Looking ahead to 2026, the AI industry is expected to remain one of the most central narrative threads in the market. In 2025, the A-share market witnessed a structural rally driven by industrial trends, where market consensus progressively deepened and transcended historical frameworks. Sectors like humanoid robots, new consumption trends, innovative pharmaceuticals, artificial intelligence, and commercial spaceflight took turns outperforming, collectively composing the main theme of "high-quality development." For 2026, we anticipate the AI industry will likely continue to be a core storyline.
AI models are poised to enter a new phase of explosive innovation. Throughout 2025, AI models continued to iterate and push the boundaries of their capabilities, even amidst industry skepticism about perceived "technological bottlenecks." Google's latest release of Gemini 3.0 confirmed that pre-training technology still holds considerable potential, while the domestic representative model DeepSeek V3.2 demonstrated the unique value of post-training pathways. Looking to 2026, we expect major model developers globally to further learn from each other and advance together, engaging in a new round of innovative competition centered on compute deployment, algorithm optimization, and the closed-source versus open-source paths. The success of Google's Gemini 3.0, in particular, has created significant catch-up pressure for other leaders, including OpenAI. The year 2026 may welcome a series of landmark flagship model releases, with model capabilities potentially evolving from text, voice, and image processing to genuine native multimodal input and output, offering the world entirely new interactive and cognitive experiences.
AI compute infrastructure investment is expected to continue its vigorous advance. Since 2023, despite intermittent market concerns about an AI industry "bubble," global tech giants have maintained robust and deepening investments in AI. This wave of AI infrastructure investment centers on accelerated computing, exhibiting distinctly different characteristics from previous digital infrastructure across key areas like computing, networking, storage, power, and cooling. Furthermore, the technological pathways are still evolving rapidly, presenting a novel challenge for the global industry. For 2026, close attention is needed on potential constraints and risks across various segments, including semiconductors, power supply, financing, and project delivery. Simultaneously, the high enthusiasm for AI investment from global tech enterprises and "sovereign AI" initiatives should be fully acknowledged. Overall, this cycle of AI infrastructure investment possesses long-cycle industrial planning attributes; therefore, it is premature to draw definitive conclusions about its short-term performance, and the industry should be granted ample time for development and innovation.
Regarding AI applications, while progress has been made over the past three years, it has fallen significantly short of the optimistic expectations held by market enthusiasts, contributing to the recurring "bubble" narrative. The market initially hoped for a virtuous cycle between compute, models, and applications. However, progress on the application side has notably lagged behind massive capital investment, and the anticipated positive feedback loop is not yet fully established. We believe AI applications are now showing initial signs of "breaking through the soil and growing." AI-native companies, represented by the likes of OpenAI and Anthropic, have demonstrated strong commercialization potential through rapid increases in Annual Recurring Revenue (ARR) and overall revenue. It has been years since we have seen revenue curves with such clear scale and vigorous growth, potentially signaling the dawn of a new era for AI applications.
Currently, many highly anticipated AI applications have yet to achieve widespread adoption. We judge the primary reasons likely stem from two aspects: First, the overall intelligence level of current models remains limited, with text processing being the most mature capability, hence faster progress in scenarios like chatbots and code generation. The anticipated breakthrough in multimodal capabilities by 2026 could become the fertile ground for incubating more innovative AI applications. Second, compute costs remain relatively high; lighter versions (flash versions) of various models are currently more popular in the market, while more powerful professional versions (Pro versions) still carry a premium price. This issue is expected to gradually ease with the continued accumulation of compute resources and further improvements in algorithm efficiency.
Looking towards 2026, we will closely monitor the evolution of industry trends and continuously optimize our product management system. We aim to respond to market changes with a pragmatic and forward-looking approach, striving to create long-term, sustainable value returns for investors.
First, we adhere to a global perspective. Although our investments primarily focus on the Chinese stock market, we will maintain close tracking of global technology companies. We will continue refining our investment research framework based on the global industrial chain, analyzing the competitive advantages, growth potential, and valuation systems of Chinese enterprises from a global viewpoint. By combining the fundamental logic of global value chain division and capital market pricing, we will conduct multi-dimensional assessments of the value and risks associated with listed companies. Compared to the significant effort devoted to outbound industrial chains over the past three years, in 2026 we will invest more energy into deeply exploring structural opportunities within the domestic industrial chain. We believe the domestic market also holds significant participation value and development space within this wave of industrial trends.
Second, we persist with a value investment perspective. In our investment selections, we favor enterprises with sustainable growth potential, paying particular attention to leading technology companies that possess global competitiveness, a pioneering spirit, and the ability to consistently create social value. We expect these companies to win market and customer trust through excellent products and services, and to reward shareholders with solid profits. We are consistently committed to allocating capital to outstanding enterprises that genuinely create long-term value for China and the world, while diligently avoiding speculative targets that might erode value or rely solely on hype.
Third, we insist on keeping pace with the times. In 2026, we will approach new market opportunities and challenges with an open and pragmatic mindset. As we continuously refine our investment research system, the goal is to transform historical accumulation into the compound interest value of our research, rather than letting it become shackles that hinder progress. Reflecting on the past three years, we effectively captured investment windows during the phased upswings in the AI compute industrial chain. Facing the new year, we believe the market will continue to nurture abundant structural opportunities, and new industrial narratives will constantly emerge. In response, we will actively embrace future changes and challenges.
Fourth, we maintain a balance between risk and return. In 2026, while pursuing investment returns, we will always place risk control in a more prominent position. We will pay particular attention to companies' ability to deliver on their performance promises and carefully consider the risks underlying potential returns. The current Chinese capital market already boasts a cohort of high-quality enterprises; therefore, we will not expend research resources on companies that rely solely on narratives but lack substantive support. Moving forward, the investment research team will focus on comprehensively assessing the potential returns and risks of investments from multiple dimensions, including industry trends, product competitiveness, order fulfillment, and performance delivery. We aim to make prudent allocations when the risk-reward ratio presents clear attractiveness.
Comments