(Central News Agency reporter Liu Yuting, Prague, 11th) In recent years, Romania has been actively building AI-related infrastructure. Expert IT, a Romanian IT company, plans to build an AI data center with an initial investment of more than 8 million euros (approximately NT$286 million). It is expected to be opened before the end of this year and will be mainly used for AI model development. Since AI model training and computing consume a lot of power, how to ensure stable power supply has become an important challenge in promoting related plans.
Romania has been actively expanding its AI infrastructure in recent years, hoping to enhance its AI industry status in Southeast Europe. According to quotes from Daily News Hungary, Romanian IT company Expert IT said it plans to build an AI data center in Reghin, Transylvania, Romania.
Mihai Farcaș Dan, CEO of Expert IT, said that the first phase of investment in the AI data center is 7 servers equipped with 56 high-performance NVIDIA B300 graphics processors (GPUs), which is the largest purchase of its kind this year. It also covers data center infrastructure such as power supply and cooling. The company has discussed expansion plans and may add 512 more GPUs in the future. The relevant servers are expected to be put into operation this year and will be used to develop large-scale AI models.
This project is part of the overall development project promoted by Expert IT in Leijin. Expert IT began planning investments 12 years ago, and in the past three to four years, it has gradually prioritized AI as a development focus. The overall development project scale is approximately 10 million euros, and the base area is approximately 1.7 hectares.
Expert IT said that medical care is one of the main areas where the AI data center may be applied in the future. The company is considering cooperating with Romanian clinics and universities to develop AI systems that can assist doctor-patient communication, support medical personnel, and analyze medical images. Related technologies are expected to be used in fields such as tumors and heart diseases in the future to help doctors identify disease characteristics and accelerate diagnosis by analyzing large amounts of data and medical images.
In addition to Regin, there are also a number of AI and data center projects being promoted in other regions of Romania. DriverAI, an American technology company, and Leviatan Group, a Romanian enterprise group, have cooperated to build a "Quantum Artificial Intelligence Data Center" in Cluj County. The investment in the first phase is about US$1 billion. Its main function is AI model training and computing. However, the project is still in the preparation stage and has been involved in legal disputes due to objections from neighboring land owners.
The Mureș County Council and Industrial Park are planning another data center. It is expected to invest approximately 45 million euros and build a facility of approximately 1,700 square meters to provide data storage, processing, server hosting and AI computing services. According to reports, the project has received the necessary funding.
In addition, a university of medicine and science and technology located in Târgu Mureș is also promoting high-performance computing infrastructure, with a total investment of approximately 11.2 million euros, which will be used for research in AI, machine learning, large-scale data analysis, communication security, and network attack detection.
According to reports, Romania is planning a "Black Sea AI Gigafactory" and is expected to build related facilities in Cernavodă and Doicești, which will eventually be equipped with more than 100,000 AI processors.
However, as the scale of the AI data center expands, power supply and grid carrying capacity have also become important issues for the development of related infrastructure in Romania. In addition to requiring a large amount of power for large-scale model training and inference, AI data centers must also maintain server operation for a long time and rely on cooling systems to control equipment temperatures. Therefore, the requirements for power supply stability and energy costs are very high, which is also the key to the smooth expansion of the data center.
