If Google’s nearly $500 billion one-day blowout over AI infrastructure spending not producing the intended results isn’t the sound of a bubble bursting, then what is?
Last week, when the search giant’s Alphabet holding company announced weaker-than-expected results, the market hammered Google, causing the search giant to lose 12% of its value. Not that Google didn’t make money. It reported revenue of $127.6 billion, but it has also spent outrageously on building giant new data centres and other infrastructure necessary to stay ahead in the AI race. Compute costs. A lot.
Losing what is now a global one-day record amount of money seems like an obvious warning sign, doesn’t it? But will that stop the rampant, unjustifiable spending on data centres and other necessary hardware to make the AI boom actually bear fruit? Probably not.
Clearly, the AI infrastructure boom is not going to benefit all the millions of people that the AI firms always argue their products will benefit. How has the life of a mother of four in Soweto changed since ChatGPT made it easier to ask the internet questions and write AI slop? Do you know anyone whose business has been as radically supercharged as everyone keeps telling us AI will do for ordinary companies?
If we analogise where AI is now to where the internet was at a similar point, it would be the equivalent of the internet boom circa 1997. But talking about it as if it’s today’s internet. We still don’t really know – apart from potential bioweapons, Lebensraum for OpenAI’s rogue agents, and astronomical listing projections for Anthropic – what the ultimate killer apps will be.
Like so many investment booms, the deep-pocketed investors will be the ultimate beneficiaries – as they always are – not the ordinary person in the street.
Is it time to face facts about the AI infrastructure boom and its stranglehold on the global economy? And will the world listen?
Scott Galloway, the NYU Stern professor who has emerged as a clear-eyed voice of reason in an age of unprecedented corporate hype about technology, always reminds his podcast listeners just how much of the global capital is invested in a small number of firms. Worse, Galloway and others warn that those firms are in a “circle jerk” of investments into each other. Nvidia sells the chips that power the boom but has stakes in the businesses that buy them.
Galloway warned last year that “when you have 10 companies responsible for 40% of the S&P’s value and the S&P makes up 50% of the total world’s value, you have 10 companies responsible for 20% of the entire valuation of public equities globally, meaning that is the string that if it gets pulled, everything goes down.”
Harvard economic policy professor Jason Furman also warned last year that “investment in information processing equipment & software is 4% of GDP… but it was responsible for 92% of GDP growth in the first half” of 2025. America’s GDP, excluding those categories, only “grew at a 0.1% annual rate.”
One sector’s 4% contribution to GDP turned into 92% of six months’ worth of GDP. Alarm bells, anyone? Especially as so much of AI infrastructure investment is being funded by companies taking on more debt.
Missed targets
There have been regular outcries when (only the listed) tech firms don’t reach their expected revenue, followed by a dip in the market as investors panic, and then a return to the mad rush to invest in data centres and AI infrastructure.
But something significant happened last week in the US midterms that is in no small part driven by a reaction to the power-hungry, water-hungry eyesores that are sprawlingly large data centres.
In many ways, these Democrat-won mid-term elections were a referendum on AI and the unprecedented power that tech firms have accrued in the US, including a cringeworthy display of sucking up to President Donald Trump.
Meta CEO Mark Zuckerberg almost gleefully dismantled Facebook’s disinformation and safety departments as soon as he could under the guise of the Republicans’ incessant complaints about conservative views being “censored” by social platforms. Zuckerberg saved Facebook hundreds of millions of dollars by not having to police the Wild West of misinformation and hate speech it created, while throwing its own users to the wolves.
From US states swooning over data centre projects by offering them lucrative tax discounts, as well as discounted water and power, the tables have turned on Big Tech, with voters demanding they pay their fair share in taxes and power bills.
In Finland, national police have started a criminal investigation into Google subsidiary Tuike Finland for clearing 330 hectares of forest for new data centres without completing the necessary environmental impact assessments. Some 400,000 trees have been chopped down in a country where forests are part of the revered natural ecosystem.
“Really, for Finland, there are no benefits,” Jussi Kangasharju, professor of computer science at the University of Helsinki, told The Guardian. “They are international companies serving international customers. Datacentres ultimately have little benefit for others except the datacentre operator.”
That is a salient point well worth remembering. If the data centres have struck deals to get cheaper taxes and utility bills, what’s the upside for the people living around the corner? Will they see more local municipal spending through higher taxes collected?
Or will they watch as a local community funds a listed multinational company’s increased profits through subsidised electricity bills?
Sam Altman, the CEO of OpenAI, whose agents autonomously attacked several companies and government websites, still wants us to be more tolerant of the errant behaviour of tech firms and their Broligarchy billionaires.
“The world should accept some bad things happening for the benefits of this technology,” he infamously said with a straight face. We should “accept some bad things” is as good a summation as you’re going to get the from the Tech Bro elite of Silicon Valley.
- This column first appeared in Business Day.
