In a remarkable ascent, Alibaba's stock price has surged from a low of HKD 77 to HKD 124, marking a 60% increase alongside a staggering market capitalization growth of approximately HKD 900 billionThis surge, however, is not attributed to the e-commerce sector as one might expect, but rather to the momentum driven by advancements in artificial intelligence (AI) models in China.

Essentially, the lifeblood of Alibaba's recent performance can be traced back to its engagement with AI.

On February 13, during the World Governments Summit 2025 held in Dubai, Alibaba's co-founder, Joe Tsai, confirmed the collaboration between Alibaba and Apple to develop AI features tailored for iPhone users in ChinaTsai noted the significance of Apple selecting Alibaba as its localized partner for AI services, signaling a monumental partnership that could reshape the landscape for both companies in the Chinese market.

Since Tsai publicly acknowledged the company's lag in the AI domain in 2024, he has made three notable appearances, each time reemphasizing the pivotal role that AI plays within Alibaba's ecosystemThis was particularly evident in his letter to shareholders for the 2024 fiscal year, where Tsai articulated Alibaba's identity as a fusion of e-commerce and cloud computing.

AI has thus cemented its status within Alibaba's strategic framework, making it a cornerstone of the company's future.

As a result of these developments, Alibaba's stock has seen continued spikes over several days, currently sitting at HKD 124.1 per share — a peak not seen in three years, bringing the company’s market value to a colossal HKD 2.36 trillion.

Alibaba's recent accolades can be attributed to two significant factors: the stimulating effects of its partnership with Apple and the inherent advantage of its Tongyi model among domestic AI models.

At the end of January 2023, the introduction of DeepSeek reignited global interest in applying large models in China, placing the focus squarely on the influential Tongyi Qianwen AI model from Alibaba.

In tandem with this resurgence, Alibaba has embarked on a redefinition of its business model, decisively pivoting away from non-core assets as of late last year

Advertisements

For example, in December 2024, Alibaba sold its stakes in Sam’s Club (Xingyun Tianxia) to the Yonghui Superstores Group for about RMB 7.4 billion, incurring a loss of roughly RMB 9.3 billion.

Previously, stock fluctuations were heavily influenced by the company’s e-commerce performanceHowever, with the ongoing intensification of competition in the e-commerce market resulting in stagnated growth, Alibaba has undergone a transformative strategic shift, steering clear from solely being a retail giant to adopting a technology-driven company profile.

A similar fate befell Alibaba when it sold its stakes in Gaoxin Retail, which likewise translated into an expected loss of RMB 30 billion.

Tsai emphasizes the dichotomy of Alibaba’s identity, stating that while e-commerce remains crucial, the company has recognized the strategic significance of its delivery services, exemplified by companies like Ele.me, even if these services aren't considered core to the company's portfolio.

A conspicuous shift has taken place; as of the first quarter of 2023 (the end of Alibaba's 2024 fiscal year), revenues derived from China commerce accounted for less than half of the company's total revenue.

The tidal wave of large models has propelled Alibaba to a noteworthy increase of HKD 900 billion in market value.

Alibaba’s stock has experienced three major rallies recently, most notably influenced by the news of the collaboration with AppleTwo previous surges were fueled by Stanford University Professor Fei-Fei Li announcing training based on the Tongyi Qianwen large model, along with the phenomenal rise of DeepSeek.

On February 6, Li's team announced the successful training of an AI inference model named S1 using Alibaba Cloud's Tongyi Qianwen model at a cost of less than $50. Their findings indicated that the model was competitive with top models such as OpenAI's O1 and DeepSeek's R1, excelling in competitive math queries by 27% over O1-preview.

This marked a pivotal moment, transforming Alibaba's large models into a talking point beyond industry conversations.

Prior to this, DeepSeek rose to global acclaim just before the Lunar New Year, with Alibaba officially releasing two key models, Qwen2.5-VL and Qwen2.5-1M, on January 28. These models were found to surpass GPT-4 and Claude 3.5 in terms of visual understanding capabilities.

Thus, the Chinese New Year gala catalyzed the visibility of the Tongyi model further

Advertisements

Remarkably, the focus shifted from traditional e-commerce to highlighting innovative capabilities in AI, amplifying Alibaba’s brand exposure both directly and through the buzz surrounding DeepSeek.

On Lunar New Year’s Day, January 29, Alibaba introduced Qwen2.5-Max, claiming its performance outshined both Llama and DeepSeekThe US markets remained unfazed during the holidays, with the stock of Tongyi Qianwen maintaining its upward trajectory.

Alibaba was among the frontrunners in the domestic landscape for unveiling large models, initiating its multi-modal pre-training project, M6, back in 2020. The primary goal was augmenting both understanding and generating capabilities in the Chinese language.

That same year, Alibaba introduced its base model with 300 million parameters, marking its initial foray into the realm of large-scale modelsOver the course of 2021, this model's parameters grew into the billions, resulting in Alibaba's emergence as one of the largest multi-modal models globallyBy October 2021, M6 expanded its parameters to a staggering 10 trillion, solidifying itself as the world’s largest pre-trained AI model at the time.

In this light, it's clear that Fei-Fei Li’s choice to utilize Alibaba's large model for training was anything but arbitrary.

Notably, the open-source AI community, Hugging Face, has recognized Alibaba's Tongyi Qianwen models’ profound impact, as all top ten models on their leaderboard are derivatives crafted through further training on Alibaba's open-source models.

Today, AI has emerged as a foundational premise in determining Alibaba's corporate identityPreviously viewed solely as an e-commerce enterprise, Alibaba's stock fluctuations were directly tied to the health of its retail operationsThis changed dramatically between 2019 and 2020, when the stock prices rose to historical highs propelled by bountiful e-commerce growth.

As the domestic e-commerce market faced stagnation, concerns over Alibaba’s declining stock arose, leading to periods where competitors such as Pinduoduo surpassed Alibaba in market value.

Pinduoduo currently boasts a market value of HKD 1.3 trillion, and based on Alibaba's projected EBITA of 194.8 billion for the e-commerce sector by March 2024, a fair valuation would place Alibaba at approximately HKD 2 trillion.

This prompted a pressing question: What can Alibaba rely on beyond e-commerce to support its foundations? Early in 2022, Alibaba’s shares plummeted below $100 USD, entering a prolonged phase of stagnation.

Throughout this time, Alibaba Cloud has maintained its position as a leading player in the domestic cloud computing market, albeit its revenue and profit contributions remain relatively modest in the grand scheme of Alibaba's operations

Advertisements

For instance, in the third quarter of 2024, Alibaba Cloud reported revenues of RMB 29.6 billion, a minor fraction of Alibaba’s overall earnings.

Even Jack Ma, the company’s co-founder, has shifted his discourse from e-commerce to AI during public appearances over the past two yearsHis keynote at the Ant Group's 20th anniversary emphasized the importance of AI.

Tsai has repeatedly underscored that e-commerce and cloud computing are the dual pillars upon which Alibaba will build its futureIn June 2024, during an interview with Norway's sovereign wealth fund, Tsai reiterated a focus on these two core business aspects.

In the shareholder letter for the 2024 fiscal year co-authored by Tsai and Wu Yongming, a profound inquiry was posed: "Who are we?" Tsai elucidated both e-commerce and cloud computing as the company's critical components.

Confidence in large models stems from a belief that through continuous aggregation of data resources and computational prowess, the performance of these models will exhibit non-linear growth trends.

This represents a historic zenith for Alibaba’s cloud computing venturesLooking ahead, Alibaba plans to continue investing heavily in core areas: accelerating the growth of their primary operations and maintaining a leadership position in foundational technology, including AI and related innovations.

Betting on AI also necessitates shedding heavy assets to streamline operationsIn tangible terms, toward the end of 2024 and into early 2025, Alibaba executed two significant divestitures that would alter the landscape of retail in China.

On December 17, 2024, the company revealed its sale of about 99% of its stakes in Intime Retail, bringing in approximately RMB 7.4 billion, albeit at an anticipated loss of around RMB 9.3 billion.

On January 1, 2025, Alibaba disclosed their divestment of around 78.7% of shares in Gaoxin Retail, a transaction valued at approximately HKD 131.38 billion, despite a previous investment nearing HKD 50 billion.

The divestiture of tangible assets illustrates Alibaba's departure from an omnipresent model to a refined concentration on core prospects.

Finally, the emergence of large models in the domestic scene has materialized into a competitive landscape characterized as a "3+2" paradigm, and within this dynamic, what advantages does Alibaba possess?

While it was initially rumored that Baidu would be Apple's partner for domestic operations, it ultimately fell to Alibaba, with indications suggesting that Baidu remained one of Apple's suppliers.

This prolonged vendor selection process of Apple inadvertently serves as a reflection of the domestic AI large model industry, marking a new benchmarking standard due to Alibaba’s inclusion.

The investment landscape concerning AI models in China can be filtered through three predominant concepts: Kimi, Doubao, and the emerging DeepSeek.

This terrain showcases two underdogs and a consolidator leading the pack.

Currently, the market for consumer-facing large models has coalesced into a 3+2 structure, primarily featuring Alibaba's Tongyi Qianwen and ByteDance's upcoming venue, Jidmeng.

DataEye’s reports highlighted that as of January 2025, in the mainland Chinese market, app downloads for major AI models indicated Doubao at 24%, Kimi at 18%, Quark at 17%, DeepSeek at 7%, and Jidmeng at 3%, with the remaining models aggregating to 31%.

The sensational rise of DeepSeek has been dubbed as China's "Alibaba IPO moment." As is widely known, the ensuing events that followed Alibaba's IPO catalyzed a prosperous internet landscape, with Chinese companies leveraging practical applications of AI thriving in the domain.

The current sentiment among investors leans positively toward the prospects of China's tech giants harnessing the large model application market, with many drawing parallels to the previous internet boom, asserting that China’s advantages in real-world applications and deployments will become increasingly evident.

In 2023, Alibaba launched the Tongyi Qianwen model, emphasizing partnerships with various industry and scenario collaborators to enhance AI capabilities while committing to a complete overhaul of all its products through the lens of the large model for modernization.

Tsai noted that Alibaba occupies a unique position, as one of the few companies globally that possess both in-house AI competencies and robust cloud services.

He pointed out the distinction among tech behemoths, where Microsoft and OpenAI function as separate entities, Amazon operates cloud services but lacks a proprietary large language model, while Meta houses LLMs but lacks cloud infrastructure; Google appears to be the only other firm with overlapping capacities, although its AI capabilities do not often garner leading recognition.

"In China, Alibaba stands out as the sole company with a leading cloud service alongside a competitive edge in artificial intelligence," he stated.

However, Alibaba's journey with its large models has not been without hurdles.

In response to talent attrition, on February 6, 2025, Alibaba proudly welcomed the notable AI scientist Xu Zhuhong, who now serves as Vice President, focusing on multi-modal foundational models and related operational solutions.

This decision comes on the heels of the departure of Zhou Chang, who previously helmed the Tongyi Qianwen model’s technical team, joined ByteDance in December 2024, taking with him over a dozen team members, driven by an attractive compensation package and senior role offered by ByteDance.

At present, offerings from Doubao and Jidmeng operate at par with Alibaba's products in terms of market reception and performance metrics, suggesting Alibaba’s competitors are not trailing far behind, even outperforming in certain consumer-facing areas.

As the landscape unfolds and new front-runners emerge within the bustling AI model arena, whether Alibaba can maintain its leadership remains to be seen.

Advertisements

Advertisements