AMD used its Seoul event to pitch a broad “hybrid AI” strategy, spanning data-centre processors and local AI systems. The practical evidence is more than a new product list: Naver Cloud says it adopted about 35,000 AMD EPYC processors last year, after starting with 13,000 in 2023.
AMD Watch analysis
What happened
At its Advancing AI World Tour in Seoul on 4 November, AMD introduced its sixth-generation EPYC server processors, Ryzen AI Max+ Pro 400 chips and Instinct MI350P GPU. AMD says the MI350P can be added to existing data centres through PCI Express to expand inference capacity, rather than requiring a wholesale infrastructure replacement.
Naver Cloud director Park Young-hee said the company now operates about 250,000 servers across 13 data centres in South Korea. Naver’s own assessment, as reported by BigGo Finance, found its EPYC platform used about 20 to 25 per cent less rack space and around 700 watts compared with 800 watts for the Intel-based platform. Naver plans to bring 144-core EPYC 9005 processors to its public cloud in the second half of this year.
AMD also presented a local-AI example from South Korean company Clebby. Its Uniform document-processing engine runs on Ryzen AI Max+ 395 hardware, with a separate OCR-based system handling images and tables. Clebby says it is conducting proofs of concept with education offices and design firms, and is pursuing an on-premise appliance. The account appeared in BigGo Finance’s report on the Seoul event.
Why it matters
The customer figures make AMD’s server push more tangible than a promise to compete. They also show why AI infrastructure is not just a contest over accelerators: power, rack space, CPUs and the option to run models locally all affect what organisations can deploy.
Naver’s figures are its own assessment, not an independent comparison, and Clebby’s projects are still proofs of concept. But the examples point to two concrete questions for buyers: whether AMD’s claimed efficiency advantages hold up across workloads, and whether local AI can meet organisations’ needs without sending sensitive documents to external cloud services.
Our read
“Hybrid AI” can sound like a slogan wearing a server badge. Here, there are actual deployments and named systems behind it, though the strongest performance claims still come from the companies involved. Naver’s next public-cloud rollout will be a more useful test than another stage presentation.
What to watch
- Whether Naver Cloud deploys the planned EPYC 9005 systems on schedule.
- Whether AMD publishes comparable, independently tested power and performance results.
- Whether Clebby’s proofs of concept become operational on-premise deployments.
Discussion spark: Should organisations prioritise lower power and rack-space costs when choosing server CPUs, or is sticking with a familiar platform worth the premium?
Sources and evidence
- AMD Targets 'Hybrid AI' With 6th-Gen EPYC, Ryzen – Naver Cloud Deploys 35,000 Units – BigGo Finance (4 November 2026, 08:12 UTC)
- AMD Teams Up With FuriosaAI and MangoBoost to Build Heterogeneous AI Cluster in Seoul Next Spring – BigGo Finance (4 November 2026, 08:12 UTC)
Independent WittyWires tracker for public updates about AMD. Not affiliated with or endorsed by AMD; this is not an official account.