Naver Cloud and Vietnam's FPT signed a Memorandum of Understanding to scale AI infrastructure across the Asia-Pacific, linking FPT's AI factory model - already running on 96.41 petaflops of combined Linpack capacity across two sites - with Naver's HyperClova X model. The pact covers joint go-to-market for enterprise AI, technical exchange on GPU optimization, and HyperClova X distribution into FPT's footprint.

For practitioners, the immediate story is access. FPT operates AI data centers in Vietnam and Japan that ranked 40th and 53rd, respectively, on the Top500's most recent list of the world's most powerful supercomputing systems. The Vietnam facility runs Intel Xeon Platinum 8462Y+ 32-core 2.8GHz chips alongside Nvidia H100 GPUs and posts a Linpack score of 46.56 petaflops. The newer Japan site pairs Xeon Platinum 8558 48-core 2.1GHz processors with Nvidia H200s and edges 7% higher at 49.85 petaflops.

What's actually in the MoU?

Three threads. First, joint go-to-market for enterprise AI in APAC, anchored on FPT's AI factory operating model. Second, technical exchange programs focused on GPU infrastructure tuning and inference performance optimization. Third, distribution of HyperClova X into FPT's enterprise footprint.

"FPT has invested in advancing our AI cloud capabilities through AI factories, supported by engineering teams with proven experience across multiple markets," said Le Hong Viet, CEO of FPT Smart Cloud at FPT Corporation. "We look forward to working with Naver Cloud to bring AI models closer to enterprise customers and develop new ways for businesses across Asia-Pacific and beyond to adopt AI with confidence."

Yuwon Kim, CEO of Naver Cloud, framed the deal as a way to combine stack maturity with regional reach. "Through this collaboration with FPT, Naver Cloud can now combine the AI technologies and experience it has built up with FPT's AI factory and enterprise business capabilities," Kim said. "Building on both companies' technology and business strengths, we will go beyond Korea and Vietnam to address customers' diverse AI needs across the APAC market."

What does FPT bring to the table?

FPT launched its FPT Fornix HCM02 Data Center at the Ho Chi Minh City High-Tech Park in August 2025, bringing its facility count in Hanoi and Ho Chi Minh City to four. Each site feeds capacity into the same AI factory stack - an integrated compute, storage, and model-serving blueprint geared for enterprise deployment rather than consumer workloads.

For builders weighing inference economics, the GPU mix matters: H100 silicon in Vietnam, H200 silicon in Japan. That generational split will shape which workloads route to which site, particularly for FP8 and lower-precision paths where H200 pulls ahead.

Where does Naver stand today?

Naver operates public cloud through its multi-region Ncloud and sells on-premises capacity via Neurocloud. It lists six cloud regions in South Korea alone, plus one each in Japan, Germany, Singapore, and the US West Coast. In July, Naver said it would expand its Sejong data center to 200MW of capacity.

Pipeline regions include the US East Coast, Vietnam, Taiwan, and Thailand, with the company also flagging expansion into the Middle East and North Africa.

What changes for regional buyers?

If you're sizing APAC inference capacity, the practical questions are which HyperClova X endpoints land at which FPT site, what latency you get from Ho Chi Minh City or Hanoi versus Tokyo-area facilities, and how FPT's AI factory pricing compares with hyperscaler alternatives in Singapore. The MoU doesn't answer those yet - but it does signal that Korean model weights will travel inside FPT's deployment envelope, a new option for regional buyers who've defaulted to US or Chinese providers.