Establishment and demonstration of heterogeneous AI computing infrastructure
Joint analysis with the National Science and AI Research Center
Establishment of ‘AMD AI Excellent Research Center’ in Korea
The government has joined hands with AMD, an Nvidia rival, to disperse the center of gravity of the synthetic intelligence (AI) computing infrastructure centered on the Nvidia graphics processing unit (GPU). Nvidia is an absolute powerhouse that’s estimated to have accounted for about 80% of the international AI accelerator market as of final 12 months, and as demand for knowledge facilities has surged due to the unfold of AI providers, even the nationwide AI infrastructure is solidifying its construction that depends on particular company GPUs.
This settlement incorporates the government’s willingness to mitigate such focus and incorporate home AI semiconductors with sturdy reasoning into the international provide chain with open and heterogeneous computing that mixes central processing items (CPUs), GPUs, and NPUs.
The Ministry of Science and ICT introduced on the twenty seventh that it met with AMD Chairman and CEO Lisa Su at the “AMD Advancing AI 2026” occasion held in San Francisco on the twenty third (native time) and signed a “MOU on AI Semiconductor Ecosystem Cooperation between the Ministry of Science and ICT and AMD” to cooperate in the AI semiconductor and computing fields.
AMD, a worldwide semiconductor firm, is pushing to broaden the AI computing ecosystem primarily based on ROCm, an open-source software program platform, and open requirements. In explicit, it’s actively creating an open AI infrastructure ecosystem by taking part as a founding member of UALink, an open high-speed AI semiconductor connection normal, alongside with a number of international corporations comparable to Google, Apple, and Microsoft (MS).
The Ministry of Science and ICT has lately joined hands with AMD as a result of the international AI semiconductor market is quickly shifting away from a single GPU-centered computing construction to a heterogeneous computing structure that mixes varied computing gadgets comparable to CPU, GPU, and NPU in accordance to service traits and cost-effectiveness due to the rising demand for reasoning due to the unfold of Generative AI and AI.
In front-line websites, infrastructure expertise that mixes varied AI semiconductors comparable to CPU, GPU, and NPU in an optimum mixture relying on computational traits relatively than counting on a single GPU to decrease the price of offering AI providers and optimize energy effectivity is drawing consideration.
The Ministry of Science and ICT defined, “This memorandum of understanding aims to establish a heterogeneous AI computing infrastructure that links AMD’s CPU and GPUs with domestic NPUs that have strengths in reasoning in response to these changes in global AI computing architectures, and to expand the foundation for global cooperation in the domestic AI semiconductor ecosystem.”
First of all, the two sides are selling the institution and demonstration of heterogeneous AI computing infrastructure linking AMD’s CPU and GPU with home NPU. Through this, it would implement an built-in computing setting that effectively combines and makes use of varied computational assets in accordance to the traits of every AI service and create an open AI computing ecosystem wherein not solely AI semiconductor corporations but additionally associated Korean corporations comparable to reminiscence, DPU, CXL, community, server, and orchestration software program take part collectively. In addition, primarily based on this, we plan to safe a sensible reference mannequin in order that home AI semiconductors can take part in the international AI computing provide chain.
It may also promote ‘AI for Science’ analysis cooperation centered on the National Science AI Research Center (NAIS). To assist AI-based analysis and difficult R&D in the area of superior science and expertise, AMD plans to present high-performance computing assets comparable to CPU and GPU, AI software program, {and professional} manpower.
In explicit, AMD is pushing for the institution of an AI Center of Excellence (AI Center of Excellence) in Korea. Based on AMD’s AI computing assets and software program capabilities, the Artificial Intelligence Excellent Research Center will function a cooperative base to assist technical cooperation, AI semiconductor software program improvement, computing expertise verification, and joint analysis. Through this, AMD’s international community and home {industry}, academia, and analysis capabilities will likely be organically linked to broaden varied cooperation alternatives primarily based on the open AI ecosystem.
At the similar time, the two sides may also cooperate to domesticate wonderful human assets in the AI area. AMD will assist AI computing platform-based training and analysis applications and computing assets to main universities and graduate colleges in Korea, and broaden alternatives for home college students, professors, and researchers to expertise the newest AI expertise and international improvement setting. In addition, it plans to broaden the basis for industry-academic-research and analysis cooperation to safe competitiveness in future AI applied sciences by selling joint analysis and technological cooperation in next-generation AI fields comparable to bodily AI.
Meanwhile, the cooperation between the Ministry of Science and ICT and AMD started in earnest with AMD Chairman Lisa Su’s go to to Korea in March, and since then, the cooperation area and implementation plan have been specified after shut session between the two sides. In addition, the Ministry of Science and ICT defined that President Lee Jae Myung’s journey to the United States has led to the improvement of discussions gathered to this point into sensible cooperation initiatives comparable to constructing AI computing infrastructure, AI analysis, and human assets coaching, and signing MOUs.
Deputy Prime Minister and Minister of Science and ICT Bae Kyounghun mentioned, “Global AI competition is changing to a heterogeneous computing and open ecosystem that combines various computational resources such as CPU, GPU, and NPU. In order for Korea to become one of the top three AI powerhouses, it is important to secure an independent and open AI computing infrastructure.”
“We will be able to cooperate with AMD to increase the competitiveness of the domestic AI semiconductor ecosystem and create a new cooperative model to respond to changes in global AI computing infrastructure,” he mentioned.
Chairman Lisa Su additionally mentioned, “Korea is a key strategic market for AMD with global competitiveness in semiconductors, AI, advanced research, and physical AI.” “AMD will actively support Korea to build an open and heterogeneous AI infrastructure by combining high-performance computing technology, open software capabilities, and global ecosystem partnerships in cooperation with the Ministry of Science and ICT.” “This cooperation shows AMD’s long-term commitment to contribute to AI innovation and economic growth in Korea through technical cooperation, talent training, and close joint R&D beyond simple hardware support,” he careworn.
In the future, the two organizations will kind a joint consultative physique for the sensible implementation of this settlement and proceed to focus on detailed cooperation duties and implementation plans via quarterly working-level conferences.