Blog

DEEP RESEARCH · S2W

S2W: From Dark-Web Intelligence to Industrial AI

A post-IPO review of DarkBERT, CTI, and SAIP as S2W expands security deep tech into enterprise AI

Written: 2025-12-11 · Post-IPO growth strategy analysis · Naver Blog

Investment decisions are your responsibility. This material is research, not a recommendation to buy or sell.

0. Bottom line first

S2W's core story is productizing analytical capability proven in the hostile dark-web data environment into XARVIS, QUAXAR, and SAIP. Near-term losses and overhang are real risks, but SaaS revenue at 81.4%, retention around 96%, and the INTERPOL reference support the thesis of monetizing trust.

Official fact: The source says S2W was founded on September 17, 2018 by researchers from KAIST NSS, became an official INTERPOL partner in 2020, and released DarkBERT in 2023. It listed on KOSDAQ on September 19, 2025 and raised about KRW 20.8 billion at an IPO price of KRW 13,200.

Interpretation: This is not only a security stock; it is closer to a data-intelligence platform company. Profitability is still unproven, as shown by the 3Q 2025 cumulative operating loss of KRW 5.159 billion.

S2W data-intelligence engineMulti-domain cross-analysis
CollectionDark web · OSINT · crypto
AnalysisDarkBERT · knowledge graph
ProductsXARVIS · QUAXAR · EYEZ
ExpansionSAIP industrial AI
Result: transfer B2G trust into the private enterprise AI market

1. Investment thesis: monetizing trust

In 2025, the technology market faces both the productivity of generative AI and the risks of data sovereignty and security. The source cites Deloitte's finding that 77% of companies worry about risks from generative AI adoption. S2W positions SAIP to reduce confidential-data leakage, prompt security, and hallucination risk.

  1. Technology moat: DarkBERT, described as the world's first dark-web language model, and validation at ACL and NAACL.
  2. Trust transfer: references from high-trust customers such as INTERPOL and Korea's National Intelligence Service.
  3. Revenue-model shift: subscription SaaS at 81.4% of total revenue as of 1H 2025 and customer retention around 96%.

2. Growth path and capital structure

  • Phase 1(2018-2020): dark-web crawling and analytics, official INTERPOL partnership in 2020.
  • Phase 2(2021-2023): QUAXAR launch, Series B funding, S2W USA establishment, and DarkBERT release in 2023.
  • Phase 3(2024-present): SAIP industrial generative AI platform, KOSDAQ listing on September 19, 2025, and expansion toward Japan, Singapore, and the Middle East.

Total shares outstanding were 10,587,956 at 3Q 2025. The largest shareholders and related parties held about 22.42%, including CEO Seo Sang-duk at 14.10%, with a voluntary three-year lock-up. Financial investors such as Stonebridge DX Business Reorganization Fund at 6.42% and LB Technology Finance Fund No.1 at 5.11% may create overhang after lock-up expiry. Daishin Securities must hold 47,400 shares, or 3% of the offering, for three months.

The roughly KRW 20.8 billion proceeds are planned for infrastructure expansion of KRW 7.5 billion, global marketing and bases of KRW 10.5 billion, and R&D upgrades of KRW 2.5 billion. The source views the H200-class GPU server plan as key to executing the technology roadmap.

3. Product portfolio

ProductTargetRole
XARVISSecurity and investigation agenciesAnalyzes criminal and threat intelligence from the dark web and closed channels
QUAXAREnterprisesCTI based on external threat data to detect leaks and attack signals early
SAIPIndustrial enterprise AICombines security guardrails and RAG for safer generative AI use
EYEZVirtual-asset analysisTracks blockchain transactions and illicit fund flows

4. Competitive map

CategoryCompetitorsS2W differentiation
Global CTIRecorded Future, Mandiant(Google), PalantirReasonable pricing, Korean and Asian-language specialization, and dark-web domain expertise
Korean securityAhnLab, Genians, Sands LabSpecialized in external threat intelligence and AI analytics, with SAIP extending into general industrial AI
AI platformNaver HyperCLOVA X, LG AI ResearchMiddleware/application layer for applying foundation models safely in enterprises, centered on security guardrails

5. Financials and turnaround conditions

Item (KRW mn)3Q 2025 cumulative20242023
Revenue6,3869,6126,267
Operating profit(loss)(5,159)(4,498)(5,671)
Net profit(loss)(5,113)(14,103)*(12,574)
Total assets31,78814,90315,869
Total liabilities8,31810,39045,619
Total equity23,4704,513(29,750)

The source explains that large historical net losses and capital impairment were largely accounting effects from RCPS derivative valuation losses and were resolved through the listing process.

Interpretation: A turnaround requires SAIP industry references, real orders in Japan, Singapore, and the Middle East, and subscription growth translating into operating leverage.

6. Growth strategy and risks

S2W aims to overcome the limits of Korea's market through an Asian Belt strategy linking Japan, Singapore, and the Middle East. SAIP starts from security but can expand horizontally into manufacturing, finance, public sector, and other industry-data platforms.

  • Overhang: VC lock-up expiry and early post-listing float pressure.
  • Policy and budget reliance: B2G references are strong but government budget cycles matter.
  • Technology competition: Microsoft, Google, Palo Alto Networks, and other large vendors are entering generative-AI security.

7. Terms

  • CTI: cyber threat intelligence, used to collect and analyze threat data for response.
  • OSINT: intelligence collection and analysis from open sources.
  • RAG: retrieval-augmented generation, where AI references external databases to improve accuracy.
  • Dark Web: web areas accessible only with special browsers, often abused for criminal activity because of anonymity.