AI Economics · · 4 minute read
In the first week of November, investors sold Asian chip stocks over American politics while Korea shipped a record month of chips. An operator should read the shipments, not the share prices.
Last week was the worst for Asian AI stocks this year. It was also one of the best months on record for the companies that actually sell into the AI build-out. Both things are true, and the gap between them is the most useful thing an operator can study right now.
The selloff was about who might slow down American data centres. The shipments were about who is still buying. Neither tells you whether AI is working inside your own business. That question has its own numbers, and they point somewhere else entirely.
In the week of 2 to 6 November, the AI trade unwound across the region. Japan's Nikkei and Taiwan's Taiex entered technical bear markets, and South Korea's Kospi fell as much as 8.8 per cent in one session, triggering a trading halt (Bloomberg, 6 November 2026). The stated causes were American: worries that Democratic midterm gains could slow data-centre construction, voter anger over power bills, and higher borrowing costs (AP via Click2Houston, 9 November 2026). By Monday the Nikkei was up 2.3 per cent again.
Bloomberg reports that Asian hedge funds and wealthy investors took steep losses, and that money is rotating back into Chinese and Indian stocks that the AI trade had left behind (Bloomberg, 8 November 2026).
So what: If a board member asks whether to pause AI spending because "the bubble burst", the honest answer is that a stock market repriced a political risk in another country. That is a reason to check your vendor exposure to US data-centre policy. It is not a verdict on your use cases.
On 1 November, South Korea's trade ministry reported October exports of US$115.3 billion, up 93.7 per cent on a year earlier. Semiconductors alone were US$56.4 billion, up 259 per cent (Korea Times, 1 November 2026). Two extra working days flattered the comparison, but not by a factor of three.
There is a softer signal underneath. Manufacturing surveys for October show the big chip exporters still expanding but cooling: Taiwan's PMI fell to 54.9 from 56.7, South Korea's to 51.3 from 53.9, and Japan's to 53.3 from 54.1. Vietnam, Indonesia and Malaysia sped up (Reuters via Manila Times, 3 November 2026).
So what: Real demand for AI hardware has not fallen, so neither will the price of the compute and memory you rent. As I wrote in October, the AI bill is going up. A falling share price at SK hynix does not make your cloud invoice smaller.
The e-Conomy SEA 2026 report from Google, Temasek and Bain, published on 21 October, counts 243 million generative AI users in the region's six largest markets, 58 per cent of the working population, adopting four times faster than mobile internet did (Google, 21 October 2026). Usage per person is about a tenth of US levels, and consumers generate about 60 per cent of the region's tokens. Enterprise workloads reach 45 to 60 per cent of usage only in Singapore and Malaysia (CRN Asia, October 2026). Only about 1 to 2 per cent of the population pays for a personal AI account (e-Conomy SEA 2026 report).
Meanwhile, companies with AI at the core of their product took 80 per cent of the region's digital-application deal value in the past six months, up from 30 per cent in the second half of 2024, per the same report.
So what: Investors are funding AI products faster than customers are paying for them. For a buyer, that is leverage. Expect discounts, free tiers and bundles to persist, and negotiate accordingly. For a seller, it is a warning: adoption is not revenue, and most of your users are individuals experimenting, not companies running workflows.
Three numbers, three different questions. Share prices measure what investors think about American politics and interest rates. Export figures measure how much hardware the world is buying. Usage per person measures whether people are doing real work with AI. Only the last one is yours to move.
Asia's problem is not that the AI trade fell. It is that adoption is broad and shallow: millions of people trying the tools, a tenth of the depth of use, and very few organisations paying for it as part of how work gets done. Markets will keep rising and falling with the news from Washington. The firms that come out ahead will be the ones that ignored that noise and kept training their own people to use the tools properly. That work does not show up in this week's market data. It shows up in next year's margins.
Keep reading
AI Economics
In five days, a reported US$50 billion pledge to Malaysia, a US$15 billion round for DeepSeek and a memory maker heading for a 78 per cent margin. The money behind Asian AI is no longer venture money, and that changes the risk an operator is buying.
AI Economics
Asia's biggest AI earners are not selling AI as a product. They sell it as a feature of something customers already pay for, and that changes how an operator should buy it.
AI Economics
Asia makes the chips, pours the concrete and writes the code for the AI boom. The profits sit with the first two. Most of the region's businesses are still on the paying side.