Korea's AI Funding Boom Concentrates on Chips and Models, Starving Services
Korean startups raised 1.02 trillion won in October, but five deals took 66 percent. AI chip and model firms now absorb 60.2 percent of AI funding as services collapse to 21 percent.

Why it matters
- Korean startups raised 1.02 trillion won ($766 million) across 86 deals in October; the five largest deals took 66 percent (670.7 billion won).
- Startup funding in the first 10 months reached 11.7 trillion won, up 91 percent year-on-year, while deal count fell 28 percent to 1,008.
- AI semiconductor and model developers' funding share rose from 25.6 percent in 2025 to 60.2 percent in the first half of 2026; AI services fell from 54.3 percent to 21 percent.
- The top three AI funding recipients took 55.8 percent of Korean AI startup funding in the first half of 2026, up from 29.6 percent in 2025.
- FuriosaAI raised 260 billion won and Dream Labs raised 120 billion won in October.
Korean startups and small and medium-sized enterprises raised 1.02 trillion won ($766 million) across just 86 deals in October, and the five largest deals absorbed 670.7 billion won of that total — 66 percent — according to market tracker The VC. The figures expose a structural shift in one of Asia's most active startup ecosystems: capital is flooding into AI, but only into a narrow band of companies that own chips, models or other foundational technologies, while AI application developers increasingly cannot raise money at all.
By Rebecca Stone
Where did October's money go?
The VC's monthly tally shows two companies dominated the month. Neural processing unit developer FuriosaAI raised 260 billion won, and physical AI firm Dream Labs raised 120 billion won. Together, the two rounds represent more than a third of all Korean startup funding in October.
The concentration holds even when the largest deal is removed. The VC reported that AI and robotics accounted for 53 percent of total October investment, and the share still stood at 37 percent after excluding FuriosaAI. In other words, the sector's weight in the market does not depend on a single outlier; it reflects a broad redirection of capital toward AI hardware and foundational layers.
How extreme is the concentration?
The October numbers extend a trend visible across the full year. Korean startup investment in the first 10 months reached 11.7 trillion won, up 91 percent year-on-year, but the number of deals fell 28 percent to 1,008. More money is chasing fewer companies.
A venture capital official described the dynamic bluntly: "While AI startups are attracting significant investment capital overall, the number of companies achieving successful fundraising rounds is limited to several top players. This is particularly so among deep-tech companies having core technologies, which are attracting more investment than companies focusing on services using AI."
The Software Policy & Research Institute (SPRi) quantified the same pattern in its report on the Korean AI venture market. The three companies receiving the largest AI investments accounted for 55.8 percent of total Korean AI startup funding in the first half of 2026, nearly double their 29.6 percent share in 2025.
The composition of that funding has inverted. Investment going to AI service and application developers fell from 54.3 percent in 2025 to 21 percent in the first half of 2026. Over the same period, the combined share for AI semiconductor and model developers surged from 25.6 percent to 60.2 percent — a swing of nearly 35 percentage points in under a year.
Why are semiconductors and models pulling away?
SPRi's report identified the specific deals that moved the market. "The domestic AI venture capital market is quickly reshaping from an application service-oriented structure toward AI semiconductors and models," the report read. "Large-scale follow-on investments in AI semiconductor companies, including Rebellions, FuriosaAI and XCENA, as well as AI model developer Upstage, have rapidly shifted the focus of domestic investment toward models and semiconductors in the first half of 2026."
That list names the beneficiaries of the boom: chip developers Rebellions, FuriosaAI and XCENA, and Upstage, a Korean AI model developer. Each attracted large follow-on rounds that, in aggregate, redrew the map of domestic venture funding in six months.
What is squeezing AI application developers?
Industry officials attributed the shift to two forces, both rooted in the rapid improvement of general-purpose AI models such as OpenAI's GPT-6 Astra and Anthropic's Claude Fable.
The first force is absorption. General-purpose models now perform tasks that once required dedicated products. "As general-purpose AI models become capable of everything from image generation to coding, they are absorbing functions that previously required separate applications," an industry official said. "Companies are now developing software tailored to their own business needs more easily with vibe coding, narrowing the opportunities for AI application developers to charge for their standalone services."
The second force is replication. A third industry official pointed to reverse engineering — analyzing the structure and operation of existing software to reproduce its functionality with new code — as a threat that AI has made cheaper and faster. "Advances in AI have made reverse engineering easier, allowing competitors to analyze the features and workings of a differentiated service and develop similar offerings with AI," the official said. "As ideas and features alone can no longer sustain a competitive advantage for long, investors are favoring companies with proprietary technologies that are difficult to replicate."
The logic is straightforward for investors. An application built on top of a general-purpose model can be absorbed by that model or cloned by a competitor within months. A semiconductor design or a foundation model represents capital, engineering depth and intellectual property that cannot be replicated with a prompt.
Why does this matter beyond Korea?
The Korean figures offer one of the clearest quantitative pictures yet of a shift now reshaping AI funding globally: the migration of capital from the application layer, which dominated early AI investment, to the infrastructure layer of chips and models. SPRi's data shows the application share of Korean AI funding collapsing from a majority position to roughly a fifth in less than a year — a compression that markets in the United States and Europe are watching in their own portfolios.
The stakes are highest for the service layer of the Korean startup economy. A 91 percent year-on-year increase in total startup funding, to 11.7 trillion won, suggests a booming market. A 28 percent fall in deal count, to 1,008, tells the opposite story for the majority of founders: fewer companies are receiving any institutional money at all.
The question heading into 2027 is whether application developers can find defensible positions — vertical expertise, proprietary data, or deep workflow integration — that general-purpose models and reverse engineering cannot easily erase, or whether the Korean market's current shape, in which three companies take more than half of all AI funding, becomes the sector's permanent structure.
Source: feed.koreatimes.co.kr
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
728 articles
Related articles
- INFevo Raises Over RMB 100M to Build AI Virtual Cells and Robotic Labs
- Foreign investors dumped $19.31 billion in Korean stocks in October
- Santander Counts €194 Million From AI as Earnings Season Puts Numbers on the Hype
- Agile Robots Closes $500M Series F at $8.5B Valuation
- Intel Can Match TSMC on Physics, Not on Economics