insights taiwan

Focus Taiwan: AI, energy and 4Q outlook 13 October 2026

• The global landscape remains complex, shaped by AI, energy prices and tariffs. • Taiwan’s AI cycle remains strong, with robust AI exports and investment and an intact semiconductor competitive moat. • AI is generating broader domestic spillovers, with Ma Tieying, CFA wealth effects and wage growth supporting Senior Economist consumption. • Taiwan is cushioning the energy shock by diversifying energy imports and absorbing much of the increase in upstream costs. • AI-driven growth is likely to moderate from 4Q26 into 2027, as hyperscaler cash-flow pressures rise and higher global rates lift investment costs. • Energy remains an inflation risk, with domestic electricity prices facing upward pressure after the November local elections. • GDP growth is expected to moderate to high-single digits from 4Q26, while we maintain our 11.6%/5.6% GDP forecasts for 2026/27. • Inflation is likely to remain above 2% in 4Q26, with our 2.0%/2.1% CPI forecasts for 2026/27 unchanged. • We maintain our call for CBC to hike 12.5bp to 2.125% in December, amid tighter liquidity conditions. • Housing should recover gradually despite credit easing, as excess new-home inventory remains an overhang.

go.dbs.com/insightsdirect/macro/home

Taiwan: AI, energy and 4Q outlook October 13, 2026

Global backdrop on the economics and pace of AI investment. The global landscape remains complex, Hyperscaler cash flow vs. capex shaped by the interplay of AI, energy USDbn prices and tariffs, with significant 400 Operating cash flow implications for growth, inflation and financial conditions. 300 Cash capex

The debate over the durability of the AI 100 investment boom continues. Tailwinds 0 include technological advances, broader 1Q22 1Q23 1Q24 1Q25 1Q26 1Q27 1Q28 adoption, and strategic and geopolitical Sources: EPOCH AI, DBS considerations. Headwinds range from Notes: Hyperscalers include Microsoft, uncertainty over investment returns and Amazon, Alphabet, Meta, and Oracle hyperscaler financing pressure to physical bottlenecks, safety concerns, US-China AI competition is also drawing and regulatory uncertainty. increasing attention. Across the five layers of the AI ecosystem, China is Hyperscaler financing pressure has making rapid progress in energy, attracted increasing attention. In the infrastructure, AI models and early stages of the AI investment boom, applications, supported by its large scale, hyperscaler capex was largely financed cost-effective domestic manufacturing by internal cash flow. This is now supply chains and huge consumer changing, as free cash flow is expected market, which provides a wide range of to turn negative from 4Q, requiring major real-world scenarios for AI adoption. The players to rely on debt issuance and emergence of a more integrated and other external financing increasingly. cost-competitive Chinese AI ecosystem could increasingly challenge the US At the same time, higher global interest ecosystem, which currently operates at a rates are raising the cost of AI relatively higher cost. investment. The Fed resumed rate hikes at its September meeting, while global China, however, remains relatively weak bond yields remain under upward in semiconductors. Around half of pressure amid persistent inflation and China's domestic AI chip market still elevated public and private borrowing relies on foreign suppliers, including needs. Our baseline assumes US rates lower-end NVIDIA chips. This will rise by 50bps this year and a further semiconductor bottleneck remains a key 25bps in 2027 to 4.50%, with global constraint on China's AI development. In long-term rates also remaining under our view, Taiwanese semiconductor pressure. This could increasingly weigh companies' competitive moat therefore remains largely intact in the near term.