文 | Tech商业
联想集团正处在一个极其微妙且充满张力的估值十字路口。
10月中下旬，摩根士丹利与中信证券相继发布针对联想的深度研报，两者的核心判断高度一致：市场严重低估了联想ISG（基础设施方案业务）的增长弹性。大摩将联想目标价上调至49港元，中信证券则维持50港元目标价，其底层逻辑是“传统服务器及AI服务器需求保持强劲，ISG有望维持高景气度”。
资本市场的定价逻辑正在发生剧烈位移。支撑这一判断的，是ISG在上一财季交出的堪称炸裂的数字：收入同比暴增98%至85亿美元，经营利润率从去年同期的负值直接跃升至9.1%，AI服务器订单储备按季增加157%，达到惊人的540亿美元。与此同时，联想在10月推出了首批搭载NVIDIA RTX Spark N1X芯片的“YOGA Pro 15 RTX Spark超级AI PC”。一条从云端AI服务器到端侧AI PC的算力叙事，似乎正在完美成型。
但在这份光鲜亮丽的AI成绩单背后，同一份财报里的另一组数据却构成了极其冷酷的对冲。根据Gartner的数据，2026年第三季度全球PC出货量同比下降22.4%，联想出货量同步下降22.4%，以约1490万台的成绩维持全球第一，份额约26%。
这并非联想一家的困境。存储芯片和内存的疯狂涨价推高了整机成本，更重要的是，AI服务器对内存和先进封装产能的疯狂挤占，使得传统PC旺季的逻辑被彻底打乱。AI基建的优先级，正在全球供应链层面重新排序。
因此，联想当下的故事，绝不能简化为“AI服务器拯救PC厂商”的爽文。它更像是一家拥有四十余年历史的硬件巨头，在两条截然不同的增长曲线之间，艰难寻找新平衡点的过程：一条是正在加速兑现、利润丰厚的算力基建曲线；另一条是周期性承压、但依然贡献着最大现金流基盘的终端设备曲线。
ISG的利润兑现：从“沉默资产”到“利润引擎”
ISG的爆发并非天上掉馅饼，而是十年磨一剑的隐忍结果。
2014年，联想以23亿美元从谷歌手中收购了IBM x86服务器业务。这笔交易在当时的硅谷和华尔街看来，更像是接手了一项利润微薄、竞争惨烈的“夕阳资产”。此后近十年里，ISG长期处于亏损或微利状态，在联想内部，其存在感和聚光灯效应远不及IDG（智能设备业务）和SSG（方案服务业务）。
然而，正是这笔充满争议的收购，为联想沉淀了极其深厚的技术底座：液冷技术（Neptune海神液冷系统）、全球企业级客户关系、以及极具韧性的全球服务器交付体系。当生成式AI的浪潮在2025至2026年彻底引爆，全球算力需求呈指数级爆发时，这些曾经被资本市场忽视的“沉默资产”，其定价被瞬间重估。
中信证券的测算给出了一个更为具体的盈利图景：联想企业客户业务的毛利率可达20%至30%，经营利润率可达15%至17%甚至更高。这意味着，ISG的利润弹性远大于其收入弹性。上一财季9.1%的经营利润率仅仅是一个起点，如果企业AI本地化部署需求、以及自有IP存储和液冷产品的商业化继续推进，ISG的利润率中枢仍有巨大的上移空间。
但在狂飙突进中，必须用审慎的框架审视其增长质量。
AI服务器订单的客户结构中，包含了相当比例的新云（Neocloud）服务商。市场对这些新云玩家的融资可持续性和偿债能力存在深深的担忧。中信证券的辩护逻辑是：联想的新云客户以股权融资为主，订单背后通常有长期下游承购合同支持。更重要的是，联想不向客户提供内部融资，并通过严格的预付款和设备回收权来控制系统性风险。
这个风控结构在财务上是极其理智的。但它同时也意味着一个残酷的现实：ISG的收入节奏，并不完全由联想自身的交付能力决定。下游新云客户的资本开支周期、美联储的利率环境、以及大模型商业化落地的速度，会直接且迅速地传导到联想订单的转化速度上。联想吃到了新云爆发的红利，但也必须承受新云波动的代价。
PC的“以价补量”：高端化与成本挤压的赛跑
在ISG高歌猛进的同时，联想的基本盘——PC业务，正在经历一场痛苦的防守战。
出货量下滑22.6%是客观事实。但中信证券测算指出，2026年第二季度联想PC平均售价（ASP）同比提升约30%。若这一趋势延续，有望基本抵消出货量下滑对整体收入的影响。联想的策略极其清晰：放弃对出货量绝对规模的执念，转而通过产品高端化和价格传导，死守利润质量。
在这场防守战中，AI PC是绝对的核心载体。它既是联想维持PC均价、对冲零部件涨价的手段，也是其将AI能力向终端延伸、重构用户交互体验的接口。联想试图通过集成本地NPU算力，让AI PC成为个人数据的本地处理中心，从而摆脱对云端算力的绝对依赖。
然而，“以价补量”的可持续性正面临严峻考验，这取决于两个核心变量的拉扯：
其一，存储等零部件成本的上涨是否会长期持续，从而彻底挤压中低端机型的利润空间。由于AI服务器对HBM和DDR5的虹吸效应，传统PC内存价格在2026年持续走高，这使得入门级PC的生存空间被极度压缩。
其二，AI PC的换机需求能否从“尝鲜”真正跨越到“刚需”。目前，大多数消费者和企业在更换设备时依然缺乏杀手级的AI应用驱动。如果AI PC的渗透率提升速度慢于整体出货量下滑的速度，PC业务的利润缓冲垫将会被迅速变薄。
估值重估的命题：18倍PE是便宜还是昂贵？
联想当前的核心矛盾，不是“有没有AI故事”，而是“市场究竟愿意为ISG的利润质量支付多少溢价”。
大摩和中信证券的逻辑是：ISG的高景气度被严重低估，联想的估值应该向纯粹的AI基础设施公司靠拢。摩根士丹利甚至直接将联想列为本周亚洲“三大可行投资主意”之一，理由是云端及Neoclouds对通用服务器和AI服务器的强劲需求，将带动ISG顶线收入实现更强劲的增长。
但另一组数据构成了折价的理由。中信证券预计联想2027至2029财年核心归母净利润分别为47.3亿、58.2亿和71.2亿美元，对应2027财年18倍的预测市盈率。
这个估值水平极其耐人寻味。放在动辄30-40倍PE的AI基础设施公司中，它显得很便宜；但放在一家PC收入仍占大头的传统硬件制造商身上，它又算不上绝对安全。折价的本质，是市场对联想的“身份认同”依然存在巨大分歧：它是一家正在被算力需求重新定价的AI基础设施新贵，还是一家PC周期尚未走完的传统制造商？
中信证券在研报中同时提示了风险：AI服务器潜在项目的转化与交付节奏、关键零部件供应的稳定性、以及新云客户资本开支的变化，都可能影响后续业绩的兑现。
联想已经用ISG的订单和利润率证明了，算力基建的需求是真实的，它的全球交付能力也是真实的。但“真实的增长”和“可持续的估值重估”之间，还隔着一道关于客户质量、利润结构和周期位置的严苛验证。
这道验证的结果，不会由一份财报的暴涨决定，而将由未来四到六个季度里，ISG利润率的走势与PC均价韧性共同书写。联想正站在旧周期与新纪元的交汇点上，这场硬仗，才刚刚开始。
本文来自微信公众号“Tech商业”，作者：Tech商业， 经授权发布。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Man, Tech Business
Think Group is at an extremely delicate and dynamic valuation crossroads.
In mid-October, Morgan Stanley and publish, successively in-depth reviews of the Syndicate, were followed by a high degree of consistency in their core judgement: the market seriously underestimated the growth elasticity of the Syndicate ISG (Infrastructure Program Operations). The UTM wants to raise the target price to HK$49, while the CIM is maintaining the target price of HK$50. The bottom logic is that “the demand for traditional servers and AI servers remains strong and ISG is expected to maintain a high level of activity”.
The pricing logic of capital markets is shifting dramatically. This judgement is supported by the fissure figures handed over by the ISG during the previous financial season: a sharp increase of 98 per cent to $8.5 billion in revenues over the same period, a direct jump in operating profitability from negative values over the same period last year to 9.1 per cent, and a 157 per cent quarterly increase in the AIS server order reserve to a staggering $54 billion. Meanwhile, in October, the Union launched the first “YOGA Pro 15 RTX Spark SuperAPC” with NVIDIA RTX Spark N1X chips. A mathematical narrative from the cloud-end AI server to the side of AI PC seems to be perfecting.
But behind this bright and bright AI report card, another set of data in the same newspaper constitutes an extremely cold hedge. According to Gartner, in the third quarter of 2026, global PC deliveries decreased by 22.4 per cent over the same period, and the number of incoming shipments dropped simultaneously by 22.4 per cent, maintaining the world ' s top score of about 14.9 million, with a share of about 26 per cent.
This is not about family dilemmas. The crazy increase in the price of storage chips and memory increases the overall cost of the machine and, more importantly, the crazy crowding out of memory and advanced containment capacity by the AI server has completely disrupted the logic of the traditional PC boom season. The priority of AI infrastructure is being reordered at the global supply chain level.
Therefore, the current story must not be reduced to “AI Server Saving PC Vendors”. More like a hardware giant with more than 40 years of history, between two distinct growth curves, it struggles to find a new balance: one that is fast-tracking and profitable; and the other that is a terminal device curve that is cyclically under pressure but still contributes to the largest cash flow base.
ISG's profit realization: from “silent assets” to “profit engines”
The explosion of the ISG was not the result of the fall of pie, but of the intransigence of ten years of grinding a sword.
In 2014, IBM x86 server operations were purchased from Valley singers for $2.3 billion. The deal, in the then Silicon Valley and Wall Street, was more like taking over a low-profit and highly competitive “sunset asset”. For almost a decade thereafter, ISG has been in a long-term state of loss or micro-interest, and its sense of presence and the effect of the spotlight is far less than that of IDG (intelligence equipment operations) and SSG (programme services operations) within the contours.
However, it is this controversial acquisition that has sunk the deep technological base for the association: Liquid cold technology (Neptune Hydraulic Cold System), global enterprise-level customer relations, and highly resilient global server delivery systems. These “silent assets”, which have been ignored by capital markets, were re-prioritized in an instant when the wave of generating AI blew up completely between 2025 and 2026 and global computing needs erupted exponentially.
The measurement of CSI gives a more specific picture of profits: the MLF of business clients can be up to 20 to 30 per cent and the profit margin can be 15 to 17 per cent or higher. This means that the profit elasticity of ISG is much greater than its income elasticity. An operating profit margin of 9.1 per cent in the previous fiscal season was only the starting point, and if the local deployment needs of the enterprise AI, as well as the commercialization of its own IP storage and liquid cooling products, continue to advance, the profit margin hub of the ISG remains vastly mobile.
But in the run-up, the quality of its growth must be viewed in a careful framework.
The client structure of the AI server order includes a significant proportion of the new cloud (Neocloud) service providers. The market is deeply concerned about the financial sustainability and solvency of these new cloud players. The defence logic for CSIs is that the prospective new cloud customer is dominated by equity financing and that long-term downstream purchase contracts are usually supported behind the order. More importantly, it wants to refrain from providing internal financing to clients and to control systemic risks through strict advance payments and equipment recovery rights.
This wind-control structure is financially sound. But at the same time it means a cruel reality: the income rhythm of ISG is not entirely determined by the ability to think about its own delivery. The capital expenditure cycles of downstream New Cloud clients, the interest rate environment of the Federal Reserve, and the speed at which LLM commercializes will be transmitted directly and quickly to the conversion rate of associated orders. The dividends of new cloud eruptions are sought, but the cost of new cloud fluctuations must also be borne.
PC's `value-added': high-end race against cost squeeze
As the ISG goes forward, the basic board of thought — the PC business — is going through a painful defensive battle.
The 22.6 per cent drop in deliveries is an objective fact. However, the CSI measure indicates that the average price of PC sales (ASP) was about 30 per cent higher in the second quarter of 2026. If this trend continues, it is expected that the overall income impact of the decline in the volume of deliveries will be largely offset. The strategy is extremely clear: abandoning the belief in the absolute size of the volume to be delivered, and instead maintaining the quality of the profits through high-end productization and price transmission.
In this defense, AI PC is the absolute core carrier. It is both a means to maintain the average price of the PC, to raise the price of the hedge parts, and an interface to extend the AI capability to the terminal and to reconstruct the interactive user experience. The Union is trying to move away from absolute reliance on cloud computing by bringing together NPUs at cost and making AI PCs a local processing centre for personal data.
However, the sustainability of “price-added” is being severely tested, depending on the convergence of two core variables:
First, whether the increase in the cost of spare parts, such as storage, will be sustained in the long term, thus completely squeezing the profit space of the medium- and low-end machine. Due to the iris effect of the AI server on HBM and DDR5, the traditional PC memory price continued to rise in 2026, which resulted in extreme compression of the entry-level PC living space.
Second, can AI PC's replacement demand really cross from “smoking” to “frequent”. Currently, most consumers and businesses still lack a killer-level AI application driver when replacing equipment. If the penetration rate of AI PC is rising at a slower rate than that of the overall output, the profit buffer for PC operations will be rapidly thinned.
Thesis of the valuation revalued: 18 times the PE is cheap or expensive?
The central contradiction is not “the existence of an AI story”, but “the extent to which the market is willing to pay a premium for the quality of ISG profits”.
The logic of the two securities is that the high profile of ISG is severely underestimated and that the underlying valuation should be brought closer to a purely AI infrastructure company. Morgan Stanley even directly listed the association as one of the “three viable investment ideas” in Asia this week, on the grounds that the cloud and the strong demand for common servers and AI servers by Neoclouds will drive a stronger growth of ISG top-line revenue.
However, another set of data constitutes the reason for the discount. The Sino-Culture securities are expected to have a net profit of $4.73 billion, $5.82 billion and $7.12 billion, respectively, for the core return for the 2027-2029 fiscal years, corresponding to a projected market gain of 18 times that for the 2027 fiscal year.
This valuation level is extremely interesting. It appears to be cheap in an AI infrastructure company that moves 30-40 times more than PE; but it is not absolutely safe in a PC traditional hardware manufacturer, which still earns a large share. The essence of the discount is that the market still has a huge disagreement over the perceived “identity”: is it a traditional manufacturer of AI infrastructure that is being reprioritized by computing needs, or is it a PC cycle that has not yet been completed?
In its development, CSI also suggests risks: the transformation and delivery tempo of potential AI server projects, the stability of the supply of critical spare parts, and changes in capital expenditures of new cloud customers may affect the realization of subsequent performance.
It is assumed that ISG orders and profit margins have proved that the need for a calculator infrastructure is real and that its global delivery capacity is real. But the “real growth” and “sustainable valuation revaluation” are also separated by a stringent validation of customer quality, profit structure and cyclical location.
The results of this validation will not be determined by a financial boom, but will be co-written by the movement of the ISG profit margin and the average price resilience of the PC over the next four to six quarters. The Union is at the crossroads of the old cycle and the new era, and the battle has just begun.
This post is from WeChatWeChat Official Account “Tech Business”, by Tech Business, an authorized publish.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.