Seoul Economic Daily EnglishDoubao · 实盘不做把握 5/10不做原因:风险收益比不佳
北京时间·🕒 发布 08:20:56(首尔 09:20)·⟳ 抓取 08:30:41 +9m45s·⚡ 决策 10-21 08:32:27 +1m46s·6.5 天前
ESTsoft 中标济州国际自由城市开发中心(JDC)内网生成式 AI 平台建设项目,金额未披露
方向性正面的中标属实且新鲜(开盘前 1 小时左右发布),但这是单个公共机构 6 个月内网 AI 建设项目、金额未披露,以韩国公共机构同类 sLLM 建设项目量级(通常数亿至数十亿韩元)对照 ESTsoft 约 1200 亿韩元年化收入和约 856 亿韩元市值,直接业绩弹性极小,主要是标杆/叙事价值;公司仍处于经营亏损、属小盘 AI 主题高波动股,在无金额、无可靠实时价的情况下追高风险收益比不足。
📄 原文Seoul Economic Daily English↗ 原文
ESTsoft Wins JDC Deal to Build In-House Generative AI Platform
ESTsoft (047560.KQ) is expanding its business of building artificial intelligence platforms that support internal opera…
不做 ESTsoft(이스트소프트)(047560.KQ)
RISK_REWARD:中标方向正面且新鲜,但单个公共机构 6 个月内网 AI 建设项目金额未披露,对照公司约 1200 亿韩元年化收入与约 856 亿韩元市值,直接业绩弹性预计极小,主要是标杆叙事;公司仍经营亏损、属小盘 AI 主题股,消息开盘前已在韩媒扩散,无可靠实时行情下追高风险收益比不足。
展开详情(成交明细 / 研究档案 / 原文全文 / 决策记录链)
研究档案
结论摘要
ESTsoft 中标 JDC 内网生成式 AI 平台项目,叙事正面且属新拓展的公共机构市场标杆,但合同金额未披露、相对公司收入与市值大概率微不足道,缺乏可交易的业绩弹性,规避追高。
消息传播链
- 英文稿于首尔时间 09:20(UTC 00:20)上线,报道 ESTsoft 宣布中标 JDC 内网生成式 AI 项目;公司口径表述(The company said),金额未披露。
- 韩文财经媒体在开盘前约 08:06 已刊发同一公司消息,说明开盘前已在韩媒传播;직썰 08:31 跟进。
关键理由
- 若交易,逻辑是 ESTsoft 借 Alan agentic AI 能力首次进入公共机构内网办公 AI 市场,可复制到更多安全敏感的公营机构
- 但单个 6 个月项目金额未披露,公共机构同类 sLLM 建设订单量级相对其约 1200 亿韩元年化收入占比极低,不构成基本面重估
主要风险
- 公司近期季度仍录得经营亏损,AI 主题股属性强,小盘资金可能短线拉高后回落(见光死风险)
- 合同金额未披露,若实际金额显著大于同类公共项目假设有上修可能,但当前无证据支撑
反向证据
先例如何影响了这次决策
对本次决策的影响:先例不足叠加订单量级微小与经营亏损,支持以 RISK_REWARD 回避而非用 probe 仓追主题
方向:未检索到 ESTsoft 同类公共机构 AI 中标可锚定首发的干净先例,不据此假设涨幅
涨幅量级:缺乏同市场、同机制(小盘软件公司中标未披露金额公共 AI 项目)的对齐先例,无法校准 tp/sl 幅度
形态:无可靠先例判断见光死概率
工具执行结果 · 成功 4 · 降级 0 · 失败 1 · 取消 0
- search_processed_events 成功 stock_tools HTTP · 12 ms
{"result": "search_processed_events:agent query=`estsoft 047560 제주국제자유도시개발센터 생성형 ai | estsoft jdc generative ai contract jeju | estsoft 사내 생성형 인공지능 언어모델 사업 수주`;结构化索引覆盖完整 72 小时;旧记录兼容扫描近 0.0/72 小时、0 条本臂决策(完整),返回 0 条候选。未命中只表示这些 agent query 表达式没有召回候选,不能据此证明消息首发或不是重复;按当前事件继续正常研判。"} - stock_quote 失败 stock_tools HTTP · 8 ms
行情服务暂不可用;本次无法核验该标的价格和可执行条件。请基于其余证据继续判断并明确标注行情降级;不得仅因工具故障自动给出 AVOID 或 ENTER,也不得编造价位。 - web_search 成功 doubao_search · 1,615 ms
豆包搜索结果(原始查询:ESTsoft JDC 생성형 AI 언어모델 사업 수주 계약 금액;API SearchType:web) - web_search 成功 doubao_search · 1,611 ms
豆包搜索结果(原始查询:이스트소프트 JDC 생성형 AI 구축 용역 입찰 금액 제주국제자유도시개발센터;API SearchType:web) - web_search 成功 doubao_search · 1,868 ms
豆包搜索结果(原始查询:이스트소프트 047560 매출액 시가총액 2026 AI 사업;API SearchType:web)
原文全文
展开原文 · 2,908 字
ESTsoft (047560.KQ) is expanding its business of building artificial intelligence platforms that support internal operations at public institutions.
The company said on the 21st that it had won a contract for an in-house generative AI language model project ordered by the Jeju Free International City Development Center (JDC).
The project centers on building a generative AI environment on JDC's internal servers, separated from the external internet, to improve the efficiency and accuracy of employees' work. It is designed to allow the institution to safely process internal materials that had been difficult to use with outside AI services because of security and personal data protection rules. The project runs for six months from the contract date.
ESTsoft will build a work-support platform on JDC's internal servers that combines a small language model (sLLM) with AI agents. It will provide chatbots specialized by major business area, including development projects, duty-free operations, budgeting, safety management, contracts and human resources. The company plans to apply a language model with more than 30 billion parameters for general questions and answers, and a model with more than 120 billion parameters for complex, long-form reasoning.
About 3.8 terabytes of documents accumulated inside the institution will also be made available to the AI. Using embedding technology that converts documents into vector data along with retrieval-augmented generation (RAG), the system will search relevant documents and present supporting evidence when employees ask questions. It will also link to existing systems such as the enterprise content management (ECM) system and to external data to keep answers current.
The project goes beyond simple information retrieval to introduce AI agents that carry out actual tasks. ESTsoft will build agents that execute work through application programming interfaces (APIs) and provide an "agent builder" that allows users to design collaboration between multiple agents themselves. The aim is to help JDC expand its own use of AI in the future. The company will also establish security and monitoring systems tailored to closed-network environments, along with a large language model operations framework (LLMOps).
ESTsoft plans to draw on the technology and AI orchestration capabilities it has gained through Alan, its own agentic AI service. With the deal, the company extends its AI business from agriculture and defense into internal work support for public institutions.
"This project will serve as an example showing that generative AI can be used safely and practically even at public institutions where security is critical," said Jung Sang-won, chief executive of ESTsoft. "Based on the agentic AI capabilities we have built through Alan, we will successfully support JDC's business innovation and expand our position in the public AI market."
The company said on the 21st that it had won a contract for an in-house generative AI language model project ordered by the Jeju Free International City Development Center (JDC).
The project centers on building a generative AI environment on JDC's internal servers, separated from the external internet, to improve the efficiency and accuracy of employees' work. It is designed to allow the institution to safely process internal materials that had been difficult to use with outside AI services because of security and personal data protection rules. The project runs for six months from the contract date.
ESTsoft will build a work-support platform on JDC's internal servers that combines a small language model (sLLM) with AI agents. It will provide chatbots specialized by major business area, including development projects, duty-free operations, budgeting, safety management, contracts and human resources. The company plans to apply a language model with more than 30 billion parameters for general questions and answers, and a model with more than 120 billion parameters for complex, long-form reasoning.
About 3.8 terabytes of documents accumulated inside the institution will also be made available to the AI. Using embedding technology that converts documents into vector data along with retrieval-augmented generation (RAG), the system will search relevant documents and present supporting evidence when employees ask questions. It will also link to existing systems such as the enterprise content management (ECM) system and to external data to keep answers current.
The project goes beyond simple information retrieval to introduce AI agents that carry out actual tasks. ESTsoft will build agents that execute work through application programming interfaces (APIs) and provide an "agent builder" that allows users to design collaboration between multiple agents themselves. The aim is to help JDC expand its own use of AI in the future. The company will also establish security and monitoring systems tailored to closed-network environments, along with a large language model operations framework (LLMOps).
ESTsoft plans to draw on the technology and AI orchestration capabilities it has gained through Alan, its own agentic AI service. With the deal, the company extends its AI business from agriculture and defense into internal work support for public institutions.
"This project will serve as an example showing that generative AI can be used safely and practically even at public institutions where security is critical," said Jung Sang-won, chief executive of ESTsoft. "Based on the agentic AI capabilities we have built through Alan, we will successfully support JDC's business innovation and expand our position in the public AI market."
决策记录链 · 只追加
每次领取产生一条记录,重新决策追加一条并以 supersedes 指向被替代的记录(ADR 0014)。
05c34bfb04· 2026-10-21 08:32 · doubao / doubao-seed-evolving-latest-version · 策略包658c0c1e2ba5当前