Digital Development Department Promotes AI “Computing Power Pool”
“Computing power” is the core engine driving AI development; however, the high cost of purchasing or renting high-performance computing resources often becomes a barrier for enterprises, especially for startup teams.
To address this, the Digital Development Department (hereinafter referred to as the “DDB”) is actively constructing Taiwan’s AI industry ecosystem. Following its application for NT$10 billion in funding from the National Development Council by the end of 2024, it has initiated efforts in five major areas: “computing power, data, talent, funding, and marketing.” A press conference for the “AI Computing Power Pool” was held on April 16, attended by DDB Minister Huang Yen-nan, Deputy Minister Lin Yi-ching, and representatives from various industry players.
Huang Yen-nan emphasized that the establishment of the computing power pool aims to solve the significant computing cost issues faced by startups and small to medium enterprises during the initial investment phase of concept verification (PoC), allowing them to quickly validate innovative ideas. “We hope to expand computing power in the coming years, targeting 70 GPUs to better assist startup development.”
Digital Development Department Minister Huang Yen-nan
Image / Digital Development Department
DDB AI Computing Power Pool Project Achievements: 88 AI Models Implemented
Since its launch, the “AI Computing Power Pool” project has received enthusiastic responses from the industry. The first opening in 2024 attracted applications from 97 enterprises, of which 64 were selected, successfully producing 88 AI models covering areas such as customer service, education, and health. Participating enterprises received an average of about six weeks of computing power usage time to bring their innovative concepts to fruition.
Regarding the role of the computing power pool, Deputy Minister Lin Yi-ching pointed out that compared to the low costs of garage startups in the past, today’s AI startups primarily face the high barriers of computing power. “The function of the computing power pool is to allow entrepreneurs to first verify whether their ideas are feasible. If the experiment is successful, the next step should be to seek funding based on the results and enter the commercial market to purchase or rent computing power.”
He emphasized that the government’s provision of computing power is meant to serve as a catalyst, assisting startups through the most challenging initial phase, with the ultimate goal of establishing a healthy ecosystem of collaboration between the government and private computing power providers.
The judging panel’s convener, Professor Chen Jun-liang from National Taiwan University of Science and Technology, explained that the review process places great importance on the project’s “real application needs,” “technical feasibility,” and “industry implementation potential,” to ensure that government resources are invested in the most impactful innovations.
Similarly, a judge, Wen Yi-ling, CEO of the Foundation for Artificial Intelligence Technology, stressed the importance of mastering proprietary data, models, and talent in the context of geopolitical challenges, stating that providing computing power this time is the “first mile” in encouraging enterprises to adopt AI, and supporting local AI ecosystems and software companies is the crucial next step.
DDB Deputy Director Chen Hui-min mentioned that in the first round of applications for 2025, 91 qualified enterprises applied, of which 54 were selected. Although the number of approved applicants has slightly decreased, she explained that this is due to this year’s support method being more flexible, allowing teams to apply for up to four GPUs for larger models requiring higher computing power, facilitating challenges in more complex AI tasks. This year’s selected project areas have also expanded to key industries such as healthcare, cybersecurity, finance, smart manufacturing, and public affairs.
In the DDB AI Computing Power Pool project, 91 qualified enterprises applied in the first round for 2025, and 54 were selected.
Image / Digital Development Department
Three Enterprises Showcase Highlights: Demonstrating AI Implementation in Healthcare, Insurance, and Transportation
Three enterprises, Kuan You Technology, Vision Travel Technology, and Rui Chuan Data, demonstrated the practical results of AI technology in various industries.
Kuan You Technology utilizes AI visual models for railway safety monitoring. Vice President Tsao Chang-yu stated that faced with 100,000 massive railway monitoring images, the model training that originally took over a day can now be completed in less than half a day through the computing power platform, significantly accelerating the development and validation of applications such as track anomaly detection.
Chen Wei-fan, R&D Manager of Vision Travel Technology, developed a professional domain language model for the insurance industry, using the computing power pool to process 5 million entries in the insurance corpus, integrating LLM and RAG technologies to create an internal AI assistant that does not require internet connectivity, enhancing customer service efficiency and stability.
As for Rui Chuan Data, focusing on smart healthcare, General Manager Huang Wei-da shared that its AI system uses the computing power pool to train 80,000 annotated disease-related data, assisting in the interpretation of chest CT and X-ray images, addressing the shortage of radiologists. The system not only improved detection sensitivity by nearly 10% but has also obtained TFDA Class II medical device certification, qualifying for clinical applications.
To meet the continuously growing industry demand, Chen Hui-min stated that the second and third rounds of applications for the AI computing power pool will start on May 1 and August 1, respectively. It is expected to provide a total of 163,000 GPU hours of computing power throughout the year, aiming to assist at least 150 enterprises, with each project receiving computing support valued at approximately NT$150,000 to NT$1.5 million, depending on the scale of the project.
This article is collaboratively reprinted from: Startup Gathering