Last week, equity markets across the globe were rattled by a marked sell-off originating in the tech sector, and the semiconductor complex bore the sharpest correction. The Nasdaq Composite plunged over 4% in a single session; semiconductor stocks shed more than $1.3 trillion in market value — the most significant correction in the AI trade of 2026. Micron lost approximately 13% of its value in a single session, a remarkable reversal for a stock that had surged nearly 800% over the preceding year on the back of AI-driven memory demand. The contagion spread eastward with equal ferocity: South Korea’s tech-heavy Kospi index closed 10% lower, dragged down by SK Hynix and Samsung, both of which ended the session with losses exceeding 12%. In Europe, ASML, Infineon and STMicroelectronics fell between 4% and 7%. The sell-off was, in the most literal sense, global.

Three causes operated within the tech sector itself, against one macroeconomic backdrop that offered no mitigation. The first tech-specific cause refers to the recent IPO of Space-X. Days after the hyper-valuation of Space X, a large market correction was expected to emerge as it often happens after large IPOs.

The second and most structurally significant is the growing investor scepticism about whether artificial intelligence can generate returns commensurate with the capital it has consumed. Combined 2026 capital expenditure across Microsoft, Alphabet, Amazon and Meta has now exceeded $452 billion — a figure that has migrated, in investor psychology, from impressive to alarming. When Apple and Microsoft simultaneously announced price increases on consumer hardware, citing soaring component and chip costs, the market read the subtext clearly: the AI infrastructure buildout is consuming margin across the entire supply chain, and those costs are now being passed downstream. Studies from MIT and the Harvard Business Review found that 95% of organisations have seen no tangible return from AI projects to date, even as cumulative industry investment approaches $1.6 trillion since the modern AI boom began. The gap between capital deployed and value demonstrably created has become too wide to ignore.

The third tech-specific cause is structural and geopolitical in equal measure: China has repeatedly demonstrated that comparable AI results can be achieved with a fraction of the investment, directly undermining the narrative that underpins hyperscaler valuations. The most prominent case beyond DeepSeek is Alibaba’s Qwen3, which ranks near the top of global leaderboards in reasoning and language tasks while operating under the same chip-constrained conditions facing all Chinese labs — and, unlike DeepSeek, is backed by a major listed company with real revenue. MiniMax produced its M1 model trained on just 512 Nvidia H800 GPUs, a figure US hyperscalers would have dismissed as inadequate for any competitive frontier model. Ant Group demonstrated something more pointed still: the ability to train AI models on domestically produced chips at lower cost with equivalent results — compute substitutability, not merely compute efficiency. Moonshot AI’s Kimi K2.5 costs four times less than OpenAI’s GPT 5.2 while scoring an identical Intelligence Index of 47. The aggregate pattern, confirmed by RAND in early 2026, is that Chinese AI models run at one-sixth to one-quarter the cost of comparable American systems.

Against this already unsettled backdrop, the macroeconomic environment offered no comfort. Geopolitical tensions in the Middle East continued to furnish a reliable volatility amplifier, sustaining fears of supply disruptions, persistent inflationary pressure and the possibility of a Federal Reserve less inclined toward accommodation than markets had priced. The US Dollar Index climbed to a new 2026 high above 101, reflecting safe-haven flows. Capital rotated accordingly — out of mega-cap technology and into healthcare, utilities and consumer staples, the perennial refuges of the risk-averse. Central banks rotated their investment from US Treasuries to gold.

This was not a repudiation of artificial intelligence as a transformative technology. It was something more disciplined: markets withdrawing the blank cheque extended on faith, and issuing instead a conditional invoice. Payment terms: verified earnings, sustainable margins, and a credible timeline from expenditure to return.