IBM如何成为AI宠儿 How IBM became an AI darling

THROUGHOUT ITS 115-year life IBM has shown itself to be a master of reinvention. In the mid-1990s the mainframe pioneer rescued itself from collapse by shifting its focus to the booming business of IT services. A decade later it sold its struggling PC division to China’s Lenovo.
在其115年的历史中,IBM已证明自己是重塑大师。20世纪90年代中期,这家大型机先驱通过将重心转向蓬勃发展的IT服务业务,从崩溃边缘自救。十年后,它将其陷入困境的PC部门卖给了中国的联想。
Over the past half decade or so “Big Blue” has been through another striking transformation. During the 2010s its business was disrupted by the rise of cloud computing, which undermined not only sales of mainframes but also the work of servicing them at a time when low-cost outsourcers from India were pinching share. Revenues and margins shrank, and investors once again lost interest.
在过去五年左右的时间里,“蓝色巨人”经历了另一场惊人的转型。在2010年代,其业务受到云计算兴起的冲击,云计算不仅削弱了大型机的销售,也削弱了维护大型机的工作,与此同时,来自印度的低成本外包商正在抢夺市场份额。收入和利润率萎缩,投资者再次失去兴趣。
That has all changed in the past three years, during which IBM’s share price has more than doubled. As a multiple of net profits, it is now valued similarly to Microsoft and other software champions. On January 28th it reported that its revenue and net profit rose by 8% and 14% in 2025—a sharp reversal from its years of stagnation. How did IBM pull it off?
这一切在过去三年里都改变了,期间IBM的股价翻了一番多。以净利润倍数计算,其估值现在与微软和其他软件巨头相当。1月28日,它报告称2025年其收入和净利润分别增长了8%和14%——与其多年的停滞形成鲜明对比。IBM是如何做到的?
The strategy began with the acquisition in 2019 of Red Hat, a platform that, among other things, helps companies manage their workloads across data centres. Rather than trying to compete with Amazon, Google and Microsoft in the so-called public cloud, IBM created a layer between that makes it easier to mix and match among the hyperscalers while continuing to use on-premise mainframes or dedicated private clouds (including those run by IBM) for sensitive tasks. Deals in 2024 and 2025 to buy HashiCorp and Confluent, two more software firms, have solidified IBM’s role as an orchestrator of hybrid clouds.
这一战略始于2019年收购Red Hat,这是一个帮助公司管理跨数据中心工作负载的平台。IBM没有试图在所谓的公有云领域与亚马逊、谷歌和微软竞争,而是创建了一个中间层,使得在这些超大规模服务商之间混合搭配变得更容易,同时继续使用本地大型机或专用私有云(包括IBM运营的私有云)处理敏感任务。2024年和2025年收购HashiCorp和Confluent这两家软件公司的交易,巩固了IBM作为混合云协调者的地位。
IBM has also created a space for itself in AI. The company has long dabbled in the technology—including using it to beat Gary Kasparov, the world chess champion, in 1997—but missed the latest wave of large language models. Rather than trying to beat OpenAI and other model-makers at their own game, it has released a series of small language models, under the name Granite, which are tailored to business applications and require less computing power. These and other open-weight models, which make their numerical parameters freely available, are accessible through its watsonx platform, which enterprises can use to build AI agents trained on their own data.
IBM还在AI领域为自己开辟了空间。该公司长期涉足该技术——包括在1997年利用它击败世界国际象棋冠军加里·卡斯帕罗夫(Gary Kasparov)——但错过了最新一波大语言模型浪潮。它没有试图在OpenAI和其他模型制造商擅长的领域击败它们,而是以Granite为名发布了一系列小型语言模型,这些模型专为商业应用量身定制,所需的计算能力较少。这些模型以及其他开放权重模型(免费提供数值参数),可以通过其watsonx平台访问,企业可以利用该平台构建基于自身数据训练的AI代理。
IBM’s growing strength in AI has been helped by another big strategic shift over the past few years—the refashioning of its services arm. In 2021 the company spun off its struggling outsourcing business, now called Kyndryl, which at the time accounted for about a quarter of its workforce. That left it with a smaller consulting division focused on technical expertise, which has come in handy as clients grapple with AI. IBM has booked over $10bn-worth of consulting contracts related to generative AI since the middle of 2023. It has also been using the technology to digitise its own consultants’ work, a move the division’s boss has described as moving to “service as a software”.
IBM在AI方面日益增强的实力还得益于过去几年的另一大战略转变——对其服务部门的重塑。2021年,该公司剥离了其陷入困境的外包业务(现更名为Kyndryl),该业务当时约占其劳动力的四分之一。这给它留下了一个规模较小、专注于技术专长的咨询部门,随着客户努力应对AI,该部门派上了用场。自2023年中期以来,IBM已签署了价值超过100亿美元的生成式AI相关咨询合同。它还利用该技术将其自身顾问的工作数字化,该部门老板将此举描述为转向“服务即软件”。
Meanwhile, IBM continues to innovate in its original metier of hardware. It is still by far the world leader in mainframes. The z17, released last year, has been a hit. It offers access to IBM’s new Spyre chip, designed for running AI models. Then there is IBM’s work on quantum computers, where it is at the leading edge. McKinsey, a consultancy, reckons the market for the technology could reach nearly $100bn by 2035. IBM thinks it can capture about 20% of the business by selling machines and renting out capacity. It hopes to deliver Starling, a “fault-tolerant” quantum computer that can spot and correct the errors the technology is prone to, by 2029.
与此同时,IBM继续在其原本擅长的硬件领域进行创新。它目前仍是迄今为止全球大型机的领导者。去年发布的z17大获成功。它提供了对IBM设计用于运行AI模型的新Spyre芯片的访问。此外还有IBM在量子计算机方面的工作,它处于领先地位。咨询公司麦肯锡估计,到2035年该技术市场可能达到近1000亿美元。IBM认为通过销售机器和租赁算力,它可以占据约20%的业务份额。它希望在2029年交付Starling,这是一台“容错”量子计算机,可以发现并纠正该技术容易出现的错误。
Lately investors in legacy technology companies have become increasingly jittery over the prospect of disruption from AI. In the past three months software businesses in America’s S&P 500 index have shed a seventh of their value. Shares in Accenture, the world’s biggest provider of IT services, are down by a quarter over the past year amid fears that bots could soon replace much of what the company does. But IBM, with its unique portfolio of businesses, looks well positioned to make it through the upheaval.
最近,传统科技公司的投资者对AI可能带来的颠覆越来越感到不安。在过去三个月里,美国标普500指数中的软件企业市值缩水了七分之一。全球最大的IT服务提供商埃森哲(Accenture)的股价在过去一年下跌了四分之一,因为人们担心机器人可能很快取代该公司的大部分工作。但凭借其独特的业务组合,IBM似乎处于有利地位,能够安然度过这场剧变。