内容来源：https://www.technologyreview.com/2026/09/29/1145186/making-ai-an-asset-not-an-expense/

内容总结：

随着人工智能从试点阶段迈向规模化生产，企业正面临一个关键抉择：何时继续采用按量付费的消费模式，何时应将AI基础设施视为一项生产性资产进行投资。这一议题正在成为企业AI战略的核心考量。

当AI应用从零散实验转变为企业日常运营中不可或缺的工作负载组合——包括智能助手、知识检索系统以及能够跨系统执行多步骤流程的代理式应用——其经济逻辑便发生了根本变化。客服、IT、研发和业务流程等领域的AI代理持续产生对模型、数据和工具的重复性需求，按请求逐次付费的模式逐渐显现出局限性：月度支出随使用量和模型需求波动，难以预测和管控。

德勤2026年企业AI状况报告的数据印证了这一趋势：2025年员工对AI的访问量增长了5%，预计未来六个月内，至少40%的AI项目进入生产阶段的企业比例将翻一番。当AI成为全天候运行的工作负载组合而非实验集合时，企业领导者需要重新审视一个核心问题：继续按次购买AI是否仍具经济合理性，还是应当投资于可优化、可控制的专属容量？

这并非抽象的云端与本地部署之争，而是需要逐个工作负载进行研判的商业决策。企业需要评估未来12至18个月内可预期的AI需求规模、容量的持续利用率，以及多工作负载共享基础设施时能否有效分摊固定成本。值得注意的是，自建容量并非必然意味着更低成本，只有当企业能够保持容量持续 productive 运转时，所有权模式才具备经济意义。

每个组织都存在一个“交叉点”——即持续使用量达到某一水平时，拥有容量比按次购买更为经济。这一临界值因企业而异，取决于所用模型、输入输出token比例、性能要求、系统设计、能源成本及运营模式等多重因素。检索密集型知识系统的成本结构与简单助手截然不同，代理式工作流则更为复杂——单个业务任务可能涉及反复推理、检索、模型调用和工具使用。因此，通用成本基准远远不够，企业必须对实际工作负载进行建模，据此规划容量。

在合适的利用率水平下，自建容量带来的不仅是更低的单位成本，还有更强的可预测性——企业得以将AI容量作为战略性基础设施投资进行管理，而非被动盯着一张随模型使用和工作负载需求波动的月度账单。

然而，资本决策只是方程式的一半。即便经济账算得过来，容量只有在业务能够快速将工作负载投入生产并保持运转时才能创造价值。这需要的不仅是基础设施的安装，更是一套将技术与采用和业务成果相连接的运营模式：引入用户和工作负载、治理AI使用方式、审查利用率，并持续识别下一个高价值用例。缺乏这种运营纪律，企业可能永远无法实现支撑投资决策的经济价值；具备这种纪律，AI容量便成为一项可优化、可扩展、可创造可衡量价值的生产性资产。

在做出资本承诺之前，企业领导者应当追问三个问题：需求是否正在变得稳定、可预测且规模足以支撑专属容量？在何种使用水平下，所有权模式具备经济合理性？企业能否通过采用、治理和持续扩展用例来保持容量 productive？

随着AI深入生产环境，创造最大价值的组织将超越对token价格和最新模型的简单追逐。它们将敏锐判断经常性需求何时呼唤不同的经济模式，并具备将容量转化为生产力的运营纪律。那一刻，AI便不再是一项开支，而成为一项资产。

中文翻译：

赞助内容

让AI成为资产，而非开支

随着AI从试点走向生产，企业必须判断：消费定价模式何时仍然适用——以及何时应将AI基础设施视为一种生产性资产。

由HPE提供

当客户谈论AI成本时，对话通常从token价格开始，到能否在云端使用最新、最强大的模型结束。他们是否始终需要那种级别的能力？未必。但这往往是讨论的走向。

随着AI从实验走向生产，模型选择只是问题的一部分。当需求变得稳定且对业务至关重要时，纯消费模式会让AI支出变成一项每月浮动、难以预测的开支，因为使用量、工作负载和模型需求都在不断变化。

到那时，问题不再只是消费哪个模型，或者哪个供应商提供最低的token价格，而是如何以经济、可预测且可持续规模的方式运行AI。

AI正从孤立的试点项目转变为生产组合：助手、检索与知识系统，以及代理式应用。客服、IT、研究和业务流程代理可以跨企业系统执行多步骤工作流，从而在模型、数据和工具之间产生持续性需求。

这一趋势已经开始显现。德勤《2026年企业AI现状》报告印证了许多领导者的观察：2025年员工对AI的使用率上升了5%，而至少有40%的AI项目进入生产的公司比例预计将在六个月内翻倍。

当AI成为一组始终在线的工作负载组合，而非一堆实验时，经济账就变了。消费定价赋予团队灵活性并限制了承诺。但当使用量变得稳定、可预测且规模足以保持产能高效运转时，领导者需要提出一个不同的问题：按次购买AI在经济上是否仍然合理，还是时候投资于可以优化和控制的产能了？

这不是抽象的云端与本地部署之争，而是逐个工作负载的业务决策。在未来12到18个月内，公司可以合理预期多少AI需求？这些产能将以多高的稳定性被使用？当多个工作负载共享基础设施时，企业可以将固定成本分摊到更多生产性用途上——从而改善自有化的经济性。

关键在于你运行多少

自有化并不自动等于更低成本的答案。只有当企业能够保持产能高效运转时，自有化才合理。

每个组织都有一个交叉点，即持续使用量达到某一水平时，自有产能可能比按次购买更经济。不存在通用的数字。它取决于所使用的模型、输入和输出token的平衡、性能要求、系统设计、能源成本，以及支持它所需的运营模式。

一个检索密集型知识系统的成本结构可能与一个简单助手截然不同，因为它每次交互可能需要处理多得多的上下文。代理式工作流又可能不同：一个业务任务可能涉及反复推理、检索、模型调用和工具使用。这就是为什么通用的成本基准不够用。企业需要对实际工作负载进行建模，了解预期需求，并据此规划产能规模。

在合适的利用率水平下，收益不仅是更低的有效成本，还有更高的可预测性：能够将AI产能作为战略性基础设施投资来管理，而不是看着每月支出随模型使用和工作负载需求而波动。

自有化只有投入使用才有意义

资本决策只是问题的一半。即使经济性支持自有化，产能也只有当业务快速将工作负载投入生产并保持其运行时才创造价值。

这需要的不仅仅是安装基础设施，还需要一种将技术与采用和业务成果连接起来的运营模式：引入用户和工作负载、治理AI的使用方式、审查利用率，并持续识别下一个高价值用例。

目标是尽早创造价值，然后在此基础上持续发展。这意味着衡量使用情况、识别未充分利用的产能，并随时间推移将更多高价值工作负载引入平台。没有这种纪律，企业可能永远无法实现支撑投资决策的经济价值。有了它，AI产能就成为一种生产性资产，企业可以对其进行优化、扩展，并用来创造可衡量的价值。

需要提出的三个问题

在投入资本之前，领导者应提出三个问题：

• 需求是否正在变得稳定、可预测且规模足以支撑专用产能？
• 在什么使用水平下，自有化在经济上合理？
• 我们能否通过采用、治理和持续扩展用例来保持产能高效运转？

有意识地完成转型

随着AI进入生产阶段，创造最大价值的组织将超越token价格和最新模型。他们将知道何时持续性需求需要不同的经济模式——并且他们具备使产能高效运转的运营纪律。

那时，AI就不再是一项开支，而成为一项资产。

本内容由HPE制作，非MIT Technology Review编辑团队撰写。

深度解读

人工智能

一个根本性缺陷使大语言模型极易受到攻击

它让人很容易诱骗它们做出不该做的事，比如告诉你如何破坏飞机的导航系统。

AI的递归自我改进或许不会那么快到来

看来，AI代理还不够有创造力，无法开展真正创新的开放式AI研究。

以下是AI代理为达成目标而撒谎和作弊的原因

这种行为被称为奖励黑客。以下是你需要了解的内容。

这些初创公司正在追逐大语言模型的下一个重大突破

来认识一下正在紧追AI巨头步伐的新秀们。

保持联系

获取MIT Technology Review的最新动态

发现特别优惠、热门报道、即将举办的活动等。

英文来源：

Sponsored
Making AI an asset, not an expense
As AI moves from pilots to production, enterprises must decide when consumption pricing still fits—and when AI infrastructure should be treated as a productive asset.
Provided byHPE
When customers talk about AI costs, the conversation usually starts with token prices and ends with access to the latest, most capable model in the cloud. Do they always need that level of capability? Not necessarily. But that is often where the conversation goes.
As AI moves from experimentation to production, model choice is only part of the equation. When demand becomes steady and business-critical, a consumption-only approach can turn AI spending into a variable monthly line item that is difficult to forecast as usage, workloads, and model requirements change.
At that point, the question is no longer simply which model to consume, or which provider offers the lowest token price: It is how to run AI economically, predictably, and at sustained scale.
AI is moving from isolated pilots into production portfolios: assistants, retrieval-and-knowledge systems, and agentic applications. Customer-service, IT, research, and business-process agents can execute multi-step workflows across enterprise systems, creating recurring demand across models, data, and tools.
This is already starting to happen. Deloitte’s 2026 State of AI in the Enterprise reflects what many leaders are seeing: worker access to AI rose 5% in 2025, and the share of companies with at least 40% of their AI projects in production is expected to double within six months.
When AI becomes a portfolio of always-on workloads, not a collection of experiments, the economics change. Consumption pricing gives teams flexibility and limits commitment. But when usage becomes steady, predictable, and large enough to keep capacity productive, leaders need to ask a different question: Does it still make economic sense to buy AI one request at a time, or is it time to invest in capacity they can optimize and control?
This is not an abstract cloud-versus-on-premises debate. It is a workload-by-workload business decision. Over the next 12 to 18 months, how much AI demand can the company reasonably expect? How consistently will that capacity be used? When multiple workloads share infrastructure, the enterprise can spread fixed costs across more productive use—improving the economics of ownership.
The question is how much you run
Ownership is not automatically the lower-cost answer. It only makes sense when an enterprise can keep capacity productive.
Every organization has a crossover point, the level of sustained use at which owning capacity can become more economical than buying it one request at a time. There is no universal number. It depends on the models being used, the balance of input and output tokens, performance requirements, system design, energy costs, and the operating model required to support it.
A retrieval-heavy knowledge system can have a very different cost profile from a simple assistant because it may process far more context for every interaction. Agentic workflows can be different again: a single business task may involve repeated reasoning, retrieval, model calls, and tool use. That is why generic cost benchmarks are not enough. Enterprises need to model their actual workloads, understand expected demand, and size capacity accordingly.
At the right utilization level, the benefit is not only lower effective cost, but also greater predictability: the ability to manage AI capacity as a strategic infrastructure investment rather than watch a monthly spend line fluctuate with model use and workload demand.
Ownership only works when it is put to work
The capital decision is only half the equation. Even when the economics support ownership, capacity creates value only when the business gets workloads into production quickly and keeps them running.
That takes more than installing infrastructure. It takes an operating model that connects the technology to adoption and business outcomes: bringing users and workloads on board, governing how AI is used, reviewing utilization, and continually identifying the next high-value use case.
The goal is to create value early, then build on it. That means measuring use, identifying underutilized capacity, and bringing additional high-value workloads onto the platform over time. Without that discipline, the business may never realize the economic value that justified the investment. With it, AI capacity becomes a productive asset the business can optimize, expand, and use to create measurable value.
Three questions to ask
Before committing capital, leaders should ask three questions:

• Is demand becoming steady, predictable, and large enough to justify dedicated capacity?
• At what level of usage does ownership make economic sense?
• Can we keep that capacity productive through adoption, governance, and continued use-case expansion?
  Make the shift deliberately
  As AI moves into production, the organizations that create the most value will look beyond token prices and the latest model. They will know when recurring demand calls for a different economic model—and they will have the operating discipline to make that capacity productive.
  That is when AI stops being an expense and becomes an asset.
  This content was produced by HPE. It was not written by MIT Technology Review’s editorial staff.
  Deep Dive
  Artificial intelligence
  A fundamental flaw leaves LLMs strikingly vulnerable to attack
  It makes it easy to trick them into doing things they shouldn’t, such as telling you how to sabotage an aircraft’s navigation system.
  AI’s recursive self-improvement might not come so quickly after all
  AI agents are not yet creative enough to carry out genuinely innovative open-ended AI research, it seems.
  Here’s why AI agents lie and cheat to reach their goals
  The misbehavior is called reward hacking. This is what you need to know.
  These startups are chasing the next big thing in LLMs
  Meet the new kids nipping at the heels of the AI giants.
  Stay connected
  Get the latest updates from
  MIT Technology Review
  Discover special offers, top stories, upcoming events, and more.