In a remarkable display of technological evolution, a revolution is rippling across the globe, driven largely by innovations in computing power, particularly through advancements in artificial intelligence (AI). This unprecedented transformation is manifested profoundly in the success of cloud service providers, with Alibaba Cloud, China's largest cloud vendor, emerging as a pivotal player in this era of AI asset re-evaluationThe recent earnings report from Alibaba Group serves not only as a reflection of corporate growth but also as a barometer for the broader market trends influenced by AI.

On the evening of February 20, Alibaba Group announced its quarterly financial performance for the period ending December 31, 2024. The company reported revenues for Alibaba Cloud totaling a significant 31.742 billion yuan, a year-on-year increase of 13 percent, which represents a substantial acceleration from the previous quarter's 7 percent growthThis impressive uptick in revenue can be attributed primarily to the enhanced demand for public cloud services driven by AI-related products, with AI revenue experiencing triple-digit growth for the sixth consecutive quarterThis marks a pivotal moment for Alibaba, signifying its increased influence and authority in the AI space.

Moving forward, Alibaba Cloud has committed to further investments in customer growth and technology innovation, particularly focusing on enhanced AI infrastructure aimed at increasing AI cloud adoption and maintaining its market leadershipThe buoyancy in Alibaba's stock price reflects investor confidence; it soared by as much as 57% since the start of the year, reaching new heights not seen in three years and translating to a market capitalization surge exceeding 600 billion Hong Kong dollars, landing at a staggering 2.3 trillion Hong Kong dollars.

In a detailed analysis on February 13, analysts from JPMorgan reviewed Alibaba Group's market valuation, suggesting there remains substantial potential for further increases

Advertisements

The favorable prospects stem from three main drivers: an upward adjustment to Alibaba Cloud's valuation multiples, improved revenue forecasts for cloud services, and a positive outlook on profitability within the Chinese e-commerce sector.

As we stand on the brink of what can be termed the AI era, Alibaba Cloud's role as a builder of AI infrastructure becomes increasingly significantIt is not merely adjusting to this paradigm shift; it is effectively at the forefront of what could be the most technologically disruptive business challenge in human history - a journey characterized by both immense potential and formidable challenges.

One of the cornerstones of this AI advancement is Alibaba Cloud's commitment to open-source technology, which seeks to widen access to AI capabilitiesRecently, Alibaba Cloud unveiled an upgraded flagship model, Qwen2.5-Max, a result of their ongoing exploration within the MoE (Mixture of Experts) model frameworkWith over 200 trillion tokens in its pre-training dataset, this new model showcases outstanding overall performance, achieving high scores across several public benchmarksImpressively, it surpasses both the leading open-source MoE models and the largest open-source dense models currently available globally.

As a pioneer in the field, Alibaba Cloud stands out as the earliest, and currently, the only cloud service provider in China to have developed large models through self-service and open-sourced them comprehensivelyThe company has spearheaded the initiative to realize "full-size, multi-modal, multi-scenario" openness in modeling technology, a move that resonates well within the broader industry.

In November of last year, Alibaba Cloud proactively released and open-sourced its complete series of code models, consisting of six iterations of the Qwen2.5-Coder modelsIndustry assessments have shown that these six coding models achieve exceptional performance across various benchmark tasks, with the flagship 32B model emerging as the best-performing open-source model globally, solidifying its reputation as a formidable presence in the AI modeling community.

Continuing its momentum from last year, Alibaba Cloud introduced additional AI inference models, such as the QwQ-32B-Preview and the QVQ-72B-Preview multi-modal inference model

Advertisements

The QVQ model has outperformed previous visual understanding models, dubbed the "open-source champion," Qwen2-VL, and its overall capabilities are comparable to leading inference models like OpenAI's O1 and Claude 3.5 Sonnet.

In January of this year, Alibaba Cloud made headlines once again with the open-source rollout of the Qwen2.5-1M model, which boasts support for an impressive context of 1 million tokensThis model comes in two sizes, 7B and 14B, outperforming the GPT-4o-mini in tasks involving long-text processingAdditionally, the company launched a new inference framework that enables nearly sevenfold acceleration in processing long-text inputs at the million-token levelThe flagship model Qwen2.5-VL, available in 3B, 7B, and 72B sizes, has succeeded in clinching top ranks across multiple authoritative assessments.

There is a clearly defined strategy by Alibaba Cloud to democratize AI access through their open-source initiativesTheir ambition is to enable a vast number of small and medium-sized enterprises, along with AI developers, to harness the power of Qwen quickly and effectively, thereby fostering widespread implementation of large model technologies across diverse sectorsNotably, the number of derivative models based on Qwen has exceeded 90,000, positioning it as one of the largest generative language model families in the worldAccording to the latest list released by Hugging Face on February 10, 2025, all top ten open-source large models are derivatives of the Qwen open-source model.

Furthermore, Alibaba Cloud's focus on AI is evidenced through its substantial investments toward developing AI infrastructureUnlike the traditional era dominated by IT practices, the AI era requires significantly enhanced performance and efficiency from infrastructureThe transition from a CPU-centric computing framework to a GPU-dominated AI computing paradigm is underwayAlibaba Cloud is at the forefront of this shift, comprehensively reconstructing fundamental hardware elements, computation, storage, network, databases, and big data to ensure a seamless fusion with AI applications, thereby accelerating development processes.

In terms of infrastructure, Alibaba Cloud has introduced the Panjiu AI server, which can support a maximum of 16 cards per machine, and a CPFS file storage system demonstrating a staggering data throughput of 20 TB/s

Advertisements

Additionally, they have implemented a high-performance network architecture capable of stabilizing connections for over 100,000 GPUs.

Building on these developments, Alibaba Cloud has launched GPU container computing for the first time under its ACS platformThis innovation enhances computational efficiency via topological-aware scheduling, while the PAI platform of Alibaba Cloud offers integrated elastic scheduling for training and reasoning at scales comparable to 100,000 cards, achieving over 90% effective utilization of AI computational power.

Through comprehensive optimizations, Alibaba Cloud has established a robust and efficient AI infrastructure system, achieving consistent training effectiveness exceeding 99% and enhancing model computational utilization beyond 20%.

In 2022, as the wave of large models began to gain momentum, Alibaba Cloud proposed the MaaS (Models as a Service) concept, which rapidly garnered industry consensusThe MoDa community, recognized as one of the best practices of the MaaS concept, quickly evolved into China's largest AI model community, with the phrase "find large models on MoDa" becoming widely embraced among AI developersAs of mid-February 2025, the MoDa community boasts over 40,000 models, spanning various domains like LLM, conversational AI, audio processing, text-to-image generation, video generation, and AI music composition, serving over 10 million developers.

Furthermore, Alibaba Cloud's international expansion strategy has shown impressive growth over recent yearsIts market footprint has multiplied more than twenty times over the past five yearsAs the market scales, Alibaba Cloud announced last May an increased commitment to the development of overseas cloud infrastructure, with plans to establish new data centers in five countries: Thailand, South Korea, the Philippines, Malaysia, and Mexico.

On February 13, 2025, Alibaba Cloud officially opened its second data center in Thailand, providing local enterprises, Chinese outbound companies, and local developers with enhanced cloud computing resources to support the increasing demand from Thailand's growing internet and fintech sectors

Just days later, on February 19, the launch of their data center in Mexico represented a significant milestone and entry into the South American marketThis center offers a range of essential cloud computing capabilities, including elastic computing, cloud storage, networking services, and container technologies at competitive prices compared to other major international cloud vendors, thus addressing the surging demand for cloud services across various sectors in Latin AmericaMoving forward, Alibaba Cloud intends to enhance the construction of local ecosystems and service systems for further expansion.