CITIC SEC has released its investment strategy for the technology sector in the second half of 2026. The most significant challenges currently are the elevated valuations of global technology stocks and crowded positioning. Additionally, the potential listings of SpaceX, Anthropic, and OpenAI within the next six to twelve months could impact market liquidity, representing a major event and observation window for global tech equities. Considering that macro liquidity in the latter half of this year may tighten marginally compared to the second half of 2025, market performance will rely more heavily on the continuous delivery and verification of corporate earnings. However, as long as the positive industry trends remain intact, any valuation corrections triggered by liquidity shocks present opportunities that outweigh the risks.
The firm maintains a positive outlook on several key areas:
1) Investment opportunities in foundational model companies driven by model iteration, application proliferation, and sustained growth in Annual Recurring Revenue (ARR).
2) Semiconductor equipment companies benefiting from the ongoing capacity expansion in domestic memory and advanced process nodes, along with increasing localization rates, as well as advanced process companies where breakthroughs in advanced technology are anticipated and capacity utilization rates continue to rise.
3) Leading companies in cloud computing, storage, optics, CPUs, PCBs, power supplies, and upstream sectors that benefit from the explosion in computing demand and sustained price increases, and which possess robust supply chain capabilities.
4) Companies in the embodied robotics and autonomous driving sectors where technology is rapidly iterating, mass production is imminent, and they are entering a capital investment window.
5) The allocation value of undervalued internet leaders among the constituent stocks of the Hang Seng Tech Index.
6) Investment opportunities in AI services and data governance companies arising from the enterprise-level implementation of AI.
CITIC SEC's primary views are outlined below:
**Market Review**
The profit-driven characteristic of the technology sector is established, but both absolute valuations and capital allocations are at high levels. According to Wind data, the STAR 50 Index has gained 31.69% year-to-date, significantly outperforming the Shanghai Composite Index's 6.9% rise, with a 25.05% surge in April alone leading the market. Comparatively, on the hardware side, electronics and communications have risen 41.17% and 52.39% year-to-date, respectively. On the software side, computer and media sectors have increased by 8.63% and 1.46%, respectively. Progress in AI Coding has consistently exceeded market expectations, driving a sustained rally in technology stocks. Since 2026, the gains in the STAR 50 and Nasdaq indices have been primarily driven by systematic upward revisions in EPS, with P/E ratios increasing only marginally by 4% and remaining stable, respectively. The valuation expansion narrative has shifted to one of profit realization. From a trading perspective, TMT currently accounts for 36.4% of active equity fund holdings. On the earnings front, the net profit for the CITIC electronics industry in Q1 2026 grew 47.4% year-over-year, with leading communication companies also reporting strong growth.
**Computing Power**
Domestic models and chips represent the most significant opportunity for growth, while overseas demand remains robust.
1) The localization of computing power will be the segment with the greatest elasticity. The expansion of advanced logic and memory production is accelerating in 2026, with the localization rate for equipment expected to rise from 30% to 40%. Full-year order growth expectations for memory equipment companies have been revised upward to 50%. Concurrently, a surge in Token usage is exacerbating computing power shortages both domestically and internationally. Domestic large language models are actively adapting to local chips like Ascend and Cambricon, with the "domestic models and domestic chips" trend driving accelerated adoption of domestic computing power. Overall, the expansion of domestic wafer fabs, increased localization rates, and the explosion in AI computing demand form a triple catalyst, with prosperity expected to continue rising across the domestic supply chain.
2) Regarding computing demand: Driven by AI Agents in 2026, the rapid explosion in global Token usage is bringing sustained incremental growth to capital expenditure (Capex). North American demand remains strong: Q1 revenues accelerated for the three major North American cloud providers, with Google Cloud Platform revenue surging 63% year-over-year. Meta, Google, and Amazon have raised their 2026 Capex guidance to $125-145 billion, $180-190 billion, and approximately $200 billion, respectively. In contrast, the absolute value and growth rate of 2026 Capex for domestic cloud providers Alibaba, Tencent, and Baidu are significantly lower than their North American counterparts. Their Capex-to-revenue ratio is only 11%-17%, indicating ample payment capacity and significant room for future Capex increases.
3) Overseas computing power demand is expected to persist, with price increases continuing to spread. Supply-demand dynamics are tightening further in core industry segments such as optical modules, storage, and PCBs. Leading companies are reporting strong earnings growth, and the binding power of long-term agreements is strengthening. The increase in the absolute scale of the AI market has driven rising prosperity and a trend of capacity shortages spreading upstream.
**Models and Applications**
The frequency of model iterations is increasing, and Chinese model developers are accelerating their catch-up efforts. A significant rise in Anthropic's Token usage has driven its rapid revenue growth, with ARR exceeding $44 billion as of May 2026. Meanwhile, in Q1 2026, model developers both domestically and abroad broke the quarterly iteration pattern. OpenAI consecutively iterated its GPT-5 series, while Google and Anthropic simultaneously launched flagship versions like Gemini 3.1 and Claude Opus 4.6/4.7. Domestically, major versions such as DeepSeek V4, Qwen3.6, and Doubao 2.0 were released in quick succession, intensifying SOTA competition. Looking ahead, Chinese model developers will accelerate their追赶 efforts. As AI application implementation deepens and the digital intelligence level of Chinese enterprises increases, the deeper integration of Chinese models with industry will create greater industrial value. Concurrently, new themes and sub-directions such as Physical AI, World Models, and video generation models are entering the early stages of commercialization, potentially opening new commercial waves for AI beyond foundational models.
**Risk Factors**
Geopolitical changes could prevent mainland China's IC design companies from using advanced process capacity at overseas foundries for tape-outs. Furthermore, observed AI applications have only generated economic benefits in relatively limited domains. Larger-scale applications depend on the integration process of AI large models with existing digital infrastructure. Corporate investment in AI may therefore be subject to cyclical fluctuations, prompting investors to fully consider the risks this investment cycle poses to technology industry investments.
CITIC SEC also highlights risks including a slower-than-expected pace of macroeconomic recovery, potential shortfalls in related industrial policies, slower-than-anticipated progress in core technology and product R&D, slower-than-expected AI application落地速度, and capital expenditure by cloud service providers falling short of expectations.
Comments