Asteromorph raises $31.4M seed to build scientific superintelligence
What's the deal? Asteromorph, a South Korean AI research lab building what it calls "scientific superintelligence," has closed a $31.4M (roughly ₩47 billion) seed round. The investment was led by BonAngels Venture Partners, with participation from FuturePlay, Mirae Asset Venture InvestmentDealroom has a profile for this one. Try Dealroom →, Mirae Asset CapitalDealroom has a profile for this one. Try Dealroom →, IMM InvestmentDealroom has a profile for this one. Try Dealroom →, Premier PartnersDealroom has a profile for this one. Try Dealroom →, Korea Development BankDealroom has a profile for this one. Try Dealroom →, KDB CapitalDealroom has a profile for this one. Try Dealroom →, SV InvestmentDealroom has a profile for this one. Try Dealroom →, and Sticks Ventures. The company had previously raised approximately ₩5 billion.
Why now? Asteromorph is developing an autonomous scientific system powered by its foundation model, Spacer. The model combines a base LLM with deep domain knowledge and multi-step reasoning to independently generate research ideas and develop scientific hypotheses in biology and chemistry — going beyond the human-intuition bottleneck that limits conventional AI tools.
The startup's flagship effort, the "Library Project," aims to run its autonomous science system at scale, uncover fundamental mechanisms of life, and compile findings into a structured database.
The company is led by CEO Lee Min-hyung, who joined Seoul National University's medical school as a researcher at age 16, skipped high school and undergraduate studies, and entered a combined master's-PhD programme at SNU's College of Pharmacy before dropping out to found Asteromorph. The core research team consists of twentysomethings from Korea's top STEM programmes, including International Olympiad medallists.
What could go wrong? "Scientific superintelligence" is a bold claim at the seed stage. Autonomous hypothesis generation in biology and chemistry demands not just capable models but rigorous experimental validation — something far harder to scale than software. The startup will need to prove its system can produce discoveries that hold up in the lab, not just on paper.
The signal: A $31.4M seed is exceptionally large for a South Korean AI startup, underscoring how autonomous scientific research — where AI generates and tests hypotheses rather than merely assisting human scientists — is emerging as a distinct, well-funded category. The breadth of the investor syndicate, spanning venture funds, state-backed development banks, and asset managers like Mirae Asset, suggests institutional conviction that the convergence of foundation models and deep scientific domains could yield outsized returns before Western incumbents fully stake out the territory.
Read more: wowtale.net
