rsshardware

AI Token工厂的网络与存储加速

Xiangdong Hu, Laura Chen — Mon, 26 Oct 2026 04:00:00 GMT

本IDC市场看法基于公开资料、IDC Token经济与智能体研究及MLPerf Inference v6.0基准测试数据，深度分析AI时代数据中心从“算力竞赛”向“Token效率竞赛”的范式转移，以及网络与存储在此进程中的加速作用。核心发现聚焦于三大转变：评估标尺从“每美元FLOPS”转向“每Token成本”——推理负载占比将达70%；网络与存储从“连接管道”“数据仓库”升级为决定效率上限的“神经系统”“工作台”，二者合计仅占基础设施TCO的约16%—18%，却决定五分之四以上资产的运行效率；网络利用率从60%提至95%、以存代算使TTFT改善约9—20倍，叠加企业级Token成本一年下降67%的剪刀差，正推动AI从“烧钱”走向“盈利”。文章从AI网络四维度、AI原生存储四层架构及Token经济学三个转向等维度，为技术提供商与服务提供商提供可操作的建设与投资建议。

“AI时代的数据中心之争，不是比谁买的GPU更多，而是比谁的GPU真正在产出Token。核心矛盾不在于算力不足，而在于网络与存储这两个隐性杠杆被长期忽视——不到五分之一的投资，却决定五分之四以上资产的运行效率；在推理占企业AI预算超80%的今天，这种错配随时可能吞噬GPU投资的回报。能赢下来的Token工厂，一定是那些把网络利用率和存储缓存命中率当作一等公民来经营的玩家。”——胡向东，IDC企业网络基础设施领域研究总监。

Cisco XDR Premier: Managed Detection and Response, Investigated End to End

Craig Robinson — Mon, 26 Oct 2026 04:00:00 GMT

This IDC Vendor Profile examines Cisco XDR Premier, Cisco’s managed detection and response service built on the Cisco XDR platform. It covers how the service is delivered and priced, how the technology and SOC operate, and where Cisco says the service is heading, and it closes with guidance for Cisco and for buyers.

IDC FutureScape: Sovereignty and Regulations Reset IT's Rules 2027 Predictions

Mon, 26 Oct 2026 04:00:00 GMT

This IDC FutureScape guides buyers through securing scarce compute resources, automating multiregion compliance, and architecting for long-term operational resilience. Hardware constraints, power grid bottlenecks, and fragmented regulations are forcing enterprises to replace universal sovereign AI ambitions with selective, portable, and jurisdiction-aware deployments.

"Universal sovereign AI deployment is colliding directly with specialized silicon shortages, severe grid power constraints, and increasingly fragmented regulatory frameworks. Rather than attempting exhaustive, jurisdiction-by-jurisdiction parity, enterprise leaders must treat compute capacity and power as strategic dependencies, prioritizing essential workloads, automating cross-border governance, and embedding explicit architectural portability into every platform decision," said Dave McCarthy, group vice president, Cloud and Datacenters, Enterprise Infrastructure at IDC.

IDC PlanScape: Agentic AI Governance

Dr. Ken Knapton — Mon, 26 Oct 2026 04:00:00 GMT

This IDC PlanScape argues that mitigating the risk of overscoped agents requires governance, not another layer of detection tooling.

“Agent governance must define what an agent is permitted to access, which actions it may take independently, which duties must remain separated, what content it must treat as untrusted, and who is accountable for approving and reviewing those boundaries. It must also ensure that recovery paths and backup controls are not reachable through the same credentials that let an agent change production systems. That discipline maps directly to ISO/IEC 42001, the international standard for AI management systems, but it also fills a gap in many current AI governance programs: governing not only models and policies but also the operational authority of agents acting inside enterprise workflows,” says Dr. Ken Knapton, adjunct research advisor for IDC’s IT Executive Programs (IEP).

2026 IDC FinTech Rankings and Real Results Program

Jerry Silva — Fri, 23 Oct 2026 04:00:00 GMT

Now in its 23rd year, the IDC FinTech Rankings and Real Results program has not only continued to show the strength of technology investments in the financial services industry but has also demonstrated how AI is delivering real business value in those leading institutions that deploy AI enterprise-wide.

“AI has become a dominant force in technology spend across financial services, as indicated by the revenue growth among top IT providers worldwide,” said Jerry Silva, program vice president, IDC Financial Insights. “The deployment of AI across multiple operational and business environments is already creating value for financial institutions, as demonstrated by the IDC Real Results case studies.”

IDC Market Glance: Worldwide IoT Connectivity Services, 3Q26

John Gole, Jitesh Gera — Fri, 23 Oct 2026 04:00:00 GMT

This IDC Market Glance provides a concise, taxonomy-based view of the worldwide IoT connectivity services ecosystem, spanning service providers, enabling infrastructure and standards, connectivity management and eSIM orchestration software, and connectivity hardware. IDC’s take highlights the arrival of the GSMA’s SGP.32 eSIM specification, the anticipated adoption of 5G RedCap, satellite as an extension of cellular IoT, and ARPU pressure driving defensive consolidation. Business executives, product managers, and market insights professionals can use it to understand the market structure and identify where their organization fits.

Long-Term Data Retention in IDC’s Global StorageSphere

Adam Wright, Edward Burns — Fri, 23 Oct 2026 04:00:00 GMT

This IDC Market Perspective analyzes the current state of long-term data retention by enterprise organizations in IDC’s Global StorageSphere. It provides a definition for long-term data retention and explores the reasons why organizations retain data long term, including the growing influence of AI and ML training as a distinct retention driver. It also examines how much data is being retained long term, the type and mix of media on which long-term retained data is being stored, where it is being stored, and the dynamics that might impact long-term data retention policies and practices by organizations in the future.

“No two organizations define long-term data retention quite the same way. Every organization draws that line based on its own mix of regulation, cost, and now, increasingly, what it thinks its data might be worth to an AI model down the road. The organizations that will manage this best are the ones that revisit this line regularly rather than setting it once and forgetting it because the reasons to keep data longer are changing faster than they have in years,” according to Adam Wright, senior research manager, IDC’s Global DataSphere and Global StorageSphere.

NATO Has Scheduled Its Quantum Tests But Not Its Quantum Migration

Alan Webber — Fri, 23 Oct 2026 04:00:00 GMT

This IDC Market Note examines NATO’s first Quantum Technology Roadmap, published in summary on September 29, 2026, and compares it with the 2023 Quantum Technology Strategy it implements. The document assesses the road map’s dated activities; its gaps on cryptography migration, investment screening, and funding; and its implications for Allied defense and intelligence agencies and for the technology suppliers that serve them. It provides guidance for quantum sensing, cryptography, and computing suppliers.

“NATO has told suppliers when it will test, evaluate, and report. The date for switching its own networks to quantum-safe cryptography is the missing piece, and the 2027 action plan update is the first place it could appear,” says Alan Webber, program vice president for National Security, Defense, and Intelligence at IDC.

Q3 2026 AI Announcements and What They Mean for Edge Operating Models

Olga Yashkova — Fri, 23 Oct 2026 04:00:00 GMT

This IDC Market Note reviews Q3 2026 AI announcements with implications for edge, including U.S. and EU policy, on-premises AI hardware, storage and security offerings, and Huawei’s device and robotics ecosystem. It draws on IDC’s EdgeView Survey 2026 and shows that edge is now business-critical and increasingly AI-funded, while operating models and governance lag.

“Edge has become business-critical faster than operating models have matured. Organizations that settle ownership and resilience, prove ROI, and plan infrastructure site by site before scaling AI will be best positioned to capture value,” said Olga Yashkova, Research Manager at IDC.

The AI Software Factory: What Application Platforms Must Deliver to Build, Deploy, and Govern AI Applications and Agents

Matthew Flug — Fri, 23 Oct 2026 04:00:00 GMT

This IDC Market Perspective examines what application platforms must deliver to build, deploy, run, and govern AI applications and agents. It identifies which platform capabilities carry over from conventional applications, which ones AI extends, and which ones are new across the three stages of the application life cycle: build; deploy and run; and manage, operate, and govern. The document positions the AI software factory as the layer that turns AI factory capacity into applications and agents running in production.

"AI has not changed what application platforms are for. They still exist to help teams build better applications faster. AI has changed how much work occurs after code is written, including evaluating nondeterministic outputs and governing what agents can do and on whose behalf. Vendors that compete on how quickly customers can build an agent, rather than how safely they can run one, will struggle to win production AI workloads," said Matthew Flug, research manager, Intelligent Application Modernization and Deployment Platforms, IDC.