2026 Outlook: Capital Accelerates AI Deployment, Tech Giants Move Beyond Showcasing Technology“

By 2025, the artificial intelligence industry accelerated its evolution amid a convergence of technological breakthroughs, practical applications, and capital surges. This year witnessed technology integrating into reality through diverse forms, marked by the emergence of phenomenon-level applications like DeepSeek and the rise of humanoid robots to the forefront. The core logic of industry competition undergoes a fundamental shift: the focus moves entirely from competing on parameter scale to a contest of deep implementation capabilities in daily life and production scenarios among companies like Alibaba, Ant Group, ByteDance, Tencent, and Baidu.

Capital serves as both the accelerator and the litmus test for this process. From the concentrated listings of domestic GPU companies to Zhizhu and MiniMax racing to become the “first large model stock,” the industry is charging ahead with capital's backing. Yet beneath the feast, the harsh realities of “high hallucination rates, high power consumption, and high costs” coupled with “low user retention,” along with the still-unfinished commercial closed loop, continue to question the sustainability of every track.

Faced with challenges, “collaborative innovation” has evolved from an option to a necessity. As the dominance of any single technology becomes increasingly difficult to achieve, leading companies are increasingly choosing to pursue system-level synergy through open-source initiatives and ecosystem development. This structural shift will deepen by 2026, becoming a key driver for scaling and industrializing AI applications.

Large models are shifting from competing on parameters to competing on applications.

From completing payments with a single voice command, to AI health assistants offering professional advice, to robots autonomously collaborating on complex industrial operations... These increasingly common scenarios clearly outline the industrial landscape of AI's “breakout year” and “embodied implementation” in 2025. If 2023 marked the “awakening year” for generative AI and 2024 served as its “exploratory phase,” then 2025 will undoubtedly witness AI technology's comprehensive transition from laboratories to industrial applications—achieving a pivotal leap from isolated scenarios to comprehensive, all-encompassing penetration.

This year, artificial intelligence underwent a fundamental transformation across multiple dimensions. Language models evolved from mechanically piecing together words into “thinkers” capable of logical reasoning; video generation models moved beyond merely “looking like” reality to building “world simulators” that understand physical laws; and most breakthrough of all, AI truly “grew legs,” entering factories and homes in the form of humanoid robots, achieving a leap from the digital realm to the physical world.

With technological breakthroughs, the focus of industry competition has shifted from parameter scale to the breadth and depth of scenario implementation. ByteDance's “Doubao” has deeply integrated into high-frequency scenarios like short video creation and intelligent customer service, achieving cross-application task coordination through system-level partnerships with smartphone manufacturers. Tencent's “Yuanbao” has transformed into an all-day “personal assistant,” deeply embedded within ecosystems like Video Accounts and WeCom. Alibaba has successively launched multiple AI-native applications including Tongyi Qianwen, Ant Linguang, and Ant Afu. Among these, “Afu” leverages “doctor AI avatar” technology to make healthcare services more accessible; “Linguang” enables users to rapidly generate interactive lightweight applications through natural language, with user-created “Flash Apps” exceeding 12 million within just one month of launch.

The continuous expansion of application scenarios has further catalyzed innovation in niche sectors like smart hardware. With the deep integration of large models and spatial computing, the smart glasses field has rapidly evolved from a niche market dominated by startups into a “hundred-glasses battle” where tech giants converge. The entry of companies like Xiaomi, Lenovo, Baidu, and Alibaba is accelerating the industrialization process in this sector.

While rapid iteration continues at the application layer, competition over underlying model capabilities remains fierce. Alibaba's Tongyi Qianwen topped the global open-source model rankings with cumulative downloads exceeding 600 million; Doubao's large model surpassed 50 trillion daily token calls, with its multimodal agent capabilities deployed across education and industrial quality inspection scenarios; Tencent's Hunyuan large model released over 30 new models throughout the year; and Baidu's Wenxin large model iterated to version 5.0.

Qiyuan, President of the Shanghai Institute of Science and Intelligence, stated that by 2025, corporate competition in the AI sector will have transcended the race for computing power and parameters, entering a deeper phase of contesting value creation capabilities. Particularly in high-demand fields like finance and healthcare, technology must achieve a triple leap from “usable” to “trustworthy” and ultimately to “highly effective.” This demands that enterprises not only possess robust technical capabilities but also deeply understand industry contexts, regulations, and core pain points to truly establish differentiated competitive advantages.

Liu Xingliang, Director of the DCCI Internet Research Institute, pointed out that by 2025AI applicationsThe core trajectory of this evolution lies in the systematic advancement of agents. These intelligent entities have evolved from early-stage tools executing single commands into organic systems capable of autonomously planning and collaboratively executing complex tasks, thereby profoundly restructuring corporate workflows and decision-making mechanisms. In his view, within the next three to six years, expert agents—tailored for vertical scenarios and equipped with deep industry knowledge—will enter a phase of large-scale implementation. They will become pivotal drivers in elevating industrial efficiency and decision-making quality.

Capital Fuels Accelerated Development of AI

Beyond deep iterations in technology and applications, the collective rush of AI companies into capital markets by 2025 has emerged as a prominent trend throughout the year. With companies like Moore Threads and Muxi Technology going public, tech firms specializing in large models and embodied intelligence have sparked a wave of IPOs and financing rounds.

According to incomplete statistics, approximately 215 new companies went public by the end of 2025. Among them, the number of companies with AI-related businesses surged from 21 last year to 51, representing a 143% increase.

In terms of specific progress: In July, Yushu Technology commenced its STAR Market IPO advisory process and completed it in November; In August, Qunhe Technology submitted its listing application to the Hong Kong Stock Exchange, with J.P. Morgan and CCB International serving as joint sponsors. In December, two large-model companies, Zhipu and MiniMax, passed the Hong Kong Stock Exchange listing hearing, competing to become the “first large-model stock.” That same month, the China Securities Regulatory Commission's website showed that Yunshenchu Technology initiated its STAR Market listing guidance, with CITIC Securities acting as the guidance institution.

These companies share common characteristics: substantial R&D investments, extended development cycles, and significant early-stage losses. They heavily rely on capital market support to build technological barriers. Financial data corroborates this: Zhipu's net losses expanded from RMB 144 million in 2022 to RMB 2.958 billion in 2024, with losses projected to widen further. MiniMax recorded losses of approximately RMB 3.61 billion through Q3 this year, with cumulative losses exceeding US$800 million over the past three years. Both companies stated in their filings that the losses primarily stem from ongoing investments in large-model R&D and computing infrastructure, with profitability and dividend distribution unlikely in the near term.

Against this backdrop, the logic of capital screening has become increasingly clear: it involves not only evaluating the technological sophistication and core barriers of a company but also assessing its sustainable operational capabilities. This implies that AI enterprises ultimately prevailing in the competitive landscape will rely not only on technological prowess but equally on their fundraising capabilities and proficiency in capital operations.

Support from capital markets has further spurred strategic investments by enterprises. Alibaba stated it is advancing a 380 billion yuan AI infrastructure initiative with plans for additional funding; ByteDance preliminarily plans to invest 160 billion yuan in AI development by 2026; Tencent and Baidu, meanwhile, announced they will enhance their R&D systems by upgrading their R&D architecture and establishing dedicated departments.

At the policy level, the “AI Plus” initiative continues to deepen. According to data from the Cyberspace Administration of China, as of early November, 611 generative AI services had completed registration nationwide, with the monthly registration rate increasing by 551% compared to last year. The industry has entered a fast lane of standardized development.

Zhang Xiaorong, Dean of the Deep Technology Research Institute, pointed out that unlike last year's prospectus which described AI as “complementary to core business,” this year AI has become a tangible revenue stream and foundational infrastructure for enterprises. As these companies transition from the primary to secondary markets, the “AI content” among listed companies is projected to continue rising through 2026, signaling a new phase in the integration of capital markets and the AI industry.

Lowering the Barrier to AI Adoption: Collaboration is Key

Despite sustained capital inflows and persistent market fervor, the AI edge computing industry must still overcome multiple challenges—including technology, ecosystem development, and cost—to transition from concept to maturity. Ensuring models operate reliably, compliantly, and profitably in the real world has become the core focus of the industry's evolution in 2026.

An anonymous AI industry insider noted that while numerous products currently flood the market, only a handful have achieved true phenomenon-level adoption. Multiple constraints underlie this phenomenon: terminal devices struggle to guarantee smooth and stable user experiences due to limitations in computing power and power consumption; The application ecosystem remains in its infancy, with most features yet to address users' high-frequency daily needs. Model capabilities also show divergence: general-purpose models require improvements in logical comprehension, task execution, and cross-modal coordination, while specialized models need further optimization in professional depth, operational reliability, and scenario adaptability.

Xu Siyan, a senior researcher at Tencent Research Institute, further noted from the perspective of embodied intelligence implementation that while robots have gradually penetrated industrial, logistics, and service scenarios to undertake high-risk tasks, the industry still faces critical bottlenecks such as high costs, insufficient technological maturity, and a lack of real-world interaction data.

Facing challenges, tech companies are exploring differentiated paths to break through the ecosystem. By leveraging “traffic aggregation and distribution,” ByteDance deeply integrates AI into its content ecosystem, focusing on entertainment and information portals. Alibaba, meanwhile, positions AI as a productivity and lifestyle gateway by building around foundational infrastructure for e-commerce, finance, and local services, thereby constructing a commercial chain of “tool generation—service implementation—payment closure.” Meanwhile, players like Tencent and Baidu are accelerating the integration of generative capabilities into high-frequency scenarios such as payments, social networking, and mapping. Through ecosystem synergy, they are building user stickiness and competitive barriers.

Regarding the core logic of future ecosystem competition, Ant Group insiders believe that competitive barriers in the AI era have evolved into an intelligent flywheel of “data-model-ecosystem”: diverse scenarios and massive user data train more precise models, intelligent models enhance user experience and attract more users, while continuous user interaction feeds back into data and model iteration. When this flywheel integrates with a platform's commercial fulfillment capabilities (such as payments and offline services), the resulting competitive moat far exceeds the boundaries achievable by purely technology-driven companies.

Technological innovation is widely regarded as the key to breaking the cost dilemma and driving ecosystem expansion. Representatives from Volcano Engine pointed out that in the future, the core mission of the large model industry will not be internal competition, but rather working together to expand the market. Continuously optimizing technology to reduce costs, lower the barriers to AI adoption, and promote the inclusive development of AI—these are the fundamental prerequisites for ecosystem collaboration.

Alibaba representatives, taking a longer-term view of the industry landscape, believe that in the transition from AGI (Artificial General Intelligence) to ASI (Artificial Superintelligence), large models will gradually become the next-generation operating systems, while AI cloud will form the next-generation computing infrastructure. Based on this assessment, Alibaba has fully open-sourced Tongyi Qianwen with the goal of creating the “Android system” of the AI era, accelerating industry-wide collaboration through an open ecosystem.

Looking ahead to 2026, Zhang Xiaorong noted that ecosystem synergy will deepen further: Leading platforms will continue integrating full-scenario resources to drive generative capabilities deep into the entire ecosystem. Simultaneously, internal division of labor within the ecosystem will become clearer. Platform enterprises with traffic and technological advantages will forge closer collaborative relationships with physical industries possessing vertical data and offline scenarios, jointly defining new commercial rules for the AI era.

This article is from WeChat“Finance” (ID: mycaijing)By Shu Zhijuan, Edited by Gao Suying,