Trillions of dollars have been wiped off global stock markets in a tech-driven rout that began on Wall Street and has now swept through Asia, forcing Korea's Kospi to halt trading and leaving investors asking whether the AI boom has tipped from bubble into crash.

The trigger, by most market accounts, was political rather than corporate. Traders spent this week digesting the consequences of the US midterms, held on Tuesday, in which data centers were as much on the ballot as the cost of living. The scale of the Democrats' victory — the so-called 'blue nami' through the world's largest economy — signalled to markets that America's AI infrastructure rollout would slow under a new Congress.

The numbers landed quickly. At the US market open on Thursday, the Nasdaq fell four per cent and the S&P 500 fell three per cent. Contagion then set in across tech-heavy Asian markets overnight: Japan's Nikkei dropped more than six per cent, and Korea's Kospi — a bellwether for AI sentiment — was forced to halt trading amid liquidity concerns.

For City AM's Midday Markets newsletter, analyst Ali Lyon chewed over what it all means. Analysts, she noted, woke on Friday reaching for their "morbid bon mots" of choice — "bloodbath", "capitulation-level event", "sea of red" or simply "extraordinary". Whatever the descriptor, there was only one markets story in town.

Is this a sudden reversal?

Lyon's first observation cuts against the panic: Friday's rout, however severe, does not represent a sudden change in direction. Tech stocks — particularly those most exposed to the excesses of AI exuberance — have been trending down for much of the past two months.

The detail bears that out. Sandisk, which makes memory chips for data centers, fell 22 per cent yesterday — but that came on top of a 45 per cent drop over the preceding month and a bit. The major hyperscalers — Nvidia, Meta, Alphabet and Tesla — are all trading considerably below where they started the year.

What does the sell-off mean for AI itself?

Lyon's second point matters for anyone building on the technology rather than trading it: even a protracted period of share price falls would not herald the end of the AI boom at large. Firms and individuals are still adopting the technology rapidly, and the productivity gains derived from it remain untouched by a shift in market sentiment.

The precedent is familiar. In the dotcom bubble of the early noughties, investors got well ahead of their skis, just as they may have now. Pets.com, the poster child of that era, reached a peak valuation of $400m despite generating only $5m in revenue and losing more than $60m a year. Its overdue crash did not stop the internet becoming the defining innovation of its time, nor slow its adoption.

The comparison gives practitioners a working framework: the equity unwind and the underlying technology can decouple, sometimes for years.

Could the rout reach the real economy?

The third element, in Lyon's assessment, deserves the closest watch: what this means for the US and global economies. Some argue the stock market is dislocated from the real economy — that Wall Street and Main Street barely interact. But the 'wealth effect', especially in the US, is powerful.

The mechanism is simple. When people look at a healthy pension pot, stocks and shares Isa or individual retirement account, they buy the car they have been pining for or go out for an indulgent dinner. When the picture looks less rosy, as it will today, the opposite holds. Demand slows, and economies grind to a halt.

Where does this leave investors?

Lyon's conclusion is deliberately restrained. A lot is going on in AI and markets at the moment, she wrote, and no one — least of all her — can say with certainty where markets will find themselves next.

For portfolio managers, the actionable read is threefold: the sell-off has an identifiable political catalyst in the midterm result; the underlying AI adoption cycle shows no sign of breaking; and the wealth effect now becomes the transmission channel to watch between falling screens and slowing consumption.