously won’t do it. It’s also a stretch to think of an existing industry or sector that would be capable of spending enough on AI to make the tech firms whole on their huge multi-trillion dollar spending splurge. Is there any customer out there big enough for the AI industry?

Suggested Reading

Bonnie Garmus: ‘Donald Trump would not understand my book’

Eleanor Longman-Rood

“They seem to assume that everybody wants it,” says O’Neill, “but I can see how western governments – and the British government epitomises it – have huge problems trying to control costs and to improve the productivity of big expensive departments – the NHS.” 

This is the AI companies’ target market – governments. Only they have the resources to spend at the level the tech firms need. It would certainly explain why Palantir recently employed Tom Watson as a senior vice president. Who better to talk to your preferred customers in Whitehall than a former deputy leader of the Labour party? 

Palantir already has contracts with the NHS and MoD and is currently in a dispute with the Mayor of London’s office over a decision to block its contract with the Metropolitan Police. On taking the role, Watson accepted that there were “legitimate questions about large technology companies and sovereign control”, but said that criticism of the tech companies was “propaganda”. 

The extent to which AI companies are fusing with the inner systems of the state is beginning to lend them an almost quasi-governmental character. And their enormous size means they are starting to deal with governments on equal terms: “Right now the global market cap of Nvidia is bigger than the whole of the UK,” says O’Neill, “which is kind of madness.” AI companies are also beginning to compete with nations for resources. “We are meant to be building all these data centres for the US’s AI needs and we can’t even build the energy needs for ourselves,” says O’Neill. 

Alongside the AI boom is the surge in US government borrowing. America’s debt pile is now over $40 trillion and the UK’s is over £2 trillion. “I’m more worried about the US,” says O’Neill. “And this is another reason why the AI thing, if it breaks, raises quite a few questions that people are just not thinking about in the states.”

One of those currently-ignored issues is that a substantial chunk of the money invested in AI has been borrowed, and US interest rates are rising, meaning that the cost of all that borrowed AI finance is going up. But on top of that is “the government itself,” says O’Neill, and its attitude to borrowing. When it comes to public debt, he says, the US administration “is even more irresponsible than we are.” 

In comments made back in August, the US Treasury secretary Scott Bessent stated that the US could “grow our way out” of its enormous public debt position. Many observers, including the IMF, do not agree. The effect of America’s huge debt pile and its apparent unwillingness to do anything about it is causing a loss of confidence in the US, which is also helping to push up interest rates. 

And here in the UK, where up until recently O’Neill was advising Andy Burnham, is there anything we can do to insulate ourselves from the storm that an AI failure could bring? At this point, something unusual happens – the conversation turns optimistic.

“We have just had our monthly data for Aug this morning,” said O’Neill, “and year on year, that shows 1.8% growth, which by the standard of the past 20 years is nearly double what it’s been. The most important thing the chancellor should do in this budget is not damage that. Because it might well be that we could get close to 2% this year.” 

The trouble is, it seems, too many people still think the economy is locked in a doom loop. O’Neill recalls speaking to “two people in the Burnham camp”, about the stronger economic news. “They were like, ‘oh, whose data is that?’ I said ‘it’s the official UK GDP statistics!’ Peoples’ minds are just stuck… they can’t lift themselves above it.” 

“What’s tentatively encouraging, there are some signs of productivity picking up in the UK. So the last thing the chancellor should do, even though the fiscal situation remains quite bothersome, is to rush to some additional tax that’s going to harm business – and I think that’s what they’re going to do.”

“Burnham saying he’s going to adjust the triple lock I think is exactly the sort of thing we need to see happening. That needs to happen on welfare, and it needs to happen on the NHS. I personally suspect that when he has talked about his ten year plan, which is coming later this year, those things will be in it.”

As for the frailties of the global AI industry, there’s little that Britain can do about that. We remain outsiders, along with the rest of the world, looking in at a very small number of American tech giants and hoping they don’t blow it all up. 

And then, a question that perhaps goes outside the usual confines of financial and economic considerations – are the AI tech bros perhaps just a bit weird? The world of global business, investment and big money has usually been a place of sober, thoughtful types. But now, I ask, is the waywardness of the whole AI story perhaps a product of the waywardness of the characters who are telling it?

Suggested Reading

My truly deranged encounter with Trump apologist Kellyanne Conway

Alastair Campbell

“I get exactly what you’re saying,” he says. The big AI companies are heavily character-based, and that has a knock-on effect, in that the main actors are unwilling to relinquish any control. “Look at the way some of these big social media and AI firms have structured their equity holdings,” says  O’Neill. “They raise more capital but they don’t actually give away any voting control. What sane investor would want to own shares in that?”

Well, I say, it seems some people do. During the floatation of SpaceX, Elon Musk’s rocket maker that also includes his AI operations, investors vacuumed up the shares straight away. “Most of it is herd mentality,” says O’Neill. “The best investors don’t get sucked into that, but 90% of investors, emotionally and psychologically, find it hard.”

“If you’re the manager of Fund X, or Insurance Fund Y, and all your competitors are buying some stocks that are rising sharply in price, and you’re not buying them, then you’re worried you’re going to get fired next quarter.” If you were looking for a description of how a bubble gets inflated, it would be hard to find a better one than that. 

It’s also a reminder that stock market investors, governments and their voters are all being told the same thing by the AI tech bros: that they are now the indispensable industry. Their fundamental message to the world is, simply, “You need this”. But it is not at all clear what this really is, what it will do, or what we stand to gain from embedding it in the most sensitive departments of the state. 

It is much clearer what the AI companies would gain. Huge government contracts and revenues, along with the protection that comes from being “too big to fail”. If an AI company is running your military, or your health service, it cannot be allowed to go bust. It doesn’t seem like an attractive deal – or an especially fair one. 

For all the hype, the brilliant mathematical proofs and videos of the US president wrestling tigers, it’s not at all clear what we would get in return for the civilisational bet we are being invited to make on AI. “Is it going to solve health problems and the cost of health in so many western societies? Is it going to help or damage education?” says O’Neill. “Is it going to kill us?” 

“And as we’re seeing in left-of-center politics in the US, society’s starting to show some signs that it doesn’t want this stuff anyhow.”