The Paradox of "Sovereign AI" and Data Sovereignty
"South Korea Attempts to Break Free from NVIDIA’s Monopoly"

Reuters Yonhap News

Reuters Yonhap News

View original image

Countries around the world, including South Korea, are developing 'Sovereign AI' in an effort to protect their data sovereignty. However, analysis shows that most of these efforts remain dependent on Nvidia, as Nvidia chips—mostly supplied by the American company—are widely used.


On August 5 (local time), market intelligence firm Counterpoint Research reported in its 'Sovereign AI Large Language Model (LLM) Research Report' that Nvidia chips accounted for a 92.4% share within sovereign AI initiatives worldwide. AMD held second place with 4.1%, followed by Cerebras at 1.7%. While countries are investing heavily to develop AI models in order to reduce reliance on the technologies, data, and infrastructure of American big-tech companies such as OpenAI and Google, they are, paradoxically, still dependent on chips from leading U.S. AI semiconductor firms. These findings are based on analysis of more than 170 sovereign AI models from 55 countries and regions, including Asia-Pacific, the Middle East, Europe, and Africa, that are competing to build their own AI models and dedicated infrastructure, excluding the U.S. and China.


Mark Einstein, Principal Analyst at Counterpoint Research, stated, "While hosting for sovereign AI (such as server operations) is increasingly being localized, the sourcing of AI chips has not followed the same pattern." He also observed, "Most domestic LLM developers are drawn to Nvidia's CUDA ecosystem, choosing Nvidia AI chips to train their large language models."

"92% of Global Sovereign AIs Rely on NVIDIA Chips" View original image

The report also highlighted the independent approaches taken by some Korean companies. South Korean AI startup Moreh developed 'Llama-3-Motif-102B' based on semiconductors from AMD, a rival of Nvidia. Motif, a Moreh subsidiary, is also building LLMs on the AMD platform. Another Korean AI startup, Upstage, has adopted the AMD GPU MI355. In the market for inference chips, Korean semiconductor company Rebellions is increasing its market share through a partnership with SK Telecom.



Nvidia's competitors, including AMD and Cerebras, are also striving to boost their market shares. Counterpoint Research explained, "AMD is targeting niche markets in the European Union and Australia, while Cerebras is gaining prominence in the United Arab Emirates (UAE) market." The report continued, "Qualcomm has joined forces with Humane in Saudi Arabia, and Google is providing its proprietary chip—the Tensor Processing Unit (TPU)—to an AI startup in Israel. Japan is also utilizing its self-developed 'Fujitsu A64FX' ARM CPU to train LLMs."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing