Minister of Digital Affairs Lin Yi-ching: AI Computing Center BOO Projects Expected to Exceed NT$100 Billion in Investment by Year-End

Taipei, Jun. 30 (CNA) Minister of Digital Affairs Lin Yi-ching announced today that the Build-Own-Operate (BOO) scheme for AI computing centers has been finalized, incorporating them into the scope of major public infrastructure under the Statute for Investment by Foreign Nationals, to guide private investment. Applications are currently under review, with total investment expected to surpass NT$100 billion by the end of this year. This will provide over 20,000 GPUs, with a certain proportion reserved for free or preferential use by government and academic research institutions.

At the Sovereign AI Summit co-hosted by Yating Smart and Taiwan AI Labs today, Lin Yi-ching stated in his speech that sovereign AI holds two significant meanings. Firstly, AI models must possess a Taiwanese perspective, including familiarity with Taiwanese languages, social environments, and values. Secondly, computing centers must be located in Taiwan, as their location subjects them to the jurisdiction of the host country's laws.

Lin Yi-ching further elaborated that the Ministry of Digital Affairs has announced the AI computing center BOO scheme to attract private investment. Applications are currently under review, with investment expected to exceed NT$100 billion by year-end, providing over 20,000 GPUs. Concurrently, a certain proportion of GPU computing power will be reserved for free or preferential use by government and academic research institutions.

He pointed out that to build AI models with Taiwanese values, the AI Product and System Evaluation Center (AIEC) has evaluated over 100 AI models using criteria such as Taiwanese values, high school college entrance exam Chinese, and high school college entrance exam social studies, to assess their understanding of Taiwanese language, society, and values. Simultaneously, Taiwan has initiated the development of large language models for finance and law, aiming for AI to respond to relevant issues based on Taiwan's regulations and institutional frameworks in the future.

Lin Yi-ching stated that countries are actively developing sovereign AI to reduce reliance on the US and Chinese technological systems. Taiwan, with its technological prowess, democratic values, and trustworthiness, is poised to become a crucial partner in the international development of sovereign AI. He also noted that Taiwanese companies have already signed Memoranda of Understanding (MOUs) with other countries, indicating that related business opportunities are gradually emerging.

Ethan Tu, founder of Taiwan AI Labs, pointed out that in the era of Agentic AI, enterprises need AI assistance. However, when a company's core processes are built upon external cloud AI services, they face operational risks if the service is discontinued, fees are adjusted, or model capabilities change.

Tu emphasized that enterprises also require sovereign AI, a core principle of which is "data stays put, models move, not data." This allows various organizations to collaborate while maintaining autonomy and privacy protection.

Tu explained that enterprise-grade sovereign AI comprises five pillars: data sovereignty, computing sovereignty, model sovereignty, system sovereignty, and governance sovereignty. Data sovereignty involves providing a sovereign AI governance platform where data remains within the enterprise and institution. Computing sovereignty entails establishing reasonable public computing power, but the mechanism must be transparent and democratic to avoid creating new computing oligarchies or allowing centralized providers to plagiarize private sector achievements.

He highlighted that model sovereignty actively supports local models and indigenous ecosystems, free from the constraints of specific multinational corporations. System sovereignty signifies the importance of indigenous framework capabilities and ecosystem building. Governance sovereignty aligns with advanced international AI governance, creating a verifiable, auditable, and trustworthy AI ethical and legal framework.

Tu stated that under sovereign AI, data, models, and applications will remain within the enterprise. In the future, AI will feature various models, which will become crucial competitive advantages for enterprises, usable internally and convertible into external sales. (Editor: Yang Kai-hsiang) 1150630)