内容来源：https://aibusiness.com/generative-ai/prompt-next-enterprise-ai-controlling-cost-of-scale

内容总结：

生成式AI应用：企业应从优化信息体验入手

随着企业加速布局生成式人工智能，如何在控制成本、衡量回报和实现规模化部署之间取得平衡，正成为与技术创新同等重要的课题。最新数据显示，企业在AI支出中约有四分之一被浪费，反映出当前挑战已不再仅仅是部署AI，而是如何在大规模运营中有效管理成本。

这一结论来自AI软件开发平台供应商Harness发布的《2026年AI FinOps现状报告》。该报告基于对700名FinOps及工程领域负责人的调查，发现超过半数企业尚未设立专门负责AI成本管理的岗位，导致支出追踪困难、资源浪费难以识别。此外，许多企业在处理常规任务时过于依赖昂贵的大型模型，而AI应用的普及虽使单次调用成本下降，但总体消耗量持续攀升。

这并非个案。安永本周发布的调查亦显示，企业在持续投入AI的同时，更加关注Token成本与投资回报率。施耐德电气与AMD联合发布了“AI工厂”蓝图，表明企业正从验证AI可行性转向建设支撑规模化应用的基础设施。这一趋势在微软、Alphabet和Meta等科技巨头的财报中同样明显，它们更强调部署效率和基础设施执行，而非单纯增加AI预算。

业内人士指出，企业AI发展正遵循类似云计算的历史路径：从早期的技术竞赛，逐步转向注重成本管控、回报衡量和可持续扩展。将AI视为一项运营能力而非无边界实验的组织，有望在未来竞争中占据优势。

本周其他AI要闻：

• 多数美国企业缺乏成熟AI治理框架：尽管AI治理投入增加，但美国企业中拥有成熟治理框架的比例依然偏低，引发对自主AI系统部署风险的担忧。
• 美国以安全为由限制中国机器人进口：美方以国家安全风险为由，扩大对中国类人机器人的限制措施，同时加快本土AI与机器人能力建设。
• Core Scientific与AMD达成140亿美元合作，AI算力翻倍至1.1吉瓦：该公司通过合作扩展AI基础设施，凸显算力需求持续增长。
• 英伟达牵头成立开放安全AI联盟：英伟达联合数十家企业推动开放AI安全工具开发，主张更大开放性有助于增强网络防御。
• AI思维：CIO增长指南：新指南指出，首席信息官应将AI视为组织转型而非单纯技术项目，需培育鼓励实验、学习和持续适应的文化。
• Uptime调查：AI加剧数据中心不确定性：2026年度调查显示，AI正迫使运营商重新审视电力、冷却和基础设施策略，以适应更高负载需求。

中文翻译：

由谷歌云赞助

选择您的首批生成式AI用例

要开始使用生成式AI，首先应聚焦于能够改善人类与信息交互体验的领域。

随着AI采用的不断扩大，企业发现，管理成本、衡量回报和高效扩展正变得与部署技术本身同等重要。

编者按：欢迎阅读《Prompt》，您每周获取AI格局变化动态的简报。我们以分析视角解读本周重大进展，并精选值得关注的重要报道。

一份新报告发现，每四美元AI投入中就有一美元被浪费，这凸显了企业面临的日益严峻的现实：挑战已不再仅仅是部署AI，而是控制其规模化运行的成本。随着AI支出在企业中不断铺开，管理成本正变得与采用技术本身同等重要。

这一发现来自AI软件开发平台供应商Harness发布的《2026年AI在FinOps中的应用状况》报告，该报告基于对700位FinOps和工程负责人的调查。报告显示，超过一半的组织缺乏专门的AI成本负责人，导致难以追踪支出或识别浪费。

此外，组织在日常任务中经常使用比实际需求更大、更昂贵的模型，而AI采用的持续增长推动Token消耗量不断攀升，即便每个Token的成本在下降。

该报告反映了企业AI领域的更广泛转变。过去两年，重点一直是将生成式AI部署到业务各处。如今，挑战在于确保这些部署随着使用范围的扩大而保持财务上的可持续性。

Harness的发现并非孤例。

安永本周发布的一项调查显示，组织在继续投资AI的同时，对Token成本和投资回报率的关注度显著提高。施耐德电气与AMD联合发布了AI工厂蓝图，这再次表明企业正从证明AI可行转向建设能够支撑其规模化运行的基础设施。

这一转变在超大规模云服务商的财报中同样明显，微软、Alphabet和Meta都强调部署和基础设施执行，而非单纯增加AI支出。

综合本周的动态来看，企业AI的下一个阶段将不再以公司投入多少来定义，而是以它们如何有效管理这些投资来定义。

企业AI似乎正遵循一种熟悉的模式。与此前的云计算类似，早期的技术采用竞赛正让位于对成本管理、回报衡量和可持续规模化部署的更大关注。将AI视为一种运营能力而非无边界实验的组织，未来可能占据更有利的位置。

本周其他AI新闻：

大多数美国公司缺乏成熟的AI治理框架：尽管对AI治理的投资不断增长，但相对较少的美国公司拥有成熟的治理框架，这引发了人们对企业部署更多自主AI系统的担忧。

美国联邦通信委员会以安全风险为由阻止中国人形机器人进口：美国扩大了对中国人形机器人的限制，以国家安全风险为由，与此同时华盛顿方面正加大力度强化本土AI和机器人能力。

Core Scientific通过140亿美元AMD交易将AI容量翻倍至1.1吉瓦：该公司通过与AMD达成的140亿美元交易，将其AI基础设施计划扩展至1.1吉瓦，凸显了对支撑日益增长的AI需求所需计算能力的持续投资。

英伟达推进AI开放安全联盟：英伟达与数十家科技公司共同发起开放安全AI联盟，致力于开发开放AI安全工具，认为随着AI系统变得更加自主，更大的开放性将增强网络防御能力。

AI作为一种思维模式：CIO成长指南：一份新指南指出，CIO必须将AI视为一场组织变革而非技术项目，建立鼓励实验、学习和持续适应的文化。

AI推动数据中心不确定性——Uptime 2026年调查：Uptime Institute的最新调查发现，AI正迫使运营商重新思考电力、冷却和基础设施策略，以支撑日益严苛的工作负载。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
As AI adoption expands, enterprises are discovering that managing costs, measuring returns and scaling efficiently are becoming as important as deploying the technology.
Editor’s Note: Welcome to Prompt, your weekly briefing on the shifting AI landscape. We provide an analytical look at the week’s biggest developments, paired with a curated roundup of the stories that matter.
A new report found that one in four AI dollars is wasted, underscoring a growing reality for enterprises: The challenge is no longer just deploying AI; it's controlling the cost of running it at scale. As AI spending spreads across the enterprise, managing costs is becoming as important as adopting the technology itself.
The finding comes from AI software development platform vendor Harness' 2026 State of AI in FinOps report, based on a survey of 700 FinOps and engineering leaders. According to the report, more than half of organizations lack a dedicated owner for AI costs, making it difficult to track spending or identify waste.
Organizations are also frequently using larger, more expensive models than necessary for routine tasks, while growing AI adoption is driving token consumption higher, even as per-token costs decline.
The report reflects a broader shift in enterprise AI. For the past two years, the focus has been on deploying generative AI across the business. Now the challenge is ensuring those deployments remain financially sustainable as usage expands.
The Harness findings weren't isolated.
An EY survey released this week found organizations are continuing to invest in AI while paying much closer attention to token costs and ROI. Schneider Electric and AMD unveiled a blueprint for AI factories, another sign that enterprises are shifting from proving AI works to building infrastructure that can support it at scale.
The shift was also evident in hyperscaler earnings, with Microsoft, Alphabet and Meta emphasizing deployment and infrastructure execution over simply increasing AI spending.
Together, the week's developments suggest the next phase of enterprise AI will be defined less by how much companies spend and more by how effectively they manage those investments.
Enterprise AI appears to be following a familiar pattern. Much like cloud computing before it, the early race to adopt the technology is giving way to a greater focus on managing costs, measuring returns and scaling deployments sustainably. Organizations that treat AI as an operational capability rather than an open-ended experiment may be better positioned going forward.
Also in AI News This Week:
Most US Companies Lack Mature AI Governance Frameworks: Despite growing investment in AI governance, relatively few U.S. companies have mature governance frameworks in place, raising concerns as enterprises deploy more autonomous AI systems.
FCC Blocks Chinese Humanoid Robot Imports, Citing Security Risks: The U.S. expanded restrictions on Chinese humanoid robots, citing national security risks as Washington intensifies efforts to strengthen domestic AI and robotics capabilities.
Core Scientific Doubles AI Capacity to 1.1 GW in $14B AMD Deal: The company expanded its AI infrastructure plans to 1.1 gigawatts through a $14 billion deal with AMD, highlighting continued investment in the compute capacity needed to support growing AI demand.
Nvidia Pushes Ahead With Security Alliance for AI Openness: Nvidia and dozens of technology companies launched the Open Secure AI Alliance to develop open AI security tools, arguing that greater openness will strengthen cyber defenses as AI systems become more autonomous.
AI As a Mindset: A Growth Guide for CIOs: A new guide argues that CIOs must treat AI as an organizational transformation rather than a technology project, building cultures that encourage experimentation, learning and continuous adaptation.
AI Drives Data Center Uncertainty in Uptime’s 2026 Survey: Uptime Institute's latest survey found AI is forcing operators to rethink power, cooling and infrastructure strategies to support increasingly demanding workloads.