AWS Korea, Kakao outline enterprise AI strategy at AI WAVE 2026
Kakao's Kanana model is trained on Amazon SageMaker HyperPod, while service inference runs on government-supported on-premises GPUs, a Kakao executive said.
At AWS AI WAVE 2026 on 23 October 2026, AWS Korea and Kakao presented their approaches to enterprise AI strategy and infrastructure at the Grand InterContinental Seoul Parnas in Gangnam, Seoul, according to inews24.com.
Kim Ki-wan, director and head of AI business at AWS Korea, said in a keynote that the success of corporate AI adoption depends more on strategy, data and operating systems than on the model itself. He said 88% of companies have adopted AI but only 6% have produced actual results.
Kim presented a four-layer enterprise AI framework covering strategy, data, platform and applications. He cited agent core, Amazon Bedrock and knowledge base features as the basis for centrally managing access rights, costs and security across multiple models.
Kakao AI communications and synergy vice president Kim Se-woong said the company trains its Kanana model on Amazon SageMaker HyperPod on AWS, while service inference runs on-premises using government-supported GPUs.
He added that Kakao plans to expand its "AI for everyone" initiative into a platform linking public services, specialized AI services and an AI agent registration and distribution system.
Quick answers
When and where was AWS AI WAVE 2026 held?
The event took place on 23 October 2026 at the Grand InterContinental Seoul Parnas in Gangnam, Seoul.
Where does Kakao train and run its Kanana model?
Kakao trains Kanana on Amazon SageMaker HyperPod on AWS, while service inference runs on-premises using government-supported GPUs, according to its vice president of AI communications and synergy.