Upstage, a generative AI company founded by former Naver AI lead Kim Sung-hun, unveiled the "Upstage Company" on the 16th, signaling a strategic shift from pure model development to everyday AI services. The Seoul-based startup, now South Korea's first AI software unicorn at a 2 trillion won valuation, disclosed that new contracts signed in the first half of 2024 have already outpaced last year's total of 15.7 billion Korean won, with over 200 companies across Korea, the U.S., and Japan adopting its technology.
Funding and national AI ambitions
During a press briefing at the Conrad Hotel in Yeouido, CEO Kim Sung-hun said the company recently secured 100 billion Korean won (about $73 million) from the National Growth Fund's Advanced Strategy Fund. This follows its selection as the only startup for the government-led "National Representative AI" independent foundation model project in August 2023, which it passed its first evaluation in January.
Kim pointed to rising U.S. curbs on foreign access to high-performance models - specifically citing Anthropic's restrictions - to stress domestic capability. "Securing independent AI models is essential to counter the U.S. and China's strategic assetization of AI," he said, adding that government support must increase tenfold. The preview release of Upstage's open-source Solar Open 2 model, built under the national project, scored 44.4 on the Artificial Analysis Intelligence Index, which Kim claimed is "comparable to GPT-5 (44.6 points)."
Agents enter the workflow
Moving beyond chatbots, the company introduced Upstage Studio, a platform where users assemble task sequences like Lego blocks to automate multi-step workflows. "The era of conversing with AI is over. Now, it's time to assign tasks to AI," Kim said. The push into AI Agents & Automation signals a broader bet on agent-based tools that IT and development teams can deploy without coding.
That bet expanded with the acquisition of Timely, an AI agent platform already used by over 600 public institutions, local governments, and educational organizations. Timely integrates agents for image and video generation, document conversion, and other tasks on top of language models including Solar. Timely CEO Kim Dae-hwan explained that the platform's agents can be activated for a task "with a single click, without coding."
Daum gets a hybrid search overhaul
Last month's acquisition of portal site Daum - whose search market share had dwindled to 3-4% under Kakao - will now tie into Upstage's generative AI roadmap. Lee Geon-su, CEO of AXZ, the entity operating Daum, demonstrated an upcoming AI Overview service where agents combine keyword context to deliver synthesized answers. Natural language queries such as "recommend a laptop suitable for college students" or "find a restaurant with parking in Seongsu-dong" will trigger agent-driven summarization from search results.
Lee added that agents will also organize time-based news alerts and market trends, and enable context-based Q&A on Daum's news pages. "While overturning the search market is challenging, our goal is to achieve small victories by focusing on user needs," he said. The approach leans heavily on the same Generative AI and LLM capabilities that power Solar and other foundational models.
IPO on the horizon
With the Daum and Timely deals, Upstage is accelerating toward a public listing. It appointed KB Securities and Mirae Asset Securities as underwriters in December and recently added foreign bank UBS. An industry source indicated the company is aiming for an IPO valued as high as 5 trillion won, though the timeline depends partly on evaluations of the national AI project and the performance of its upcoming AI services. A preliminary review filing could come as early as the second half of this year.
Why this matters for IT and development professionals
Upstage's shift to agent-based workflow automation and hybrid search points to a concrete change in how AI gets integrated into enterprise stacks. For developers, the open-source Solar model and the Lego-like Studio platform lower the barrier to building compound AI systems that can handle multi-step tasks - think a chain of prompts, API calls, and document generation - without managing each component from scratch. The Timely acquisition further reinforces low-code agent deployment, which could influence how internal tools get built in organizations that prioritize speed. Keeping an eye on Upstage's national AI project milestones and the Agentic Search direction for Daum can help teams anticipate the next wave of AI-augmented search and productivity patterns.
Your membership also unlocks:
