Abstract: GPU Cloud Strengthens the Computing Infrastructure, Powering the Global Expansion of AI Applications
BEIJING, Aug. 12, 2025 /PRNewswire/ — In recent years, the global AI market has experienced rapid expansion. According to Bain & Company, the global AI software & hardware market reached USD 185 billion in 2023, and is expanding at an annual growth rate of 40% to 55%. By 2027, the market is expected to surpass USD 780-990 billion, with AI applications alone accounting for over USD 407 billion. Amid this surge, Chinese AI application companies are accelerating their global expansion, leveraging technological breakthroughs across generations, experience in innovative domestic scenarios, and strong policy support. They are rapidly emerging as key players in the global AI ecosystem.
However, the global expansion of Chinese AI application companies is far from smooth. According to research by 36Kr Research Institute, these companies encounter three structural barriers: 52.7% report insufficient global computing infrastructure deployment, leading to high latency and low data collaboration efficiency across regions; 52.0% struggle with high costs and long cycles of cross-border payments and settlements, which limit cash flow and profitability in international markets; and 44.3% face limited global marketing channels, making it difficult to overcome cultural barriers and achieve precise user acquisition. Among these, computing infrastructure— the backbone of AI applications—directly affects model training efficiency, inference response speed, and service coverage, making it a critical variable that determines the success or failure of global expansion.
For AI application companies to succeed in their global expansion, they should establish a closed loop of “computing infrastructure, market penetration, and monetization”. Among these, enhancing computing infrastructure stands as the primary breakthrough. At the same time, delivering scenario-driven, customized computing infrastructure—tailored to the unique needs of different industries and regions—will be essential for achieving differentiated competitiveness and securing a solid foothold in overseas markets.
Against this global backdrop, 36Kr Research Institute conducted a survey of 700 Chinese AI application companies expanding globally, and released the 2025 Insight Report on the Development Demands of Chinese AI Application Companies Expanding Globally (hereinafter referred to as “the Report”). The Report offers a comprehensive analysis of the current state, core demands, and future trends of AI application companies expanding globally. It focuses on seven high-growth sectors, including AI productivity tools, emotional companions, audio & video generation, education, gaming, AI terminals, and embodied artificial intelligence, and unpacks their diverse computing capacity needs across the training and inference phase. By doing so, the report provides a valuable basis for Chinese AI application companies developing their global expansion strategy.
1. Computing Infrastructure: Globalized, Elastic, and Scenario-Driven Infrastructure in Parallel
With the widespread global adoption of AI applications, enterprises are experiencing an explosive surge in demand for computational infrastructure. According to the Report, over 70% of surveyed companies allocate more than 10% of their R&D budgets to computing infrastructure, and inference-related demand is increasing by more than 70% annually. These figures clearly highlight that computing infrastructure has become the “most fundamental and urgent need” driving the growth of AI application companies.
However, the surge in computing infrastructure demand has also brought a range of challenges for AI application companies. High latency for global user access (58.7%) often leads to lag, slow response times, and degraded user experiences when accessing AI applications; low efficiency in cross-regional data collaboration (57.0%) limits companies’ ability to integrate global resources and streamline operations; insufficient capacity for computing resource orchestration (52.3%) makes it difficult to handle sudden spikes in user traffic, resulting in underutilized or overloaded computing resources; and significant cost pressure related to computing infrastructure (42.3%) continue to erode profit margins, becoming a major concern for companies seeking sustainable global expansion.
To address these challenges, AI application companies are actively exploring new computing solutions. GPU cloud, with its advantages of rapid deployment, elastic scaling, and pay-as-you-go pricing, has become a key component for global expansion plans of AI application companies. Currently, 87% of surveyed companies rely on GPU cloud services to support their global businesses, while only 1.0% have not adopted them. The core driver behind adopting GPU cloud lies in solving inference deployment challenges through cloud-based computing resources. 60.0% of companies value the cloud’s cluster management and resource scheduling capabilities, using dynamic compute resource orchestration to manage cross-regional traffic loads. 51.0% prioritize global node coverage, enabling low-latency, multi-region inference deployment to meet the demands of real-time user interaction. This underscores the importance of GPU clouds which have evolved beyond merely being a computing resource. It now functions as an integrated solution for technical coordination, cost optimization, and global adaptability, making them a critical path for overcoming computing bottlenecks in the global expansion of AI applications.
Chinese AI application companies expanding globally consider a combination of factors—cost, technical support, efficiency, and compliance—when selecting computing infrastructure providers. Survey data shows that: 59.6% of companies prioritize cost competitiveness, 58.7% value technical support and O&M guarantee, and 58.3% focus on product and service delivery efficiency, demonstrating companies’ pursuit of a balance between cost reduction and high-efficiency responsiveness in their computing infrastructure investments. Meanwhile, demand for global computing resource deployment (45.3%) and compliance certifications (31.0%) reflects the growing need to adapt computing services to varied regional regulatory environments.
Notably new providers are beginning to gain traction alongside major players like Alibaba Cloud, Google Cloud, and Amazon Web Services (AWS). GMI Cloud was the 3rd most considered provider with 36.3% of surveyed companies including it in their evaluation. GMI Cloud’s rise suggests ongoing shifts in provider dynamics across core international markets.
Scenario-based customization is a key direction for computing solutions. The Report’s analysis across seven high-growth sectors reveals that while a broad technological consensus has formed around key capabilities, namely “low-latency interaction, elastic scheduling, heterogeneous computing resources collaboration, and global compliance readiness,” demand varies significantly by application scenario. During the training phase, AI productivity tools focus on multimodal fusion, while emotional companion emphasizes cross-cultural emotion recognition and character persona fine-tuning. Audio & video generation models require multi-device compatibility, and low-resource language data augmentation. In the education domain, attention is directed toward cross-age model adaptation and low-resource language annotation. Gaming prioritizes high-fidelity rendering and cross-platform debugging. AI terminals stress multi-device collaboration and model lightweighting, whereas embodied artificial intelligence targets safe and compliant simulation environments and sensor data annotation. During the inference phase, AI productivity tools aim for low-latency real-time interaction. Emotional companions require 24/7 dynamic response, low-power optimization. Audio & video generation models prioritize Real-time stream processing, burst traffic scaling-up. In education, the focus shifts to real-time teaching interaction and compliance auditing. Gaming demands millisecond-level real-time interaction and optimization of graphical computing capacity. AI terminals should support millisecond-level on-device response and device-cloud collaborative inference. Embodied artificial intelligence emphasizes microsecond-level motion control and on-device real-time decision-making. Given these scenario-specific demands, companies should develop tailored computing solutions to ensure efficient operation of AI applications.
In terms of cost optimization, technologies such as mixed-precision inference and heterogeneous computing resources collaboration are being increasingly adopted. These approaches significantly reduce computing costs, enabling companies to maximize economic efficiency without compromising on computing performance.
2. Market Penetration: Social Media-Driven, AI-Enabled, and Localization-Deepened Strategies
In global marketing, social media operations (63.0%), channel-led market access (61.7%), and localized content marketing (60.3%) have become the primary channels for AI application companies to acquire new users. Social media platforms, with their massive user bases and broad content reach, offer companies vast marketing opportunities; channel-led market access leverages local partners’ resources and influence to help companies gain rapid market access; and localized content marketing strengthens brand affinity by delivering content tailored to local user needs, grounded in a deep understanding of regional culture, consumer behavior, and preferences.
However, companies still face significant capability gaps in global marketing. High costs of social media operations (64.0%) remain a key challenge, as companies must invest heavily in manpower, tools, and budgets to manage accounts, promote content, and maintain active user engagement. Insufficient audience segmentation (57.7%) makes it difficult for companies to accurately identify target customer segments, significantly undermining the effectiveness of their marketing efforts. Meanwhile, limited visibility into advertising ROI (57.3%) prevents companies from accurately evaluating the impact of their ad placement, making it challenging to optimize marketing strategies.
To bridge these capability gaps, AI technologies are increasingly becoming a powerful enabler for marketing penetration. 67.7% of surveyed companies expect AI to enhance social media sentiment monitoring. With AI technologies, companies can track brand-related sentiment on social media in real time, enabling early detection of negative signals and timely crisis response. 57.0% of companies seek AI-driven intelligent ad placement. By analyzing user behavior and interest patterns, AI enables highly targeted ad placement, ultimately improving conversion rates. In addition, automated multilingual content generation has become a key enabler in overcoming bottlenecks in localized content production. AI technologies allow companies to efficiently create marketing content across multiple languages, enabling them to meet the needs of diverse regional audiences and significantly enhance marketing efficiency.
3. Monetization: Compliance First, Efficiency Upgraded, and Ecosystem Integrated
As a critical link in completing the commercial loop of global expansion, cross-border payments remain fraught with challenges. Complex compliance reviews (61.3%) pose a significant challenge, as varying payment regulations and policies across countries and regions require companies to dedicate substantial time and resources to ensure cross-border payment compliance. Limited multi-currency settlement capabilities (54.0%) limit companies’ ability to operate globally, as they fail to meet the diverse payment preferences of users across different regions. Meanwhile, exchange rate volatility (51.7%) also introduces uncertainty to corporate revenue streams.
In response to these cross-border challenges, companies have expressed clear core demands. 65.0% seek one-stop compliance management, aiming to streamline compliance processes and reduce regulatory risks by integrating policies across different countries and regions. 57.7% require real-time financial instruments, such as foreign exchange hedging instruments, to stabilize revenue expectations and mitigate exchange rate volatility risks. 24.7% prioritize localized payment adaptation, supporting regional mainstream payment methods including digital wallets, to improve the user payment experience. Only by addressing these core demands can companies establish an efficient and stable cross-border payment loop, providing strong support for seamless and sustainable global business operations.
4. Data-Driven Insights Rooted in Real-World Practice: A Practical Guide for Diverse Stakeholders
Unlike reports that focus solely on macro-level analysis or a single technical trajectory, the 2025 Insight Report on the Development Demands of Chinese AI Application Companies Expanding Globally delivers practical, field-tested insights grounded in real-world operations. Based on in-depth research involving 700 frontline Chinese AI application companies expanding globally, the Report goes beyond theoretical modeling to provide a systematic overview of actual market demands. In addition to rich data and comprehensive analysis, the Report delivers differentiated value to a wide range of stakeholders, including corporate decision-makers, AI technology R&D and engineering teams, investors, and industry researchers.
For corporate decision-makers, the Report outlines a closed loop of “computing infrastructure, market penetration, and monetization”, with computing infrastructure identified as the primary point of leverage. Grounded in real-world data from 700 Chinese AI application companies expanding globally, the Report offers decision-makers key trend insights, such as the adoption rate and strategic value of GPU cloud, and the growing importance of scenario-specific customization, as well as resource allocation priorities, including computing investment ratios and cost optimization strategies. In addition, the Report provides an in-depth breakdown of differentiated computing infrastructure demands across the training and inference phases in seven high-growth AI sectors, along with practical recommendations for scenario-based computing resources deployment strategies. These frontline insights serve as a critical foundation for crafting targeted global expansion strategies and mitigating core risks, ultimately helping decision-makers build robust, globally competitive AI businesses.
For AI technology R&D and engineering teams responsible for execution, the value of the Report lies in its high degree of scenario adaptability and technical guidance. It not only identifies the core technical challenges faced across different sectors, but also provides targeted recommendations for computing resources configuration and optimization strategies. The Report further explores cutting-edge trends such as scenario-defined infrastructure, software-driven hardware value realization, and multimodal fusion, offering fresh perspectives for technology selection and system architecture evolution. These insights empower technical and engineering teams to enhance product performance and user experience with greater precision and relevance.
For investors and industry researchers, the Report offers a comprehensive view of the global expansion landscape and key drivers behind the Chinese AI application ecosystem. It quantifies market opportunities, pinpoints core challenges, and reveals the strategic paths taken by leading players. The in-depth analysis of divergent development trajectories and computing demands across the seven major AI sectors provides valuable benchmarks and forward-looking perspectives for identifying high-potential sectors, assessing technical barriers, and evaluating the sustainability of business models. Moreover, the Report’s discussion of infrastructure optimization strategies, including solutions such as GMI Cloud, also highlights the critical service layers that underpin the entire global expansion ecosystem.
It is important to note that AI technology evolves at a much faster pace than traditional sectors, and both application formats and user demands can shift significantly within a short time frame. The Report centers its core value on supporting decision-making during the next 1–2 years—a critical strategic window for global expansion. Regardless of how technologies evolve, the demand for efficient, stable, and cost-effective computing infrastructure remains a pillar underpinning the global scalability of AI applications. The Report’s in-depth analysis of computing infrastructure challenges, its evaluation of solutions such as GPU cloud, and its emphasis on scenario-specific deployment collectively address this enduring strategic imperative.
Together, these insights provide companies with a foundational logic for building long-term global competitiveness—a logic that remains robust and relevant, even as technologies continue to evolve.
These frontline insights offer solid data support and practical guidance to help companies craft informed globalization strategies, empowering them stand out amid fierce international competition and succeed in expanding their AI applications globally.
The full report, including raw data and detailed insights into computing requirements across the seven featured sectors, is available for download here.
