|
Global markets have continued to take a battering this week, with the sell-off in technology and semiconductor stocks intensifying to a degree not seen since the early days of the AI boom. South Korea's KOSPI index recently plunged nearly 11 percent, dragging down chip giants SK hynix and Samsung by double digits. Japan's Nikkei, Taiwan's benchmark and mainland Chinese shares all fell in sympathy, while on Wall Street the Nasdaq flirted with correction territory before clawing back most of its losses.
The trigger : a major Shanghai listing for Chinese memory chipmaker ChangXin Memory Technologies, whose shares soared roughly 500 percent on debut, followed by reports that another Chinese firm had begun mass-producing lithography technology long monopolised by the Dutch giant ASML. Layered on top was the release of a new open-source Chinese AI model, Kimi K3, which some experts suggest rivals Western AI systems at a fraction of the cost.
Together, the developments have revived a question that has hovered over the tech markets for two years: is the AI tech sector about to unravel?
A narrow, and narrowing, bet
What makes the moment unusual is the sheer concentration of money involved. The five largest companies on Wall Street, all tech giants, are now worth a combined sum approaching the size of China's entire economy. That concentration means a wobble in a handful of boardrooms in Silicon Valley can move markets from Seoul to Frankfurt to Auckland within hours, and it has left analysts describing current valuations as sitting well above levels seen even at the peak of the dot-com bubble.
The financing model underpinning the boom has also shifted in ways that make observers uneasy. Six months ago, the biggest AI spenders were buying back their own shares, a classic signal of surplus cash. Now those same companies are turning to debt to fund the enormous cost of data centres and chips.
Estimates suggest close to half a trillion US dollars in AI-related debt has already been issued this year. So far, the amounts being borrowed relative to these companies' overall size remain manageable, but a rise in US interest rates could make servicing that debt considerably more painful.
There is also growing unease about so-called circular financing, a term once associated with the sustainable economy concept, but which today describes large technology firms investing in AI start-ups which then spend much of that money buying cloud computing and other services from the very companies that funded them.
Amazon's multi-billion-dollar investment in Anthropic, paired with Anthropic running its computing primarily through Amazon's cloud, is one example analysts point to: money cycling between balance sheets in a little money-go-round which looks fine but has the potential to crumple.
It's not the only one. OpenAI's near-$500 billion data centre in Ohio comes with roughly $250 billion in backing from Nvidia, which means the chipmaker is effectively bankrolling a huge chunk of the demand for its own hardware. Microsoft, similarly, holds a large stake in OpenAI while also being one of its biggest cloud suppliers.
Each individual deal has a commercial logic. Put together though, they start to resemble a closed loop: a handful of firms trading capital back and forth, each transaction flattering the other's balance sheet, without much fresh money or genuine external demand entering the system at all.
The concern isn't that any single deal is improper, it's that the arrangement makes it hard to tell how much of the AI boom's apparent size is real growth versus the same dollars doing several laps around the same small group of companies.
Analysts warn that this kind of structure can hold together for a long time, right up until one link in the chain, a delayed product, a disappointing earnings report, and the trouble starts.
Genuine wobble or overdue breather?
Not everyone reads the past few weeks as the beginning of the end. Some strategists argue the sell-off looks more like profit-taking and a valuation reset after an extraordinary run-up, rather than evidence that the fundamentals of the AI business have broken down.
Demand for the high-bandwidth memory chips that power AI systems remains strong, and the largest technology companies have given no sign of pulling back their spending plans. What seems to have changed, is simply the market's appetite for what may come next.
Still, some of the warning signs are hard to dismiss. Oracle recently suffered its worst week since the dot-com crash of the early 2000s, a comparison that inevitably raises the spectre of 2001. And the economics of the AI labs themselves remain strained: OpenAI is reported to be burning through billions of dollars a month, with subscription and advertising revenue nowhere near covering the cost of building and running its infrastructure. It has already had to shut down one product, its video-generation tool Sora, after the running costs proved unsustainable.
Bigger than the last one
If the AI trade does eventually break, several observers argue the fallout would probably look different from the dot-com collapse of 25 years ago. That earlier bust largely wiped out smaller, speculative companies; today's boom is concentrated in the very largest tech firms, the kind that sit inside the retirement accounts and index funds of millions of ordinary savers across the Asia-Pacific. A serious correction in these stocks would not stay contained to Silicon Valley.
Beyond the disagreements over timing and severity, there is a consistent thread running through the commentary: nobody credibly claims to know exactly when a downturn, if it comes, will arrive. Bubbles are only ever definitively identified in hindsight.
For now, the current jitters give analysts no consensus on whether a crash is imminent and with earnings from Microsoft, Meta, Apple, Amazon and the major Asian chipmakers coming soon, that debate will continue.
-Asia Media Centre
|
