The construction of the economy of the future has hit a bit of a snag. Case in point: IBM’s Black Tuesday.
It was the single worst trading day in the 115-year-old company’s history, with its shares down a shocking 25 per cent to US$217.07 apiece, erasing nearly US$70-billion in market value in a matter of minutes.
As a software provider to the artificial intelligence ecosystem, IBM’s nosedive says a lot about how the limitless potential of AI is colliding with physical constraints.
There is a global shortage of memory chips, and it’s getting worse. Prices for chips are soaring, and cloud providers and data centre operators are frantic to get their hands on supplies while they still can.
IBM was caught off guard by just how quickly corporate spending priorities have changed. As the race for storage and memory has heated up, budgets have shrunk for the kind of enterprise software IBM provides.
The memory supply crunch is at the core of several major disruptions: a new wave of consumer inflation, havoc across the consumer electronics industry – what Apple chief executive Tim Cook calls a “hundred-year flood” – and a major shuffling of the deck across financial markets.
IBM is a good example of the latter. Samsung Electronics is another, but in the opposite direction. The South Korean tech giant, which is now more of a memory chip company than a smartphone company, made 20 times more money in its latest quarter than in the same period last year. That’s about 2,000 per cent growth, year over year, all because of the run-up in memory prices.
Such is the power of AI mania to reshape fortunes. You don’t drop trillions of dollars in capital spending in short order without causing some serious distortions.
Recall in the early days of the pandemic, when policymakers flooded the economy with support. Trillions of dollars in stimulus probably staved off a depression, but also warped the economy in countless ways. Household incomes increased despite mass unemployment, for example, and all that cash helped prop up economies and stock markets for years.
In the same manner, the scale of the AI buildout is leaving plenty of unintended consequences in its wake. Global spending on data centres is expected to reach US$750-billion this year, and possibly US$1-trillion next year.
All those new data centres need to be stuffed with memory chips that the industry is increasingly struggling to supply.
“The situation has become unsustainable,” Apple‘s Mr. Cook recently told The Wall Street Journal. “I’ve never seen anything like it in any area in over 40 years.”
Companies like Apple are already passing the costs on to consumers, since those same chips are used in smartphones and computers. The price of a Macbook Neo laptop in Canada, for example, recently went up by almost 20 per cent.
For years, the prices of consumer electronics have generally trended downward. Now, households have a new inflation pain point to add to groceries and gasoline. In May, prices for computer equipment, software and supplies rose by 3.9 per cent over last year, according to Statistics Canada. It was the first increase in the category since 2020.
This is a problem without a solution, at least in the short term. Memory chip manufacturing is dominated by just three giants – Samsung, U.S.-based Micron Technology, and South Korea’s SK Hynix, which had a splashy debut on the Nasdaq exchange last week.
Just a few years ago, all three companies were cutting costs and production in the face of an industry downturn. Now, they cannot scale up nearly fast enough to keep pace with the insatiable demand of the AI giants.
“We forecast that next year will be the worst year in the industry’s history from the supply perspective,” SK Hynix chief executive officer Kwak Noh-Jung told Reuters last week. The supply shortfall is likely to last “even beyond 2030,” he said.
Building new factories takes time, and the AI giants are not likely to let up. They will pay whatever cost to secure their supply, bidding up the price of limited resources for the rest of us, and knocking around companies like IBM in the process.