文 | Tech商业
在硅谷，大厂进行人员调整从来不缺剧本。从马斯克收购推特时的暴力清洗，到扎克伯格宣布“效率之年”时的大规模通报，科技巨头在面临利润压力时，往往选择最直接、甚至有些残酷的断腕方式。
但这一次，微软选择了一种截然不同的路径——向员工提供一份带有“自愿”标签的买断方案。
近日，微软首席人事官艾米·科尔曼（Amy Coleman）向员工发送了一份内部备忘录，宣布启动一项针对约6万名员工的自愿离职计划（Voluntary Exit Offer，简称VEO）。这一数字，大约占据了微软全球员工总数的四分之一。
备忘录的措辞极其温和：“自愿离职计划体现了我们变革方式的一个重要方面：创造更大的灵活性，并在可能的情况下，为员工提供更多塑造自己新篇章的方式。”
看似给足了体面，但剥开这层温情脉脉的公关话术，背后依然是一个庞大的商业现实：在AI基础设施上疯狂砸下近两千亿美元（2026自然年计划资本开支约1900亿美元）的微软，正在寻找一切可能的空间，来重新平衡其人力成本与算力开支。
这既是一场精准计算的财务博弈，也是大厂在AI时代进行结构性换血的缩影。
精准圈定的买断方案
这份买断方案并非面向所有微软员工，它带有极强的条件限制，精准地圈定了一个特定的“舒适区”。
根据内部文件及媒体报道，符合条件的是加拿大、印度、英国和美国参与计划的团队中，70级及以下的员工（包含公司高层领导及以下级别）。而销售部门、部分工程团队（如部分数据中心工程、量子团队）以及部分职能部门（如安全、市场营销）被明确排除在外；CoreAI、Copilot、Windows+Devices等AI与核心产品工程团队仍在参与范围内。
对于这部分符合条件的员工，微软给出了与标准遣散条款相似的补偿方案：至少60天的基础工资，在此期间员工依然保留在工资单上；根据资历和任期，大多数员工最多可获得39周的基本工资。
具体的补偿细节按内部级别划分：
内部级别64及以下的员工，每服务满半年可获得一周的基本工资补偿； 级别65至67的员工，每服务满半年可获得两周的基本工资补偿。 一旦接受这一方案，员工必须在12月初做出决定，并在年底前正式离职。微软发言人明确表示，目前暂无再次进行买断的计划。
这并非微软今年第一次尝试“体面分手”。在更早的时候，微软曾推出过一项自愿退休计划，向约8750名符合条件的员工提供了退休买断方案。据知情人士透露，最终超过30%的人选择接受。通过这种方案，微软得以在避免大规模强制裁员引发舆论危机的前提下，完成了人员的优化与精简。
算力账单与人力成本的置换
微软为何要在这个节点，花大价钱“请”这6万名员工离开？
答案藏在微软最新季度的财报，以及硅谷当下最宏大的叙事里：AI军备竞赛。
今年，微软在AI数据中心、GPU采购以及云基础设施上的资本支出（CAPEX）达到了惊人的规模。为了支持OpenAI的模型训练，以及自家Copilot等AI产品的推理需求，微软必须不断建设更庞大、更密集的算力集群。
算力是当前科技界最昂贵的资源之一。当微软将上千亿美元砸向数据中心和芯片时，华尔街的分析师们自然会拿放大镜审视它的利润率。在营收增长无法立刻跟上资本支出狂飙的情况下，CFO必须找到平衡账本的筹码。
减员，永远是硅谷大厂最快见效的降本手段之一。
与强制裁员（往往需要支付高昂的赔偿金，且面临极大的公关与团队士气风险）相比，自愿买断是一种更温和、也更具财务弹性的“人员清洗”方式。对于微软而言，给工龄较长、薪酬较高或处于非核心业务的员工一笔几十周工资的补偿，促使其主动离开，能够有效优化人力成本结构，腾出资金去供养那些真正稀缺的AI算法科学家和算力工程师。
这是一场典型的资源置换：用旧时代的人力成本，去填补新时代的算力账单。
结构性换血与岗位折叠
微软的买断计划，不仅是一家公司的孤立行为。它深刻揭示了全球科技巨头在AI时代正面临的一种结构性撕裂。
在过去的移动互联网时代，大厂通过不断扩张编制来抢占市场。但进入AI时代后，巨头们的招聘与用人逻辑发生了根本性逆转。如今的微软、谷歌、Meta，都在做同一件事：给核心AI团队开出天价薪酬包，同时对非核心业务线和初级岗位进行无情的削减。
微软的Copilot和各类AI编程工具，已经在内部极大地提升了开发效率。当AI能够快速生成代码、处理基础的数据清洗和客服咨询时，那些重复性的白领工作自然成了冗余。
因此，买断计划精准排除了部分非核心工程团队和职能部门，而AI主干力量并未被排除；被暗示可以考虑“开创新篇章”的，往往是那些从事传统业务、或者可以被AI工具轻易替代的中间层员工。
不仅是降本增效，这更是大厂在AI时代的结构性换血。它们正在将庞大的员工金字塔，削成一个哑铃型：一端是极少数的顶尖技术天才，另一端是掌握核心算力和数据的系统，中间那些只会按指令执行的“打工人”，正在被挤压出局。
铁饭碗神话的终结
微软的这份自愿离职备忘录，像是一个缩影，宣告了硅谷大厂“终身雇佣”神话的最终破产。
曾几何时，进入微软、谷歌意味着拿到了通往中产阶级安稳生活的金钥匙。体面的薪水、优厚的福利、良好的工作生活平衡，让这些巨头成为了无数名校毕业生的终极归宿。
然而，当AI革命的洪流席卷而来，连微软这样的老牌巨头，也必须在“算力的无限需求”与“人力成本的有限预算”之间做出艰难抉择。
买断方案中的那句“创造更大的灵活性”，换个角度理解，就是公司的未来不再需要这么多固定的人力配置。
微软并没有强迫任何人离开。员工依然可以选择留下，但在接下来的日子里，大概率需要面对更严苛的绩效考核、更高强度的AI工具使用要求，以及随时可能被AI替代的焦虑。而对于那些工龄较长、或者想趁此机会拿一笔补偿金转型的人来说，提前主动离开，或许是这场温和博弈中，双方都能保留最后体面的最优解。
结语
当微软把巨额的资本开支砸向AI数据中心时，它所呈现出的景象是极其割裂的：一边是科技界最宏伟的基建狂飙，一边是6万名员工正在等待那封决定他们去留的邮件。
AI革命创造着前所未有的奇迹，也伴随着前所未有的破坏。它带来了千亿级别的算力红利，却也让大厂的员工们第一次如此真切地感受到，所谓的“铁饭碗”在资本的财务模型面前，是多么的脆弱。
留给所有科技从业者的启示是残酷的：在算力吞噬人力的时代，不要试图用苦劳去换取安稳。真正能让你在硅谷屹立不倒的，不是你在哪家大厂上过班，而是你是否具备了驾驭AI、或者不可替代的底层能力。毕竟，在买断协议上签字的那一刻，资本的算盘已经打得清清楚楚。
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
In Silicon Valley, the major plant was never short of scripts for staff adjustments. From the violent purges during the acquisition of Twitter by Mask, to the large-scale announcement by Zuckerberg of the “Year of Efficiency”, the technology giants often choose the most direct and even some cruel way to break their wrists in the face of profit pressures.
But on this occasion, Microsoft chose a very different path — providing employees with a buyout with a “voluntary” label.
Recently, Amy Coleman, the Chief Personnel Officer of Microsoft, sent an internal memorandum to staff announcing the launch of a voluntary separation scheme for some 60,000 employees (Voluntary Exit Offer, known as VEO). This figure represents about a quarter of Microsoft ' s global workforce.
The memorandum is very modest: “The voluntary separation scheme reflects an important aspect of our way of changing: creating greater flexibility and, where possible, providing staff with more ways to shape their own new chapter”.
It seems decent, but the public relations jargon that rips off the warmth is still a huge business reality: Microsoft, which has razed nearly $200 billion on the infrastructure of AI (about $190 billion in planned capital expenditure for the natural year 2026), is looking for all possible space to rebalance its human costs and calculus expenditures.
This is both an accurate financial game and a microcosm of structural blood exchange in the AI era.
Precision buyout.
This buyout programme, which is not intended for all Microsoft employees, is subject to very strong conditions, precisely encircling a particular “comfort zone”.
According to internal documentation and media reports, eligible employees (including senior corporate leadership and below) in Canada, India, the United Kingdom and the United States of America participating in the plan are those in levels 70 and below. Sales, some engineering teams (e.g., partial data centre engineering, quantum teams) and some functional units (e.g., security, marketing) are explicitly excluded; the AI and core product engineering teams, such as Core AI, Copilot, Windows+Devices, are still in the range.
For this part of the eligible employees, Microsoft offers compensation schemes similar to standard severance clauses: basic wages for at least 60 days, during which the employee remains on the payroll, and basic wages for up to 39 weeks, depending on seniority and tenure.
Specific compensation details are classified by internal level:
(b) For employees in grades 65 to 67, two weeks of basic salary compensation. Once this option is accepted, the staff member must take a decision in early December and be formally separated before the end of the year. The Microsoft spokesperson made it clear that there was no plan for another buyout.
This is not Microsoft's first attempt this year to break up with dignity. Earlier, Microsoft had introduced a voluntary retirement scheme that provided retirement buyout programmes to some 875 eligible employees. According to sources, more than 30 per cent of the population eventually chose to accept it. Through such a programme, Microsoft has been able to optimize and streamline its staff while avoiding a massive crisis of public opinion caused by sanctions.
Replacement of manual bills and manpower costs
Why would Microsoft, at this node, “please” the 60,000 employees at high cost to leave?
The answer lies in Microsoft's latest quarterly financial statements and in Silicon Valley's greatest narrative: the AI arms race.
This year, Microsoft ' s capital expenditure on the AI data centre, GPU procurement and cloud infrastructure (CAPEX) reached alarming proportions. In order to support the model training of Open AI and the reasoning needs of its own AI products, such as Copilot, Microsoft must constantly build larger and more intensive clusters of computing power.
This is one of the most expensive resources available to the scientific and technological community today. When Microsoft throws hundreds of billions of dollars into data centres and chips, Wall Street analysts naturally look at its profitability with a magnifying glass. The CFO must find leverage to balance the books when the growth of revenue cannot immediately keep up with capital spending.
Loss of staff will always be one of the fastest drop-outs at Silicon Valley.
Voluntary buyouts are a more moderate and financially resilient “human cleansing” approach than strong sanctioners, who often have to pay high compensations and face high public relations and team morale risks. For Microsoft, the compensation of dozens of weeks of wages for long-serving, better-paid or non-core employees prompts them to leave on their own initiative, effectively optimizing the human cost structure and freeing up funds to support truly scarce AI algorithmic scientists and computer engineers.
This is a typical change of resources: to fill the new age's power bill with the human costs of the old age.
Structural blood exchange and job folding
The Microsoft buyout plan is not just an isolated act by a company. It is a profound illustration of the structural tore that the global technological giant is facing in the AI era.
In the past era of mobile Internet, large factories seized markets through expanding production. But when we entered the AI era, the recruitment and use logic of the giants was fundamentally reversed. Microsoft, Google and Meta are doing the same thing today: handing out a sky-priced pay package to the core AI team, with relentless cuts in non-core lines of business and junior positions.
Microsoft ' s Copilot and various AI programming tools have greatly improved development efficiency internally. When AI is able to quickly generate codes, process underlying data cleansing and client counselling, the repetitive white collars naturally become redundant.
As a result, the buyout plan accurately excludes some non-core engineering teams and functional units, while the AI backbone is not excluded; it is suggested that “new chapters” could be considered, often mid-level employees who are engaged in traditional business or who can easily be replaced by AI tools.
This is not only a reduction of efficiency gains, but also a structural blood exchange in large plants during the AI era. They are cutting large employee pyramids into a dud bell: one end is a very small number of top-tech geniuses, the other is a system with core computing power and data, and the middle is being squeezed out of the game by “hitting workers” who only follow instructions.
The End of the Iron Man's Myth
This voluntary exit memorandum by Microsoft, like a microcosm, declared the “life-long employment” myth of Silicon Valley Industries to be final bankruptcy.
Once upon a time, entering Microsoft and Google meant getting the gold key to a stable life for the middle class. Decent salaries, a generous freebie and a good work-life balance have made these giants the ultimate destination for countless graduates.
However, when the flood of the AI revolution has swept over, even old giants like Microsoft have to make difficult choices between “calculated limitless demand” and “limited budget for human costs.”
The phrase “to create greater flexibility” in the buyout programme is understood from the perspective that the future of the company does not require so much fixed manpower.
Microsoft did not force anyone to leave. Staff members still have the option to remain, but in the coming days the probability will have to face more stringent performance appraisals, more intensive AI tool use requirements and anxiety that may be replaced by AI at any time. For those who have been working longer, or who wish to take advantage of this opportunity to make a change, leaving ahead of time is perhaps the best solution to this gentle game that can last with dignity.
Concluding remarks
When Microsoft throws huge capital expenditures at the ACDC, it presents an extremely fragmented picture: On the one hand, the scientific community's greatest infrastructure boom, and on the other hand, 60,000 employees are waiting for the e-mail that they decided to leave.
The Al-Air revolution has created unprecedented miracles, accompanied by unprecedented destruction. For the first time, it has brought hundreds of billions of calculus dividends, but for the first time, the vulnerability of so-called “iron jobs” to the financial model of capital has also been so acutely felt by workers in large factories.
The message left to all science and technology practitioners is cruel: in the age of calculation that consumes human beings, do not try to trade hard labour for stability. What really stands you up in Silicon Valley is not your work at the main factory, but your ability to handle AI or an irreplaceable bottom line. After all, at the moment when the purchase agreement was signed, the balance of capital was clear.
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.