AI is being passed around as the next best thing to come from the world of technology. The most recent pronouncement on the topic comes from SoftBank CEO Masayoshi Son. Speaking at the company’s annual conference in Tokyo this week, Son claimed that artificial intelligence will require $5 trillion per year by 2040. Not to turn a profit. Simply to function.
“Every year $5 trillion, or 800 trillion yen, you might think that’s a lie, but I am confident that’s what it will cost,” Son said, as reported by Reuters. The SoftBank CEO also claimed that the idea of an AI bubble is “absurd”. Of course, given how much the Japanese entity relies on artificial intelligence succeeding, there’s nothing else he can say.
SoftBank goes hard on AI
The company’s fate is hinged particularly hard on the success of OpenAI’s ambitions. If a bubble exists, and it bursts, it will obliterate SoftBank as an entity. That’s not all such an event will do, of course, but Son’s company is especially vulnerable. A bubble isn’t even needed to kill off the Japanese bank. OpenAI just has to fail on its own to wipe out most of the company’s value. Projecting confidence is something its CEO is obliged to do. Even if it’s delusional.
Reuters points out the obvious in its report. Masayoshi Son, in his statements on AI, doesn’t explain how he gets to the annual cost figures for the tech. He also doesn’t detail how his explanation for why the $5 trillion figure isn’t a problem.
“The business model will be viable because by 2040, if AI revenue makes up 20% of global GDP, spending 800 trillion yen a year is a rounding error,” said Son. There’s some reason for his optimism, even if it’s poorly explained. A United Nations Trade and Development report reckons that AI could generate $4.8 trillion annually by 2033. The report is presented in generalised terms, and even it doesn’t explain exactly how artificial intelligence is supposed to achieve this. The report treats AI expansion as a foregone conclusion, based mostly on previous (and entirely speculative) growth. But it has fancy graphics, so you don’t have to think too hard about it.
Natural stupidity
Artificial intelligence doesn’t do anything to justify its expense. And make no mistake, it’s massively expensive. The technology has yet to make more money than has been spent on it. It will be years, assuming spending stops today and folks keep using it at its current rate, before it’ll cover its own expenses. And that’s globally, not just for one company’s spending. AI doesn’t work the way that it’s supposed to. It actively makes many aspects of human learning worse. Most of the numbers involving usage and uptake are fudged by companies like Google and Microsoft sticking AI services into their products. Whether you’re a user or not (and whether they work or not), they’re allowed to count you as a participant.
It’s already obvious to folks who handle money for a living that artificial intelligence is too expensive. According to the Bank of International Settlements, “Disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust, with potential knock-on effects on financial conditions…should hyperscalers slow or halt the aggressive pace of capex deployment, many borrowers across the supply chain could struggle to replace lost revenue and service their debt.”
In other words, there’s significant risk of a bubble exploding, taking loads of investors and the companies they’re connected to with it. The current global bill for AI, over just the five largest companies, is expected to hit $1 trillion between 2025 and 2026. Current annual turnover? Those are estimated at a maximum of $300 billion for the same period. That’s… not a good deal by anyone’s maths.
C’mon, do something
To justify spending of $5 trillion — Son’s figure is actually believable, based on current expenses for developing and training AI — artificial intelligence will have to account for a significant portion of global GDP. The trouble is that nobody seems able to explain how, exactly, the technology will do that. If it follows the Meta/Microsoft/Google route, it’ll be because people are forced to use it. Google broke its Search product and replaced it with AI. Microsoft will stuff Copilot anywhere that stands still long enough. Meta… doesn’t really know what it’s doing, but it’ll do it anyway. Few are asking whether this time and expense is worth it.
Let’s look at a few comparisons, shall we? The Manhattan Project, if conducted today, would have cost in the region of $35 billion. The Apollo Space Program would clock in at around $340 billion. Building the internet’s backbone, in 2026 terms, would sit around $100 billion. Those three projects collectively gave the world the nuclear bomb (and nuclear power), put human beings on the moon, and delivered space exploration to mankind. You could throw the lifetime cost of the International Space Station ($150 billion) on top of these achievements, and you still wouldn’t reach the amount of money expected to be spent on AI in 2025-2026.
These historical projects have all had some sort of impact on the world. There’s no arguing that artificial intelligence hasn’t had some sort of impact. But can you really class a technology that makes its users dumber and doesn’t have a clear function (and isn’t completely functional, even when it does have a use) with these achievements? At some point, folks are going to start asking whether AI is worth it. Unfortunately, unless you’re Sam Altman and his colleagues, the answer is becoming increasingly obvious.
