Summary: With AI large model parameters surpassing the trillion-parameter mark, global computing demand is growing exponentially, fueling demand for GPUs, HBM, advanced packaging, and other chip categories. In Q2 2026, global semiconductor sales grew 18.7% YoY, with AI-related chips contributing over 60% of the increment. This article offers an analysis from the four dimensions of technological barriers, domestic substitution, policy support, and long-term market demand to explain why the present is a go
On July 29, 2026, Singapore — The global semiconductor industry is undergoing a structural transformation driven by artificial intelligence. According to the latest data from the World Semiconductor Trade Statistics (WSTS), global semiconductor sales in the second quarter of 2026 reached $162 billion, up 18.7% year-on-year and 6.2% quarter-on-quarter. Among them, AI-related chips (including GPUs, AI accelerators, and HBM high-bandwidth memory) accounted for over 40% of sales, becoming the core engine of industry growth. Analysts point out that this round of growth is not cyclical fluctuations but a long-term demand explosion triggered by technological paradigm upgrades, fundamentally changing chip investment logic.
Computing Demand Expands Exponentially, AI Chips Become "New Oil"
After large models such as ChatGPT-5 and Gemini 3.0 successively surpassed the trillion-parameter threshold, the computing power required for AI training and inference is growing at a rate of 4-5 times per year. OpenAI's latest "Orion" model has 12 trillion parameters, and the computing power consumed in a single training session equals the total annual computation of a traditional internet company. Global leading cloud service providers — Amazon AWS, Microsoft Azure, and Google Cloud — will invest a combined over $350 billion in capital expenditure in 2026, with about 70% directed toward AI infrastructure, including purchasing high-end GPU servers such as NVIDIA H200, B200, and AMD MI400.
"The current global AI computing gap is still widening. Even with TSMC's 3nm capacity fully loaded and CoWoS advanced packaging capacity doubled, demand cannot be met," said Chen Weiming, a renowned semiconductor analyst and visiting professor at the National University of Singapore. "This supply-demand imbalance will persist for at least 2-3 years, creating huge investment opportunities across the upstream chip design, manufacturing, packaging, and testing supply chain."
Domestic Substitution Accelerates, China's Chip Supply Chain Ushers in Golden Period
Against the backdrop of escalating U.S. export controls on China, the autonomous development of China's semiconductor industry has significantly accelerated. In July 2026, the Ministry of Industry and Information Technology officially released the "Three-Year Action Plan for High-Quality Development of the Semiconductor Industry (2026-2028)," planning to invest 1.5 trillion yuan to support key areas such as chip design, manufacturing equipment, and materials. Companies like HiSilicon, SMIC, and YMTC are expanding capacity. SMIC's three 12-inch wafer fabs in Beijing, Shanghai, and Shenzhen are expected to be fully operational by 2027, increasing mature process capacity by 50%.
Chip-related companies listed on the Singapore Exchange (SGX) have also benefited. Semiconductor equipment and parts suppliers such as UMS Holdings and Frencken Group saw first-half revenue growth of 32% and 28% year-on-year respectively, mainly due to massive purchases of domestic equipment by Chinese local wafer fabs. Analysts point out that domestic substitution is not only policy-driven but also an inevitable choice under geopolitical risks, and related supply chain companies will maintain a compound growth rate of over 20% in the next three years.
Policy Dividends Continue to Release, Global Countries Increase Semiconductor Support
Besides China, major economies such as the United States, the European Union, Japan, and South Korea are all stepping up support for the semiconductor industry. The second round of subsidies under the U.S. CHIPS and Science Act landed in April 2026, with Intel, Samsung, and TSMC receiving a total of $28 billion to build advanced process factories on U.S. soil. The EU's European Chips Act aims to increase the global market share to 20% by 2030, investing over €43 billion, focusing on R&D for 2nm and below processes.
The Singapore government is also actively deploying. In June 2026, the Singapore Economic Development Board (EDB) announced the establishment of the "Semiconductor Innovation Fund," with an initial injection of SGD 5 billion to attract top global chip design companies to set up R&D centers in the Jurong Innovation District and support local startups in developing RISC-V architecture chips. Currently, Singapore has gathered over 50 semiconductor design companies, including regional headquarters of giants like Broadcom and MediaTek.
Long-Term Demand Highly Certain, Investors Should Focus on Three Main Lines
Looking ahead, the long-term logic for chip investment is clear and solid. From the demand side, in addition to AI, automotive electronics (over 1,000 chips per new energy vehicle), IoT (over 50 billion connected devices globally), and 5G/6G communications will continue to drive semiconductor demand. It is predicted that the global semiconductor market size will exceed $1.5 trillion by 2030.
For investors, three main investment lines deserve attention:
- AI Computing Core Track: Including GPU designers (NVIDIA, AMD), HBM memory (Samsung, SK Hynix), advanced packaging (TSMC CoWoS, ASE), benefiting from the AI infrastructure wave with high order visibility.
- Key Domestic Substitution Links: Focus on semiconductor equipment (NAURA, AMEC), materials (NSIG, Anji Microelectronics), and EDA software (Empyrean), with huge growth potential driven by dual drivers of policy and localization rate improvement.
- Power Semiconductors and Analog Chips: As electric vehicles and new energy installations continue to rise, demand for IGBTs, SiC devices, and power management chips is strong. Companies like Infineon, STMicroelectronics, and domestic players such as StarPower and Will Semiconductor are worth attention.
However, experts also remind investors of risks. While AI demand has reduced the semiconductor industry's cyclicality, uncertainties such as geopolitical conflicts, technology roadmap changes, and overcapacity still exist. It is recommended that investors adopt a "core + satellite" strategy, using index funds or ETFs (e.g., Philadelphia Semiconductor Index SOX, SGX-listed Nikko AM Semiconductor ETF) as core positions, supplemented by in-depth research on specific sub-sectors.
Conclusion
In the new AI-driven technological era, chips have become the "physical foundation" of the digital economy. Whether from the perspective of technology barriers, policy support, or long-term market demand, now is the strategic window for chip investment. Investors should seize the historic opportunity of the deep integration of AI and semiconductors, seeking certain growth opportunities amid industrial transformation.
