JPM Asia Tech Strategy A

J P M O R G A N Asia Pacific Equity Research 05 August 2026

This material is neither intended to be distributed to Mainland China investors nor to provide securities investment consultancy services within the territory of Mainland China. This material or any portion hereof may not be reprinted, sold or redistributed without the written consent of J.P. Morgan.

Asia Tech Strategy AI Upcycle isn’t ending; reshuffling the pack for the next leg up

This is the third 20+% drawdown in Asian Tech stocks and the SoX index that we Technology and Telecoms have witnessed in the current AI driven upcycle that started in late 2022. Given the Gokul Hariharan AC strong momentum in EPS revisions, heavy investor crowding into price-hike ( narratives within Tech hardware names and lingering concerns about the sustainability of hyperscaler AI capex, investor concerns are understandably J.P. Morgan Securities (Asia Pacific) Limited/ J.P. Morgan Broking (Hong Kong) Limited higher this time round, as compared to the prior drawdowns. Stepping away from Jennifer Hsieh the share price moves, we do not see any fundamental indicators that signal ( meaningful weakness in the next 6-12 months. First, the key tenet of this upcycle, “AI Scaling Laws”, are still very much alive, with multiple AI labs competing for J.P. Morgan Securities (Taiwan) Limited the frontier with meaningful model performance increases every few David Chou months. Second, token demand for AI inference remains very strong, be it for ( premium tokens from Frontier AI Labs or cheaper ones from Open Source vendors, J.P. Morgan Securities (Taiwan) Limited with AI adoption continuing to rise steadily across various industries. Third, profitability across the AI ecosystem appears to have improved, with hardware Jason Chen ( vendors, AI cloud providers and AI Labs all seeing improving unit economics, with the advent of Agentic AI. Capex sustainability and funding ability is J.P. Morgan Securities (Taiwan) Limited understandably a key concern, with hyperscaler FCF turning negative in 2H26 and Subham Singhania 2027, but we do not anticipate any of the hyperscalers stepping back on AI compute ( investments in 2027 and think they are likely to tap into equity and debt markets to finance their AI infra rollout. Usual indicators of cycle peaks, such as inventory J.P. Morgan India Private Limited build of bottleneck components, a slowdown in demand growth, and a fast catch up of supply, aren’t flashing yellow, either. After the recent correction, we believe that the stocks are already anticipating an imminent EPS downgrade cycle or hyperscaler capex cuts in the next 3 months – we believe we are likely to see the reverse, with broadening EPS revisions and continued capex upgrades. Positive sentiment to break us out of the current downdraft could come from (1) Software and Internet vendors highlighting increasing adoption of Gen AI, LLMs and agentic workflows; (2) More indications of widespread adoption beyond software / coding automation, especially in Financials and Healthcare; (3) Continued model innovation, especially with leading frontier labs pointing to advancements towards recursive self-improvement (RSI) for AI models. We are buyers of Asian Tech stocks at these levels, and our key picks would be TSMC, MediaTek, Unimicron, Ibiden, Tokyo Electron, ASMPT, Aspeed, AMEC, Naura, Hon Precision, Chroma, Accton and Delta. Memory is a sub-segment where the top-down narrative has become more muddled due to concerns about demand destruction due to high prices (in AI chips) even though supply-demand fundamentals remain solid.

Stocks are now pricing in an imminent EPS cut, but we don’t see this through the end of 2026 or early 2027: After the 25-30% correction in Asian Tech stocks / SOX index, we believe that tech hardware stocks are pricing in a high likelihood of imminent EPS cuts, either due to diminishing component shortages or weaker demand (through capex cuts). We believe the inverse is likely true - EPS estimates are likely to keep rising through the next few quarters, and revision breadth is likely to broaden out, with analog, tier-2 foundries, wafers, MLCC etc. also joining the broader EPS upward…