The week of Oct. 12–16: What happened, what matters, what's next

6 Min Read

Technology companies and regulators made decisions this week that could shape the cost, availability and control of enterprise infrastructure. Microsoft is investing in its own AI chips, while a shortage of semiconductor manufacturing equipment could constrain how quickly the industry expands capacity. Meanwhile, U.S. energy regulators weighed in on how data centers connect to the power grid, and China proposed new restrictions on technology exports.

These developments affect different parts of the enterprise technology landscape, from infrastructure investment to international operations. Together, they offer a look at the commercial and regulatory forces that CIOs must account for when planning technology strategy.

Here are four key developments worth watching this week.

News Highlights Oct. 12–16

Microsoft unveils next-generation AI chip as it expands custom silicon

What happened: On Oct. 12, Microsoft announced Maia 300, the latest generation of its custom AI accelerator. The company plans to begin deploying the chips at multi-gigawatt scale in 2027. The company expects the hardware to deliver 40% more tokens per megawatt than leading GPU-based systems. Microsoft is developing the chips to support AI inference across its products and services, including Copilot, Teams and GitHub, as well as workloads running on Azure.
Why it matters: Custom silicon gives Microsoft greater control over the performance, power consumption and cost of running AI services. It also creates another way for cloud providers to differentiate their infrastructure as demand for AI computing grows. For CIOs, the practical implications will depend on how these investments translate into cloud pricing, service performance and workload options. Microsoft's efficiency claims are company estimates, but they signal how aggressively infrastructure providers are trying to improve the economics of AI at scale.

Related:The week of Oct. 5–9: What happened, what matters, what's next

ASML warns semiconductor equipment supply will lag demand

What happened: ASML, the Dutch manufacturer of advanced chipmaking equipment, said Oct. 15 that its supply would fall short of expected demand in 2027 and 2028. The company plans to increase production capacity by about 30% over that period, and expand output of its extreme ultraviolet lithography machines, which are used to manufacture advanced chips. Even with those increases, ASML expects demand to outpace its ability to deliver equipment. ASML's semiconductor manufacturing customers, including Taiwan Semiconductor Manufacturing Co., SK Hynix, Samsung and Micron, are competing for capacity as demand for AI processors and memory grows.
Why it matters: Expanding chip production requires specialized manufacturing equipment, so additional investment by semiconductor companies cannot immediately translate into more chips becoming available. ASML's outlook points to a supply constraint that could persist even as manufacturers invest to meet demand. For CIOs, that means procurement forecasts and infrastructure roadmaps may need to account for extended lead times and the possibility that supply remains tight beyond the next budget cycle.

Related:Nonprofit cost-cutting exposes outdated data infrastructure

FERC rejects key parts of PJM's plan for data center power demand

What happened: The Federal Energy Regulatory Commission on Oct. 9 largely rejected PJM Interconnection's proposed framework for connecting data centers and other large electricity users to its grid. The commission rejected several provisions, including proposed interim service and requirements for some large loads to bring new generation capacity. It did, however, approve PJM's request to develop a database tracking data centers across its territory, including their locations, power sources and expected growth in electricity demand. PJM serves a large area of the eastern and Midwestern U.S., where data center development is driving rapid growth in power demand.
Why it matters: The decision leaves major questions unresolved about how quickly new data centers can connect to the grid and how the costs of meeting their electricity needs will be allocated. For organizations planning large computing deployments, power availability and connection timelines can become constraints on expansion even when financing and hardware are in place. CIOs working on infrastructure plans will need to coordinate with facilities and energy teams early, particularly where projects depend on new data center capacity in regions experiencing rapid demand growth.

Related:The week of Sept. 21-25: What happened, what matters, what's next

China considers export controls covering advanced AI technologies

What happened: China proposed new export restrictions Oct. 16 covering 40 technologies, including 23 related to AI. The draft rules would restrict or prohibit the overseas transfer of certain advanced AI models, AI agents used in sensitive settings, chip-design technologies and other capabilities. The proposal reflects Beijing's effort to protect technologies it considers strategically important amid intensifying competition with the United States. The measures are proposed rather than fully implemented, and their eventual scope will depend on how the rules are finalized.
Why it matters: The proposal could introduce additional uncertainty for companies that develop, source or deploy technology across national borders. Depending on the final rules, restrictions could affect technology licensing, research partnerships, product development and the movement of technical expertise. CIOs with global operations or suppliers in China should monitor the consultation and assess where their technology roadmaps depend on cross-border access to advanced tools or intellectual property. The immediate task is to identify potential exposure before the proposal becomes a binding constraint.

From InformationWeek: Worth the read

You'll find a wealth of insight and analysis on the InformationWeek site, but here are a few curated picks from the week that should be at the top of your reading list.

Airtable's post-acquisition layoffs put vendor risk in focus
This week's Big Tech Story looks at how enterprises can — and should — assess new ownership at their vendor partners, particularly following early signals like layoffs.

T-Mobile CTO on injecting AI into network resiliency
Senior editor Kelsey Ziser speaks with T-Mobile's John Saw about leaning on AI, network slicing and backup satellites to speed disaster recovery and overcome CIOs' network resiliency concerns of coverage, capacity and congestion.

InformationWeek Podcast: Fighting bad actors with access to similar AI
In this latest episode, host and senior editor Joao-Pierre Ruth is joined by Phison Electronics' CTO to discuss how AI can make the difference in cybersecurity when both parties have access.

SPIFFE can identify your AI agents. But can it control them?
Contributing writer Pam Baker looks at the uses and limitations of the Secure Production Identity Framework for Everyone for identifying and controlling AI agents and explores the additional controls CIOs should be implementing.

Coming Up: Dates to watch

Oct. 19–22: Gartner IT Symposium/Xpo brings CIOs to Orlando to discuss technology strategy, cybersecurity and business transformation. InformationWeek's Ziser will be there covering the event.

Oct. 20–22: NVIDIA GTC Berlin explores AI infrastructure, accelerated computing and enterprise AI deployments, with CEO Jensen Huang's keynote on Oct. 21 likely to bring announcements relevant to CIOs' AI roadmaps.

Oct. 21: Meta, IBM and SAP report quarterly results, offering fresh signals on AI infrastructure investment, enterprise software demand and technology spending.

From the economics of AI infrastructure to the realities of power supply and global technology policy, this week's developments show how external forces continue to shape enterprise IT decisions. Which of these issues is having the biggest impact on your organization? Let us know at [email protected].

About the Author