S2w

S2W was founded in 2018 as S2W Lab Inc. and renamed to its current form in 2021, listing on KOSDAQ on August 19, 2025 under the technology growth company special listing track.

The firm is an AI data intelligence company that analyzes unstructured data collected from hidden channels such as the dark web and Telegram, based on knowledge-graph-driven cross-data analysis technology, and has developed its own dark-web-specialized AI model 'DarkBERT' and security-document-specialized model 'CyBERTuned.' Its core solutions consist of three products: 'XARVIS,' a security platform for public and government institutions, 'QUAXAR,' an enterprise security solution, and 'SAIP,' an industry-specific AI platform.

As of 2024, revenue mix stood at 42.1% for QUAXAR (enterprise security AI), 40.5% for XARVIS (security AI), 15.2% for SAIP (industry-specific AI), and 2.3% for other.

A government- and public-institution-centered customer base is regarded as the starting point of its growth strategy, with the company having built references by supplying solutions to global government agencies including Interpol, Japan, Singapore, and Saudi Arabia, using this foundation to expand into private enterprise customers.

Founded by a team centered on KAIST researchers, the company is recognized as one of the largest collectors and analyzers of dark-web data globally, and drew industry attention in 2023 by unveiling 'DarkBERT,' the world's first dark-web-specialized AI language model, at ACL, a top computational linguistics conference.

Its revenue structure carries a high SaaS weighting; at the time of listing the company had posted an average annual revenue growth rate of 81% since founding, with SaaS accounting for 69% of revenue.

Competitively, it coexists with large domestic security firms such as AhnLab while emphasizing technical differentiation in the narrow dark-web-specialized domain, alongside global partnerships such as integration with Microsoft's 'Copilot for Security.'

2025 consolidated revenue came in at KRW 10.089 billion, up 5.0% from KRW 9.612 billion the prior year, but the operating loss widened to KRW 5.640 billion from a loss of KRW 4.498 billion the year before.

In contrast, the net loss attributable to owners narrowed sharply to KRW 5.291 billion from a loss of KRW 14.103 billion, a 62.5% reduction.

FnGuide attributed the pattern to increased revenue from defense, security, and cyber-threat-response solutions along with expanded enterprise adoption of the SAIP platform, offset by a wider operating loss driven by higher R&D investment and market-expansion costs.

By product, enterprise security AI 'QUAXAR' grew to KRW 4.7 billion (+16.5% year over year) and security AI 'XARVIS' grew to KRW 4.3 billion (+10.4%), while industry-specific AI 'SAIP,' which generated about KRW 1.5 billion in 2024, produced almost no revenue in 2025 amid intensifying competition from low-cost LLM-based solutions.

On a quarterly basis, 4Q25 revenue was KRW 3.702 billion with an operating loss of KRW 0.482 billion.

Revenue then fell sharply to KRW 2.177 billion in 1Q26 versus the prior quarter while the operating loss jumped to KRW 2.252 billion, and in 2Q26 revenue declined slightly further to KRW 2.085 billion while the operating loss widened again to KRW 3.163 billion.

As a result, the cumulative operating loss for the first half of 2026 reached roughly KRW 5.414 billion, nearly matching the entire full-year 2025 operating loss of KRW 5.640 billion.

The net loss attributable to owners also widened for two consecutive quarters, reaching KRW 1.955 billion in 1Q26 and KRW 3.069 billion in 2Q26, marking a phase where stagnant revenue and widening losses are occurring simultaneously.

The global big-data-analytics market is expected to grow at a compound annual rate of 12.9% through 2032, suggesting continued market-expansion opportunities for AI solutions.

In Korea, demand for intelligent security solutions is expanding as AI-based hacking threats grow more sophisticated, with policy support strengthening as the Ministry of Science and ICT and the Korea Internet & Security Agency (KISA) published an 'AI Security Red-Teaming Guide' presenting attacker-perspective verification procedures and methodologies.

Indeed, the National Intelligence Service, National Police Agency, KISA, and the Financial Security Institute jointly issued a 'joint cybersecurity advisory' together with private security firms including AhnLab, S2W, NKWhitehat, and Plainbit, illustrating the company's position within the government-private cooperation network.

According to the company's own tally, advanced persistent threat (APT) issues attributed to state actors such as North Korea, China, and Russia numbered 158 in the first half of 2026, up 7.5% from the prior half, suggesting structurally growing demand for countering state-sponsored threats.

In the competitive landscape, large integrated domestic security firms such as AhnLab stand alongside niche specialists like Raonsecure, SandsLab, and Huneshion, with S2W positioned to differentiate itself in the narrow domain of hidden-channel data analysis across the dark web and Telegram.

The company is also noted for a relatively high overseas revenue share among domestically listed AI firms, with export revenue reaching 25% on a cumulative basis through the third quarter of 2025.

However, competition from low-cost, general-purpose LLM-based solutions is intensifying in industry-specific AI areas such as SAIP, making it a key point to watch whether the company can sustain its premium security-specialized positioning.

The company said it is significantly expanding its AI service vulnerability assessment business in response to the agentic AI era, explaining that it provides diagnostic services applying the international standard 'OWASP Top 10 for LLM,' based on offensive-security expertise that applies real large language model (LLM) attack techniques collected from the dark web, Telegram, and hacking forums directly to inspection processes.

Concurrent with launching this business, the company said it had won an AI service vulnerability assessment contract with a large domestic financial institution, and plans to expand the scope of diagnostic services into various industries going forward, starting with the financial sector.

Sell-side estimates vary, but a number of analysts projected 2026 full-year revenue in the range of KRW 19.5–24.0 billion with a swing to operating profit (roughly KRW 1.2–3.9 billion).

However, actual combined revenue for the first half of 2026 was only about KRW 4.262 billion, roughly one-fifth of that annual target, while the first-half operating loss grew large enough to nearly match the entire full-year 2025 loss, making the strength and timing of any second-half turnaround the key variable.

In an April 2026 report, Shinhan Investment & Securities noted that XARVIS retention stood near 100% and QUAXAR near 91%, arguing that high SaaS-based customer retention would serve as a foundation for earnings improvement, and projected overseas revenue would more than double year over year on the back of a large Indonesian government-related deal and customer wins in Japan, Taiwan, and Saudi Arabia.

The same report also forecast that construction-type orders for the security-focused premium AI 'SAIP' would begin materializing from the second half of 2026, as corporate security incidents from the prior year were expected to lead to a stricter government regulatory stance.

These, however, are forecasts as of the report date, and actual second-half order intake and revenue recognition will need to be confirmed through future disclosures.