April 10, 2025
By Our Correspondent
Ant Digital Technologies, the technological division of Ant Group, has selected Hong Kong as its international headquarters, demonstrating its dedication to artificial intelligence (AI) and Web3 technologies, reported by The South China Morning Post.
The subsidiary highlighted Hong Kong’s significance as a “bridgehead” for its global strategy. The company intends to collaborate with local universities and research institutions to establish joint laboratories. “By establishing our international headquarters here, we reaffirm our commitment to the AI and Web3 industries,” remarked Zhao Wenbiao, CEO of Ant Digital Technologies. In recent years, Hong Kong has been striving to position itself as a technology hub, particularly emphasizing blockchain and virtual assets. This decision aligns with Ant’s broader initiatives, founded by Alibaba’s Jack Ma two decades ago, to enhance the international expansion of its blockchain and AI technologies, especially in light of the global AI surge ignited by OpenAI’s ChatGPT and more recently, China’s DeepSeek.
Ant intends to utilize Hong Kong’s position as a “strong financial ecosystem to promote Web3 and AI technologies while improving the circulation of digital assets,” as stated by the company. On April 6, 2024, individuals walked past a display for the Hong Kong Web3 Festival. The city has been actively working to establish itself as a hub for virtual assets.
In the last two years, Ant Digital has significantly increased its investments in the development of AI models and the enhancement of essential capabilities, such as trusted intelligent agents, according to Zhao. In the Web3 domain, the firm introduced an asset tokenization platform named “Two Chains and One Bridge,” which links green assets to the international capital market, he noted. The fintech leader has recently attracted global attention for its proficiency in training large language models (LLMs)—the technology that powers generative AI chatbots—by utilizing locally sourced graphics processing units, thereby reducing training costs by 20 percent, as reported by various media outlets. Ant’s Ling team, which is tasked with LLM development, has disclosed that the Ling-Plus-Base model can be “effectively trained on lower-performance devices,” according to a research paper from the company. While the team did not specify the devices used, Bloomberg reported that this accomplishment was achieved with domestic chips.
