HSBC flags AI capex mismatch, others warn of ‘irrational exuberance’

As companies pour billions of dollars into artificial intelligence, HSBC CEO Georges Elhedery warned on Tuesday about the mismatch between investments and revenues.

Speaking at the Global Financial Leaders’ Investment Summit in Hong Kong, Elhedery said the scale of the investment poses a conundrum for companies: While computing power is required for AI, current revenue profiles may not justify such large expenditures.

Morgan Stanley’s July forecast He said that global data center capacity will grow sixfold in the next five years, and the cost of data centers and hardware will reach 3 trillion dollars by the end of 2028.

McKinsey said in a report In April, it was estimated that by 2030, data centers equipped to handle AI processing loads would require $5.2 trillion in capital expenditures to meet computing demand, while those powering traditional IT applications would require $1.5 trillion in capital expenditures.

Consumers aren’t prepared to pay the price for this, and businesses will be cautious because the productivity benefits won’t be realized for a year or two, Elhedery said.

“These are like five-year trends, and so the increase means we’ll probably start to see real income benefits and real readiness to pay for that later than investors’ expectations,” he said.

William Ford, chairman and CEO of General Atlantic, speaking on the same panel, agreed: “In the long run, you’re going to create a whole new set of industries and applications and productivity will pay off, but that’s a 10-20 year game.”

Major tech firms Alphabet, Meta, Microsoft and Amazon have lifted their guidance on capital spending and now collectively expect that figure to rise above $380 billion this year.

OpenAI, which started the artificial intelligence craze with the launch of ChatGPT in November 2022, announced that it has made infrastructure agreements worth approximately $ 1 trillion with partners including Nvidia, Oracle and Broadcom.

Ford said the big spending on the industry shows people are realizing the long-term impact of AI. But this sector will be capital-intensive to begin with, he added, and “you kind of have to make a down payment for the opportunity that’s going to present itself.”

Ford, “misallocation of capital, destruction, overvaluation… [and] “Irrational enthusiasm in the early stages,” he said, adding that it might be difficult to pick winners and losers now.

“You’re betting that this is a broad-based technology, more like railroads or electricity, that has profound impacts over time and reshapes the economy, but it’s very difficult to predict exactly how it will happen in the first few years.”