Hong Kong launches DeepSeek-based AI model designed to run on domestic chips

The Hong Kong Generative AI Research and Development Centre has officially launched a new large language model built on DeepSeek’s architecture, designed specifically to run on domestic Chinese chips. The move marks a significant step for the government-backed lab as it pushes to commercialise its products and export Chinese AI technology beyond the city’s borders. Established in October 2023 under Hong Kong’s InnoHK funding programme, HKGAI has focused on developing AI models tailored to local needs. Its latest model, HKGAI-V3, represents a departure from the reliance on Western hardware. The centre’s leader, also the provost at the Hong Kong University of Science and Technology, confirmed last month that the model has been optimised to operate on both Western and domestic hardware, including Huawei Technologies’ Ascend 910C chips. This dual compatibility is no small feat—it positions the model as a viable alternative for markets where Nvidia’s chips are restricted or prohibitively expensive. The implications extend beyond hardware. HKGAI has embedded Hong Kong’s “cultural DNA” and local knowledge into the V3 model, aligning it with the linguistic context and habits of local users. This localisation effort is not merely cosmetic; it reflects a broader strategy to make the model relevant in a city where Cantonese and written Chinese coexist with English. The centre’s earlier V1 model, launched last year, already demonstrated that Hong Kong was not missing out on the AI wave, Technology and Industry. What a casual reader might overlook is the geopolitical undercurrent. Hong Kong’s public sector has been increasingly adopting mainland Chinese technology solutions amid growing risks of US export controls, which treat the special administrative region the same as the mainland. The ability to run AI systems on domestic chips allows government bodies and enterprises to operate on their own preferred infrastructure rather than relying on public cloud platforms. This is not just about cost or convenience—it is about sovereignty in an era of tech decoupling. Interestingly, at Wednesday’s launch event, HKGAI did not explicitly mention the model’s ability to run on Chinese AI chips. Instead, it emphasised the provision of “diversified inference computing power and intelligent services.” The omission is telling: the centre may be downplaying the hardware angle to avoid drawing unnecessary attention from US regulators, even as it quietly positions itself as a bridge for Chinese AI exports. The push for sovereign AI—the idea that jurisdictions should develop and operate AI systems on their own infrastructure—is central to HKGAI’s ambitions. The centre now aims to replicate its experience localising DeepSeek for Hong Kong in overseas markets. If successful, this could offer sanctioned countries and enterprises a path to deploy advanced AI without depending on Nvidia’s ecosystem. The model’s ability to run on Huawei’s Ascend chips makes it particularly attractive in markets like Russia, Iran, or parts of Southeast Asia where US export controls are a barrier. Hong Kong’s role as a testing ground for this strategy is deliberate. The city’s unique position—a global financial hub with deep ties to mainland China and exposure to international sanctions—makes it an ideal laboratory for proving that Chinese AI can compete on both performance and compliance. The V3 model is not just a product; it is a proof of concept for a new kind of tech export, one that bypasses the traditional hardware monopoly. Whether this model gains traction in markets beyond Hong Kong will depend on how quickly the centre can scale its deployment and convince potential buyers that domestic chips are not a compromise, but a strategic advantage.

The Hong Kong Generative AI Research and Development Centre (HKGAI) has officially launched a new DeepSeek-based large language model that can run on domestic chips, as the government-backed lab seeks to commercialise its products and export Chinese AI over...

Hong Kong’s DeepSeek-based model on domestic chips challenges Nvidia’s hardware monopoly, offering a viable alternative for AI deployment in sanctioned markets.

The development adds to a wider Hong Kong ai & machine learning story in which companies are being judged on execution, capital access, regulatory fit and the credibility of their regional expansion plans.

For business readers, the important question is whether this becomes an isolated announcement or part of a more durable operating pattern across customers, financing channels, partners and public-market expectations.