大华股份:扣非净利润同比增长39.75%,大模型切换AI估值

（以下内容从天风国际证券《大华股份:扣非净利润同比增长39.75%,大模型切换AI估值》研报附件原文摘录）
　　Zhejiang Dahua Technology (002236 CH) 扣非净利润同比增长39.75%，大模型切换AI估值 Switch-over to large-model AI security solutions could kickstart new valuation growth cycle TP:RMB31.80 BUY(maintain) 投资要点/Investment Thesis 投资要点/Investment Thesis 事件：公司发布了2023年半年度业绩快报。2023年上半年，公司实现营业收入1,463,354.73万元，同比增长3.88%；实现归属于上市公司股东的净利润197,600.95万元，同比增长30.01%；实现扣除非经常性损益后的归属于上市公司股东的净利润175,476.09万元，同比增长39.75%。2023年第二季度，公司实现营业收入861,791.74万元，同比增长4.60%；实现归属于上市公司股东的净利润148,065.83万元，同比增长27.17%；实现扣除非经常性损益后的归属于上市公司股东的净利润135,665.58万元，同比增长49.10%。 Q2 inflection point on downstream demand uptick; 20-30x valuation potential Financial metrics: Dahua Technology released 23H1 results: revenue came to RMB14.63bn, up 3.88% yoy, net profit to RMB1.97bn, up 30.01% yoy, while ex-nonrecurring net profit arrived at RMB1.75bn, up 39.75% yoy. Within this, Q2 revenue was RMB8.62bn, up 4.60% yoy, net profit was RMB1.48bn, up 27.17% yoy, and ex-nonrecurring net profit grew 49.10% yoy to RMB1.36bn. 点评：1）二季度开始行业逐渐回暖，业绩处于历史相对底部；2）估值切换逻辑清晰，AI逻辑下公司估值有望从20倍提升到30倍。我们认为公司的业绩和估值有望呈现良好的上升态势。 We have two observations: (1) the industry began to recover in Q2 when business was treading a historical bottom; and (2) Dahua showed a clear thematic valuation switchover to AI drivers. We expect the company’s business and valuation trends will track firmly upwards henceforth and we see its valuation rising 20x-30x on AI drivers. 下游需求复苏开始启动，二季度业绩宣示拐点到来。我们认为，二季度业绩增长超预期印证了业绩见底背后的三个逻辑：1）下游To B业务回暖。2）减少主动集成业务，优化盈利结构。3）海外需求恢复，竞争格局改善。基于此，我们认为业绩复苏有望将在三季度中得到延续。 We attribute the following three key factors to Dahua above-expectation growth and bottoming out in Q2, and expect the recovery trend to continue in Q3E. ·pick-up in B2B downstream business; ·dialing down of proactive business integration to optimize profit mix; ·recovering export demand, along with a more favorable competitive landscape. 从传统安防到AI，估值切换逻辑清晰。大华股份将发布百亿级参数的大模型，今年在部分高价值行业中试点，明年在多行业实现应用，凭借算法开发领域的深厚积累，大华股份有望在AI领域继续领跑。公司已经在大模型和多模态技术取得连续突破，并取得了数据驱动的算法方案生产流水线的显著成果，在机器视觉领域也表现出技术优势。公司面向垂直行业应用，开发行业视觉大模型，结合行业专业知识，在效果精度方面有显著优势。大模型技术能从庞大、多类型的场景数据中学习，总结出跨场景、跨业务的通用能力，学习出最普遍的特征和规律，成为具有强泛化能力的通用大模型。它能够更好地捕捉数据中的细微差异和模式，准确性更高。在一些任务中，大模型已经超过了人类的表现，具体场景可以实现连续复杂行为的识别。我们认为，大华大模型发布在即，公司的估值逻辑将从传统安防切换到AI，我们此前预计大华股份在2023年将实现35亿归母净利润，据此计算公司当下的PE仅20倍左右。在AI逻辑下，我们认为大华股份的估值有望达到30倍PE，对应市值1050亿元。 Valuation uptrend potential with switch-over from traditional to AI-enabled security We are expecting an AI catalytic impact on the rollout of a large model with tens of billions of parameters. This would first be a pilot project involving several high-value industries, before rolling out to multiple other industries next year. Dahua has amassed deep algorithm development resources, so we expect it will lead the market in AI deployment. Large model advantage: successive breakthroughs in large-model and multimodal technologies have led the company toward remarkable data-driven algorithmic solutions in its production lines, demonstrating its technological advantage in machine vision. Dahua’s business is oriented toward vertical industry applications and the development of a large-scale industry vision model synergizes industry expertise with accuracy to significant advantage. Large model technology involves combing enormous multiple-scenario data, summarizing general capabilities across different scenarios and businesses, and learning common characteristics and laws, so as to form a large model with strong generalization capabilities. This enables a more accurate capture of data nuances and patterns. In some tasks, large models have surpassed human performance, and in certain scenarios they have achieved continual and complex behavior recognition. Imminent rollout: we expect Dahua will roll out its large model soon, which would bring a switch in its valuation theme from traditional to AI security. Based on our current net profit projection of RMB3.5bn in 2023E, PE comes to only about 20x. Adding the AI catalyst, we believe Dahua’s share valuation could reach 30x PE, corresponding to market cap of RMB105bn. 降本、提精度能力将提升解决碎片化问题的效率，大模型产业化路径更加明确。许多To B业务场景下游链条较长，需要面对大量碎片化细分场景的复杂问题，以煤矿领域为例就有一千多个细分场景。大模型可以大幅缩减算法的交付数量，降低人工智能解决方案的研发成本，将有效提升人工智能面对碎片化问题时的研发效率和供给能力。公司已开发SAML技术，可以根据算法需求，自动在先验库中搜索出最优网络结构，与去年同期算法相比，在实现同等效果的前提下可以节省50%的人力，模型效果平均提升5%，模型耗时平均降低5%。随着大模型降本、提精度的能力进一步增强，我们认为，通过更有效地解决长尾碎片化问题，大模型在垂类场景中的应用将快速扩张。 Lower cost & higher accuracy add fragmentation efficacy, monetization potential Many B2B application scenarios have long value chains, especially in the downstream market, with complex problems due to the large number of fragmented market segments. Coal mining, for instance, has over 1,000 segmental scenarios. A large model could sharply reduce the number of delivery algorithms, lower R&D cost in AI solutions, and thus effectively raise R&D efficiency and AI supply capacity in handling fragmentation problems. Dahua developed SAML technology, which automatically searches for the optimal network structure in a library, based on algorithm requirements. Compared with the algorithms of a year ago, it saves 50% of manpower while achieving a similar effect: the model’s effect increases 5% on average and its time consumption is 5% less on average. As large models get better at lowering cost and raising accuracy to more effectively solve long-tail fragmentation problems, we believe large model applications in vertical market scenarios will expand rapidly. 投资建议/Investment Ideas 投资建议：公司业绩拐点或到来+估值从传统安防向AI切换的逻辑将成为公司价值增长的重要驱动力，我们预测23/24/25年公司归母净利润分别为35.29亿元/45.79亿元/61.22亿元，维持“买入”评级。 Valuation and risks We believe the key driver of Dahua’s growth stock value lies in the potential impact of its performance inflection point, combined with its thematic valuation switch-over from traditional to AI-driven security. We forecast net profit of RMB3.53bn/4.58bn/6.12bn in 2023/24/25E and maintain our BUY rating. 风险提示：产业数字化发展不及预期；市场竞争加剧；技术应用进度不及预期风险。业绩快报是初步测算结果，具体财务数据以公司披露公告为准。 Risks include: weaker industrial digitalization than expected; intensifying competition; slower-than-expected progress in technology applications. Note: the performance report is based on preliminary estimates and subject to the release of detailed financial data in the company’s disclosure announcement. Email: equity@tfisec.com TFI research report website: (pls scan the QR code) 本文件由天风国际资产管理有限公司（证监会中央编号：ASF056）编制，所载资料可能以若干假设为基础，仅供专业投资者作非商业用途及参考之用途，会因经济、市场及其他情况而随时更改而毋须另行通知。任何媒体、网站或个人未经授权不得转载、链接、转贴或以其他方式复制发表本文件及任何内容。已获授权者，在使用本文件或任何内容时必须注明稿件来源于天风国际集团，并承诺遵守相关法例及一切使用的国际惯例，不为任何非法目的或以任何非法方式使用本文件，违者将依法追究相关法律责任。本文件所引用之数据或资料可能得自第三方，天风国际集团将尽可能确认资料来源之可靠性，但天风国际集团并不对第三方所提供数据或资料之准确性负责。且天风国际集团不会就本文件所载任何资料、预测及／或意见的公平性、准确性、时限性、完整性或正确性，以及任何该等预测及／或意见所依据的基准作出任何明文或暗示的保证、陈述、担保或承诺而负责或承担任何法律责任。本文件中如有类似前瞻性陈述之内容，此等内容或陈述不得视为对任何将来表现之保证，且应注意实际情况或发展可能与该等陈述有重大落差。本文件并非及不应被视为邀约、招揽、邀请、建议买卖任何投资产品或投资决策之依据，亦不应被诠释为专业意见。阅览本文件的人士或在作出任何投资决策前，应完全了解其风险以及有关法律、赋税及会计的特点及后果，并根据个人的情况决定投资是否切合个人的投资目标，以及能否承担有关风险，必要时应寻求适当的专业意见。投资涉及风险。敬请投资者注意，证券及投资的价值可升亦可跌，过往的表现不一定可以预示日后的表现。在若干国家，传阅及分派本文件的方式可能受法律或规例所限制。获取本文件的人士须知悉及遵守该等限制。

点此查看微信公众号原文全部内容>>

大部分微信公众号研报本站已有pdf详细完整版：https://www.wkzk.com/report/（可搜索研报标题关键词或机构名称查询原报告）