Alibaba Cloud revenue surged 13% to 31.7 billion, driven by AI, returning to double-digit growth

Author | Liu Baodan

Editor | Zhou Zhiyu

The wave of computing power revolution sparked by DeepSeek is sweeping the globe, with cloud vendors being the biggest beneficiaries. At a critical moment for the revaluation of AI assets in China, Alibaba Cloud's financial report serves as a significant barometer as the largest cloud vendor in the country.

On the evening of February 20, Alibaba Group announced its new quarterly results for the quarter ending December 31, 2024, with Alibaba Cloud achieving revenue of 31.742 billion yuan, a year-on-year increase of 13%, nearly doubling the growth rate of 7% from the previous quarter. This growth is attributed to the increase in public cloud revenue driven by AI-related products, with AI revenue experiencing triple-digit growth for six consecutive quarters.

Alibaba Cloud will further invest in customer growth and technological innovation, particularly in AI infrastructure, to increase the adoption rate of AI in the cloud and maintain its market leadership.

Since the beginning of the year, Alibaba's stock price has risen by as much as 57%, reaching a three-year high, injecting a strong dose of confidence into the capital market. Alibaba's market value has surged by over 600 billion Hong Kong dollars, with the latest market value reaching 2.3 trillion Hong Kong dollars (approximately 2.17 trillion yuan).

On February 13, JP Morgan analysts conducted an in-depth analysis of Alibaba Group's valuation, believing that its stock price still has room for further increases, mainly benefiting from three factors: an upward adjustment of Alibaba Cloud's valuation multiple, an upward adjustment of cloud business revenue expectations, and an improvement in profitability expectations for China's e-commerce business.

The AI era has begun, and Alibaba Cloud, as a builder of China's AI infrastructure, has started to lead this historically unprecedented business challenge characterized by technological disruption, marking an unparalleled journey ahead.

Promoting Model Open Source and AI Popularization

Recently, Alibaba Cloud's Tongyi Qianwen flagship model Qwen2.5-Max was newly upgraded and released. The Qwen2.5-Max model is the latest exploration result of Alibaba Cloud's Tongyi team on the MoE model, with pre-training data exceeding 20 trillion tokens. The new model demonstrates exceptionally strong overall performance, achieving high scores on multiple mainstream public model evaluation benchmarks, surpassing the currently leading open-source MoE models and the largest open-source dense models globally.

Alibaba Cloud is the earliest and only "big company" in China to open-source its self-developed large models, and it is also the only cloud computing vendor globally that actively develops advanced AI models and fully open-sources them, achieving "full-size, full-modal, multi-scenario" open-source ahead of the industry.

In November last year, Alibaba Cloud open-sourced the entire series of Tongyi Qianwen code models, including six Qwen2.5-Coder models. Relevant evaluations show that all six code models achieved the best results in the industry at the same size, with the flagship code model of size 32B achieving the best open-source results in over a dozen benchmark evaluations, becoming the strongest open-source code model globally.

In the same month, Alibaba Cloud successively released and open-sourced the new AI inference model QwQ-32B-Preview and the multi-modal inference model QVQ-72B-Preview. Multiple evaluation data indicate that QVQ has surpassed the previous visual understanding model "open-source king" Qwen2-VL, with overall performance comparable to "full-blooded" inference models like OpenAI o1 and Claude3.5 Sonnet In January of this year, Alibaba Cloud Tongyi launched the open-source Qwen2.5-1M model, supporting a context of 1 million tokens, with two sizes: 7B and 14B, both of which stably outperform GPT-4o-mini in handling long text tasks; at the same time, an open-source inference framework was introduced, achieving nearly 7 times the speed when processing long text inputs at the million-level. In the same month, Alibaba Cloud Tongyi Qianwen released a new visual model Qwen2.5-VL, with three size versions: 3B, 7B, and 72B. Among them, the flagship version Qwen2.5-VL-72B won the visual understanding championship in 13 authoritative evaluations, surpassing GPT-4o and Claude3.5 comprehensively.

Alibaba Cloud hopes to promote the widespread use of Tongyi Qianwen among a large number of small and medium-sized enterprises and AI developers through open-source, accelerating the inclusive application of large model technology and facilitating the construction of China's large model ecosystem. Currently, the number of derivative models of Qwen in the domestic and international open-source community has exceeded 90,000, surpassing the Llama series of derivative models, making Tongyi Qianwen Qwen the world's largest generative language model family. According to the latest global open-source large model ranking by Huggingface on February 10, 2025, all of the top ten open-source large models are derivative models developed based on the open-source model of Tongyi Qianwen Qwen.

Continuously Increasing Investment in AI Infrastructure

Unlike the traditional IT era, the AI era demands higher performance and efficiency from infrastructure, with the CPU-dominated computing system rapidly transitioning to a GPU-dominated AI computing system. Alibaba Cloud is centering on AI, comprehensively reconstructing underlying hardware, computing, storage, networking, databases, and big data, organically adapting and integrating with AI scenarios, accelerating model development and application, and building the strongest AI infrastructure for the AI era.

At the infrastructure level, Alibaba Cloud launched the Panjiu AI server, which supports up to 16 cards per machine, a CPFS file storage system with a data throughput of 20TB/s, and a high-performance network architecture HPN7.0 that supports stable connections for over 100,000 GPUs.

At the platform level, Alibaba Cloud ACS introduced GPU container computing for the first time, enhancing computing affinity and performance through topology-aware scheduling. Alibaba Cloud's artificial intelligence platform PAI achieved integrated elastic scheduling for training and inference at the level of 100,000 cards, with effective utilization of AI computing power exceeding 90%.

Through full-stack optimization, Alibaba Cloud has built a stable and efficient AI infrastructure, with continuous training effective duration greater than 99% and model computing power utilization increased by over 20%.

At the same time, Alibaba Cloud Bailian has integrated over 200 large model APIs, covering mainstream open-source and closed-source models both domestically and internationally. To further reduce the cost of using text large model APIs for users, Alibaba Cloud Bailian has also launched a new KV Cache billing model, which can avoid redundant calculations by automatically caching contexts, further lowering model invocation costs, suitable for long texts and code completion In scenarios such as multi-turn dialogue and specific text summarization, over 700,000 enterprises and institutions, including FAW, Kingsoft, Hello Group, and the National Astronomical Observatory, are currently using Alibaba Cloud's Bai Lian.

At the end of 2022, with the initial wave of large models, Alibaba Cloud's MaaS concept, which was proposed first in the industry, quickly developed into an industry consensus. The MoDa community, as one of the best practices of the MaaS concept, rapidly grew into the largest AI model community in China. "Find large models on MoDa" has become a tacit understanding among AI developers. As of mid-February 2025, the total number of models in the MoDa community exceeded 40,000, covering various fields such as LLM, dialogue, voice, text-to-image, image-to-video, and AI composition, serving over 10 million developers.

AI-Driven Continuous Growth

Relying on a full-stack advanced AI cloud infrastructure and the Qianwen large model, Alibaba Cloud is accelerating its integration into various industries. After serving the vast majority of clients in the pan-internet, automotive, retail, education, and research fields, Alibaba Cloud continued to expand rapidly in finance, electricity, energy, transportation, cultural media, and consumer electronics this quarter, engaging in extensive cooperation with State Grid, China Central Radio and Television, China Merchants Bank, China Pacific Insurance, Hangzhou Metro Group, Meizu, and Thunderbird around deep cloud usage and AI application implementation.

In addition, to better support Chinese enterprises going global and MNCs (multinational companies), Alibaba Cloud has also accelerated the layout of global infrastructure and the construction of security and compliance capabilities, engaging in deep cooperation with Midea, Transsion, Xiaohongshu, Paper Games, Tuya Smart, Adidas (China), and General Mills (China).

China Merchants Bank X Alibaba Cloud

China Merchants Bank has widely applied Alibaba Cloud's Qianwen large model in multiple business scenarios, including the intelligent investment research assistant "Zhaoyin Zhiku AI Xiaoyan," intelligent customer service, general office applications, and a bank-wide knowledge base. For example, the intelligent investment research assistant "AI Xiaoyan" relies on the bank's largest research resource platform "Zhaoyin Zhiku," utilizing Alibaba Cloud's Qianwen large model to build site-wide intelligent Q&A robots, Chat research reports, and hot topic focus functions, solving issues such as difficulty in information retrieval, complexity in report extraction, and time-consuming analysis in traditional investment research processes. It achieves precise understanding and generation of financial semantics, significantly improving query efficiency and investment research accuracy, and automatically generates hot topic summaries and professional popular science knowledge, greatly enhancing work efficiency and decision quality, becoming an important innovation in the fintech field.

China Pacific Insurance X Alibaba Cloud

On December 23, 2024, China Pacific Insurance's new generation distributed core system was fully launched, with a scale and complexity rarely seen in the industry, and the results were remarkable, fully validating the feasibility of the internet cloud-native technology route in the insurance industry. On the same day, Alibaba Cloud and China Pacific Insurance reached an agreement on comprehensive digital intelligence upgrade cooperation, which will involve deep cooperation in intelligent applications, exploring foundational models and model management platforms, intelligent computing infrastructure, and disaster recovery system construction, providing customers with a more personalized service experience.

Xingji Meizu X Alibaba Cloud

On December 3, 2024, Alibaba Cloud and Xingji Meizu announced a comprehensive cooperation, with both parties promoting the implementation of cloud computing and Qianwen large model products and services in multiple fields such as mobile phones, smart wearables, and smart cars, to provide users with better product and service experiences RayNeo X Alibaba Cloud

On January 2, 2025, consumer-grade AR leading brand RayNeo held a strategic signing ceremony with Alibaba Cloud. Alibaba Cloud designed an overall multimodal interaction architecture based on the Tongyi series large model foundation, customizing the development of "cloud + end" models to help products achieve high responsiveness, low latency, and high-quality multimodal AI interaction experiences. At the same time, both parties will also carry out functional customization and development around the actual usage scenarios of the glasses products, and continue to iterate and update on RayNeo's AI glasses products.

State Grid X Alibaba Cloud

In December 2024, State Grid Corporation of China launched the "most comprehensive power knowledge, largest parameter scale, and strongest professional capability" billion-level multimodal industry large model—Guangming Power Large Model—in Beijing, with technical support from Alibaba and others for the large model and development platform. This model serves as an artificial intelligence "expert" in the energy and power sector, providing a "super brain" for the safe and stable operation of the power grid, promoting the consumption of renewable energy, and ensuring quality power supply services. At the same time, State Grid signed a strategic cooperation framework agreement with Alibaba Group to jointly build the Guangming Power Large Model and promote the integrated development of energy and power technology innovation and industrial innovation.

Hangzhou Metro X Alibaba Cloud

In November 2024, Hangzhou Metro Group and Alibaba Cloud Intelligence Group held a strategic cooperation agreement signing ceremony at Alibaba Cloud Valley Park. Both parties will focus on further cooperation in exploring the value of large models and data in rail transit, as well as innovative applications in rail transit logistics, achieving a higher level of win-win cooperation.

Accelerating Global Layout

In recent years, Alibaba Cloud has experienced rapid growth in overseas markets, with a market scale increasing more than 20 times in five years. As the market scale expands, Alibaba Cloud announced last May that it would increase investment in overseas cloud infrastructure construction, planning to build new data centers in five countries worldwide, located in Thailand, South Korea, the Philippines, Malaysia, and Mexico.

On February 13, Alibaba Cloud announced the official launch of its second data center in Thailand, which will provide stronger cloud computing resources for local enterprises, Chinese outbound companies, and local developers to meet the growing demand from emerging sectors such as the internet and fintech in Thailand.

On February 19, Alibaba Cloud announced the official opening of its data center in Mexico, marking its first data center in the Latin American region and an important step into South America. The first batch of products launched at the Alibaba Cloud Mexico data center includes core cloud computing capabilities such as elastic computing, cloud storage, network services, and container technology, with core product prices lower than those of other mainstream international cloud vendors to meet the rapid demand for cloud services across various industries in the Latin American market. In the future, Alibaba Cloud will continue to increase investment in local market ecosystem and service system construction.

Transsion X Alibaba Cloud

Transsion Holdings (hereinafter referred to as Transsion) has reached a cooperation agreement with Alibaba Cloud, with the Tongyi Qianwen large model already integrated into the AI phone PHANTOM V Fold2 launched by Transsion's technology brand TECNO, creating a deeply localized "practical AI." Through a specific AI button on the device, users can call the edge large model with one click, allowing for smooth multi-turn AI conversations even in offline environments, as well as document and call summaries.

Midea X Alibaba Cloud

As a leader in smart home technology in China, Midea will collaborate with Alibaba Cloud to create a digital foundation for overseas regional business located in Singapore. In recent years, globalization has become a core strategy for Midea Group. To meet the demands of its global business, Midea will independently deploy some of its overseas IT applications in Singapore, utilizing Alibaba Cloud's public cloud computing capabilities, including storage, cloud networking, cloud security, and cloud databases, to build a digital foundation for overseas regional business. Since 2014, Midea Group has been collaborating with Alibaba Cloud. In the future, both parties will engage in more cooperation in overseas expansion and AI, accelerating Midea's innovative global outreach