Google Cloud showcases AI successes in Seoul

Google Cloud Korea director Chi Ki-sung speaks during a media briefing at COEX in Seoul, Tuesday. Courtesy of Google Cloud

Since launching a regional office in Seoul in 2020, Google Cloud has continually expanded its computing capabilities to meet the evolving needs of businesses in Korea.

Celebrating its fifth year, the company held Cloud Day Seoul on Monday and Tuesday at COEX in southern Seoul, bringing together people in tech to explore its latest advancements in artificial intelligence (AI) and cloud technology, and sharing success stories from its client enterprises.

“Google Cloud is the only company in the industry that provides a full-stack service from hardware to software as a service. Our distinguishing feature is that all intellectual property is proprietary,” Google Cloud Korea director Chi Ki-sung said during a media briefing at the event’s venue on Tuesday.

“We provide companies with not only models and hardware, but also open and interoperable software tools and platforms to support the development of multi-agent systems.”

The Seoul office boasts large-scale interconnected infrastructure, equipped with servers, custom silicon chips, storage devices and advanced networking equipment. It serves as the physical backbone for delivering Google Cloud’s high-performance services, playing a central role in ensuring the availability, security and accessibility of applications and systems.

The office is also connected to Google’s global private network that spans more than 200 countries and regions with over 2 million miles of fiber-optic cables, providing a resilient global infrastructure for AI-driven workloads and operations.

A participant demonstrates Uplus' ixi-O in the Gemini Playground exhibition at COEX in Seoul, Tuesday. Courtesy of Google Cloud

Last week, the office officially launched its Gemini 2.5 Flash model, enabling AI inference and machine learning processing to be carried out entirely within Korea.

“By offering sovereign cloud solutions, organizations now have the choice to operate services in Seoul, if data boundaries are important to them,” Chi said.

Chi announced that Google Cloud will soon be rolling out Ironwood, a tensor processing unit specifically designed for large-scale AI inference.

“Ironwood marks a significant shift in the advancement of AI and the infrastructure that supports it,” he said.

“This development signals the ‘era of inference,’ where the industry is transitioning from reactive AI models, which provide information in real time for humans to interpret, to models that proactively generate analysis and insights.”

Korean AI start-up WRTN Technologies joined the event to share its testimony of implementing Google Cloud’s product in its latest service WRTN 3.0, which features strengthened AI search, productivity tools and personalized AI functions.

“We needed a large language model for the new service that has strong performance, reasonable price and high stability, and the Gemini 2.5 model line satisfied these requirements,” WRTN Chief Operating Officer Yoo Young-joon said.