AMD and NAVER Cloud have forged a new strategic partnership to expand artificial intelligence infrastructure in South Korea, a move that aligns closely with the global race for so-called sovereign AI. Announced on March 18, the deal includes NAVER Cloud increasing its use of AMD EPYC processors, early access to upcoming AMD Instinct MI455X GPUs, and joint efforts to optimize AI services and software stacks on AMD and ROCm platforms.
There is more depth to this news than a standard press release suggests. NAVER Cloud is not a minor player in the Korean ecosystem: it is part of NAVER, one of the country’s major tech conglomerates, and markets HyperCLOVA X, its proprietary foundational model. Meanwhile, South Korea has been strengthening its strategy for building national AI and computing capabilities with greater control over data, infrastructure, and software for months.
A deal combining CPU, GPU, and software
According to AMD, NAVER Cloud will expand its deployment of EPYC processors to include the upcoming 6th-generation EPYC “Venice,” a family not yet commercially launched but already considered a key element in AMD’s roadmap for data centers and AI systems. Additionally, AMD will provide NAVER Cloud with early access to the MI455X, the next-generation accelerators that will also be part of the rack-scale Helios platform.
This detail is significant because the announcement isn’t just about updating servers. AMD’s proposal involves a collaboration touching all three critical layers of the new AI data center: CPUs for orchestration and general services, GPUs for training and inference, and software for running and optimizing models. In this context, the software layer relies on ROCm, AMD’s open ecosystem aimed at establishing itself as an alternative to NVIDIA, which dominates much of the accelerated market.
It’s also worth noting that this move doesn’t mean NAVER Cloud will depend solely on AMD. NVIDIA announced in October 2025 that NAVER Cloud was expanding its AI infrastructure with deployments exceeding 60,000 GPUs, including Blackwell GPUs, and was preparing for the next phase of its sovereign development with open models like Nemotron. In this broader context, AMD’s alliance appears less as a replacement and more as an expansion of NAVER’s technological palette to support its strategy.
South Korea seeks greater control over its AI infrastructure
The term “sovereign AI” has become a key slogan in the sector in 2026, but in South Korea, it has a very specific translation. In 2025, the Ministry of Science and TIC launched the K-Cloud R&D project with a budget of 400 billion won to reinforce infrastructure based on domestic semiconductors, and just days ago confirmed that NAVER Cloud is among the teams advancing to the next phase of the national sovereign foundational models program.
This context helps explain why AMD mentions “national-scale AI capabilities” in its statement. The company is not just selling chips but promoting an open platform capable of powering cloud services, proprietary models, and new enterprise or sector-specific solutions with greater local control. For a country aiming to reduce external technological dependence without isolating from the global market, this proposition makes both political and industrial sense.
Furthermore, for AMD, South Korea is becoming a strategic market. On March 18, AMD also announced an expanded collaboration with Upstage, another Korean player focused on AI models and engines, to enhance sovereign capabilities in the country. Alongside the agreement with Samsung on HBM4 and DDR5 memory for next-generation systems, a clear strategy emerges: to gain ground in a market where national AI, open infrastructure, and provider diversification are now priorities.
AMD aims to strengthen its position in a market still dominated by NVIDIA
From an industrial perspective, this partnership also offers insight into AMD’s current position in the AI infrastructure race. The company ended 2025 with record revenues and declared in February 2026 that it was starting the year with strong momentum driven by EPYC adoption and rapid expansion of its AI data center business. Still, it continues to compete in a market where NVIDIA maintains a dominant position in acceleration, software, and ecosystem.
That’s why agreements like NAVER Cloud’s carry value beyond immediate volume. They enable AMD to place its future roadmap—Venice, MI455X, and ROCm—into real production conversations, not just lab prototypes. They also provide visibility in one of the most critical scenarios today: countries and large cloud providers striving to build their own scalable, independent infrastructure less tied to a single technology stack.
In the short term, this announcement alone won’t drastically shift the global market balance. No figures on volume, investment, or deployed capacity are provided, and both Venice and MI455X are part of products still under deployment or early access. However, it sends a clear signal: the race for sovereign AI is no longer confined to Washington, Beijing, or Brussels. It’s also happening in Seoul, around who supplies the infrastructure, what software supports it, and to what extent a country wants to control how it trains, serves, and governs its own AI.
Frequently Asked Questions
What exactly have AMD and NAVER Cloud announced?
They announced a strategic collaboration to expand AI infrastructure in South Korea through more deployment of AMD EPYC processors, early access to AMD Instinct MI455X GPUs, and joint optimization of AI services and software on AMD and ROCm platforms.
What does “sovereign AI” mean in South Korea here?
It refers to developing AI models, services, and infrastructure with greater national control over data, scalability, operations, and technological capacity. South Korea supports this through public programs like K-Cloud and its sovereign foundational models initiative.
Does NAVER Cloud already use other AI technologies besides AMD?
Yes. NAVER Cloud has also announced expansion of its infrastructure based on NVIDIA GPUs for its next phases of sovereign AI, so the AMD agreement should be seen as an expansion of its technological strategy, not an exclusive replacement.
What role do EPYC Venice and the MI455X GPUs play in this deal?
They represent the most advanced components of AMD’s future data center roadmap for AI. NAVER Cloud will utilize the upcoming Venice generation of EPYC and gain early access to MI455X accelerators for its cloud and production environments.