Simular sees Korea’s hardware edge in AI agent era

Simular CEO Ang Li said Korea's hardware manufacturing strength could give it an edge in the AI agent era, where autonomous computers that act on behalf of users are replacing chatbots. The startup, backed by Samsung NEXT, sees Korean companies like Samsung, LG, and Hyundai as potential partners for providing the physical 'bodies' for AI agents.

CEO Ang Li says shift from chatbots to 'autonomous computers' could open new opportunity Korea’s next artificial intelligence opportunity might not stop at chips, smartphones or devices. It may hinge on whether those devices can become the “bodies” for AI agents that operate computers and carry out work on behalf of humans. For Ang Li, CEO and co-founder of California-based AI agent startup Simular and a former Google DeepMind researcher, the shift now underway in AI is not simply from one chatbot to another. It is from AI that responds to users to AI that can use computers for them. “The question is, do humans still need to use the computer if AI has become so capable in the future?” Li said in an interview with The Korea Herald in Seoul on Thursday. “A chatbot is only solving one piece of the puzzle, like speaking. But what about typing and moving the mouse and keyboard? That kind of action can also be automated by AI.” “That’s the problem we try to solve, and we call it autonomous computers,” he said. Simular, founded in 2023 by Li and other former Google DeepMind researchers, has raised about $27 million to date from investors including Felicis, NVentures, South Park Commons and Samsung NEXT. Its first product, Sai, is an always-on AI agent built on computer use. Rather than simply generating text, it automates workflows by controlling local and cloud computers — clicking, typing, navigating apps and running terminals and code. The aim is to move AI from conversation to execution. Simular has not announced a Korean office or local commercial rollout. But Li said Korea is already an important long-term market and potential partner ecosystem for the startup. “I think in the long term, definitely,” he said when asked about doing business in Korea or with Korean partners. “It’s not a yes or no question. It’s more about when — when is the right time to enter Korea?” As a startup, Li said, Simular has to be selective about where it deploys its resources. Still, he suggested Korea could move faster if the right partner appears. “It could be much faster than we expected,” he said. “It’s definitely a yes. It’s just about what’s the right time.” The company already has a Korean connection through Samsung NEXT, which invested in Simular’s seed round. Li said Samsung NEXT, the venture capital and innovation arm of Samsung Electronics, has helped the company better understand Korea’s technology ecosystem and business culture. “Samsung NEXT helps introduce people around. It’s definitely very helpful for me to know more about the ways to work with Korean companies,” Li said. Simular also held a public event Friday with the Artificial Intelligence Institute at Seoul National University, introducing its technology progress and research roadmap with members of its research and engineering teams. Li’s interest in Korea is not limited to market entry. He sees the country’s strength in semiconductors, memory, smartphones, displays, home appliances and automobiles as directly relevant to the AI agent era. “We are working on initially an intelligence layer, which is the software,” he said. “And now you can see the agents become more concrete, meaning the agent has a body.” “Basically, intelligence needs the body,” he said. “And Korean companies have this huge advantage — you manufacture hardware. So hardware is basically the body.” That could be significant for companies such as Samsung Electronics, LG Electronics and Hyundai Motor, all of which already have global device footprints. If AI agents begin operating across PCs, smartphones, cars and connected appliances, those products may become more than hardware endpoints. They could become the environments in which the agents carry out tasks. Li said the industry has spent the past few years focused heavily on GPUs, but computer-use agents point to a broader infrastructure challenge. “In the past few years, everyone has been talking about GPUs,” he said. “But when we develop this agent, we know that there’s a bottleneck in CPU and memory as well. So these are the strengths in Korea.” For Simular, that makes Korea more than a market to watch: It is one of the few places where the hardware base for agent-driven computing is already in place, if the right software layer and partners emerge. yeeun@heraldcorp.com