
Technology-led selling pulls Asian markets sharply lower on Friday, following Wall Street’s overnight retreat. Enthusiasm over Anthropic’s record-breaking stock market debut failed to offset persistent concerns about whether enormous artificial intelligence investments would prove profitable and sustainable. Major regional indices came under pressure as technology heavyweights suffered the heaviest losses.

Japan’s Nikkei 225 sank nearly 5% to around 56,630, taking its weekly decline beyond 7% and reaching its weakest level since April. Sharp falls hit technology and AI-related names such as Kioxia Holdings, Advantest, and SoftBank Group. Fresh inflation figures added to the selling pressure by reinforcing expectations of further Bank of Japan interest rate increases. Japan’s annual inflation rate rose from 2.1% in September to 2.3% in October, the highest reading since November 2025.
A nearly 5% fall in South Korea’s benchmark KOSPI took the index to about 4,890, wiping out the previous session’s gains. Samsung Electronics and SK Hynix headed the decline among domestic chipmakers, with investors remaining uneasy about AI-related capital spending despite Anthropic’s robust market debut.
Substantial losses also marked the close for Chinese and Hong Kong shares. Broad-based technology selling pushed Hong Kong’s Hang Seng Index down 1.7% to roughly 22,450, its lowest since May 2025. Disappointing third-quarter results at Tencent, NetEase, and Kuaishou weighed on sentiment. Adding to those concerns, Tencent warned that computing constraints and tighter US semiconductor export restrictions could hold up the training of next-generation AI models.
Mainland indices retreated as well, with the Shanghai Composite losing 1.57% to 3,520 and the Shenzhen Component shedding 2.05% to 10,370. A trilateral meeting involving Presidents Xi Jinping, Donald Trump, and Vladimir Putin offered investors few catalysts. Beijing and Moscow reported broad agreement across four areas: United Nations reform, refined petroleum supply, International Criminal Court overreach, and nuclear risk reduction. However, according to a Kremlin aide, separate bilateral talks between Trump and Putin on the war in Ukraine yielded no breakthroughs.
The staging of events at APEC reflected changing geopolitical alignments, according to Rabobank analysts. They noted that Donald Trump was “greeted by Xi Jinping with a hug and pom-poms”, whereas Vladimir Putin “was met by a lower-level, taller Chinese official, which looked a deliberate diplomatic snub.” The analysts viewed the contrast between the US delegation’s warm welcome and the more restrained reception for Russia’s president as reinforcing the summit’s role in showcasing a changing global hierarchy, with possible implications for markets.
At its core, artificial intelligence is an academic field concerned with reproducing human cognitive functions, logical understanding, perception and pattern recognition in machines. Commonly shortened to AI, it encompasses several sub-fields: artificial neural networks, machine learning or predictive analytics, symbolic reasoning, deep learning, natural language processing, speech recognition, image recognition and expert systems. Creating artificial general intelligence, or AGI, is the ultimate objective of the field. In practical terms, this would mean building a machine capable of tackling arbitrary problems it has not been trained to solve.
Artificial intelligence has a wide range of applications, with generative AI platforms among the best known. Trained on large language models (LLMs), these platforms respond to text-based queries; examples include ChatGPT and Google’s Bard platform. For image creation, Midjourney produces original visuals from text supplied by users. Probabilistic techniques underpin other AI systems that assess an entity’s characteristics or how it is perceived. Upstart’s lending platform, for example, evaluates applicants’ creditworthiness through an AI-enhanced credit rating system that searches the internet for information on their careers, wealth profiles and relationships. In pharmaceutical research, other AI systems draw on extensive databases of scientific studies to propose potential new drugs for laboratory testing. Content aggregators, including YouTube, Spotify and Facebook, collect and organise viewing-habit data through AI applications to recommend personalised content to users.
Among the companies operating in AI is Nvidia (NVDA), a semiconductor manufacturer that produces AI-focused computer chips as well as some of the platforms engineers use to develop AI applications. Because it supplies the tools needed for further AI applications, many supporters consider Nvidia a picks-and-shovels investment in the AI revolution. Another participant, Palantir Technologies (PLTR), specialises in “big data” analytics and holds substantial contracts with the US intelligence community. That community uses its Gotham platform to analyse data, identify intelligence leads and support pattern recognition. Major corporations use Palantir’s Foundry product to monitor employee and customer data for predictive analytics and anomaly detection. Microsoft (MSFT), meanwhile, holds a sizeable stake in OpenAI, the privately held creator of ChatGPT. OpenAI’s technology has been incorporated into Microsoft’s Bing search engine.
AI-linked shares rallied after ChatGPT became available to the public in late 2022. In the six months after its release, Nvidia, for example, gained well over 200%, prompting immediate debate among Wall Street commentators over whether another technology bubble was taking hold. Prominent investor Stanley Druckenmiller, who has held sizeable positions in both Palantir and Nvidia, argued that bubbles do not last just six months. If AI enthusiasm did turn into a bubble, he said, extreme valuations would persist for at least two and a half years or longer, as they did during the DotCom bubble in the late 1990s. As of mid-2023, the best assessment was that the market was not in a bubble, at least for the time being. Although Nvidia was then valued at 27 times forward sales, analysts expected exceptionally strong revenue growth over the coming years. For comparison, the NASDAQ 100 traded at 60 times earnings at the DotCom bubble’s peak, against 25 times earnings in mid-2023.