Tired of ads? Enjoy an ad-free experience by signing up.
👩‍🍳 How we use AI at Tech in Asia, thoughtfully and responsibly.
🧔‍♂️ A friendly human may check it before it goes live. More news here

Hong Kong develops DeepSeek-based AI for local chips

Hong Kong Generative AI Research and Development Centre, a government-backed AI lab in Hong Kong, plans to launch its HKGAI-V3 model in the first half of 2026 and said the system can run on Chinese-made chips as it explores overseas expansion.

The model is based on DeepSeek V4 architecture with full-parameter fine-tuning for localisation and has been optimised to run on both mainstream and domestic hardware, including Huawei’s Ascend 910C chips, said director Guo Yike at a Tsinghua University alumni event in Hong Kong.

Founded in 2023 under the InnoHK program, the lab develops large language models for Hong Kong and launched Cantonese chatbot HKChat in November.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

DeepSeek V4 open weights, Huawei support, lower pricing

  • DeepSeek V4, the base model behind HKGAI-V3, uses open weights. Labs such as HKGAI can download it and tailor it for local needs, including Cantonese language support 1.
  • The model runs on Nvidia graphics processing units (GPUs) as well as Huawei Ascend AI accelerators. That fits HKGAI’s plan to use Chinese hardware instead of depending only on Nvidia chips that face export restrictions 1.
  • Cost also shapes this decision. DeepSeek’s application programming interface (API) pricing sits well below rivals such as OpenAI, which suggests Chinese AI firms are competing through efficiency and lower prices 2.

HKGAI-V3 launch backs sovereign AI under export controls

  • HKGAI-V3 is part of a wider effort to build a Chinese AI hardware and software stack. The Hong Kong government-backed lab is preparing a DeepSeek-based model that can run fully on Chinese-made chips 3.
  • HKGAI’s approach adapts a Chinese model for Hong Kong’s needs. That gives other governments and public-sector labs a template for “sovereign AI,” with more local control over models, infrastructure, and data, while matching HKGAI’s plan to explore overseas markets 3.
  • DeepSeek V4 is also rolling out amid U.S. allegations of intellectual property (IP) theft and model distillation, a technique that trains a new model on another model’s outputs, involving DeepSeek and other Chinese AI firms. That adds geopolitical risk for governments and businesses weighing this technology stack 4.

Stay ahead in Asia’s tech landscape

You've reached your 2 free content limit for the month. Sign up for free to read the full story.

🏄 For casual readers / 👶 Free

Basic

US$0

Free forever

Get instant access to this article and more every month

0 premium content

Unlimited news briefs

5

5 articles

Ad-free reading experience

Just US$0 per day

⌛Sign up in 20s. No payment details needed.

📖 For learners / 👍 Starter

Lite

US$4.92/month

Billed annually at US$59/year

Get instant access to this article and more every month

4

4 premium content

Unlimited news briefs & articles

Ad-free reading experience

Just US$0.17 per day

Cancel anytime

Our subscriber community includes professionals from these companies:

Stay updated on the go with our mobile app.

Get latest insights with smoother, more personalized experience through TIA mobile app.