Susana Schwartz
Fiber connectivity — a make-or-break variable in AI
The blazing speed needed for autonomous vehicles, real-time financial trading, robotic surgery, virtual assistants, chatbots, and real-time fraud detection doesn’t come easy. AI’s insatiable appetite for low latency and bandwidth is triggering a surge in interest around fiber optic connectivity, something we point out in two of our top-three stories, below.
As STL’s chief business officer of hyperscale data centers writes in a recent RCR Wireless Reader Forum, “The pressure points are no longer just compute-heavy; they’re connectivity-heavy,” with latency and real-time data movement seen as “make-or-break variables” in modern data center design.
As he points out, legacy cabling and improvised interconnects no longer cut it, which is why companies like Microsoft Azure consider fiber connectivity a major component of their data center design strategies. Other hyperscalers, like AWS, Google, Meta, and Oracle are also investing billions in fiber infrastructure.
As AI workloads get bigger and more complex, fiber connectivity will continue to be seen as crucial to AI data centers and the extreme bandwidth, low latency, and scalability requirements that are coming our way.
Susana Schwartz
Technology Editor
RCRTech
AI Infrastructure Top Stories
STL on AI-ready DCs: In this RCR Wireless Reader Forum, STL’s hyperscale data center expert Jimi Barker talks about fiber routes, density, latency, and real-time data movement as “make-or-break variables” in modern data center design.
Azure sees HCF as critical: Microsoft Azure is serious about high-capacity fiber (HCF) because it will be critical to data center connectivity, AI, 5G/6G networks, quantum, defense communications, IoT smart cities, and high-frequency trading.
OpenRAN in 2026: Open RAN isn’t yet delivering on its original promises, and this Market Pulse Report examines AI-native RAN, AI, and automation in the RIC, and the challenges of energy efficiencies and multi-vendor approaches.
AI Today: What You Need to Know
Nvidia training and inference: The new Rubin platform uses “extreme codesign” across six chips for a 10x reduction in inference token cost and 4x reduction in number of GPUs to train MoE models.
NIST and MITRE partner for AI: The National Institute of Standards and Technology (NIST) and MITRE will work to bolster U.S. AI leadership. NIST will invest $20 million to advance manufacturing and cybersecurity for AI infrastructure.
Data center rebellion is here: The Washington Post covers the groundswell of opposition from local communities against data centers, and how its reshaping the political landscape as communities anticipate impact on water and energy.
Fermi’s 5,800 energy grid: Fermi America, a startup led by former Texas governor and U.S. Energy Secretary Rick Perry and Dallas billionaire Toby Neugebauer, is developing the world’s largest private energy grid and AI campus.
New Samsung DC in SK: Samsung’s IT services unit, Samsung SDS, will build a facility in Gumi, a city in North Gyeongsang Province. The $295 million data center in eastern South Korea will be built on the site of a former manufacturing plant.
Consolidation of Chinese chip factories: In a push for semiconductor self-sufficiency, leading Chinese chipmakers like SMIC and Hua Hong are pursuing large domestic acquisitions that break the constraints of recent U.S. export controls.
