AFCOM’s “State of the Data Center” report is in its 10th edition, and made available during Data Center World in Washington, D.C. In the report, organizations report an average facility size approaching 38 MW, up from 32 MW last year. Additionally, rack density was found to have climbed to 27 kW per rack, up from 16 last year. According to AFCOM, nearly 70% of respondents to the survey indicate they are most commonly increasing density through liquid cooling and airflow optimization.
In its findings, AFCOM cites the S&P Global list, which summarizes where the industry is globally with data centers, with 13,500 facilities worldwide, operated by 2,700+ providers, across more than 130 countries.
RCRTech will delve deeper into its findings in upcoming days.

Susana Schwartz
Technology Editor
RCRTech
AI Infrastructure Top Stories
Marvell momentum continues: Following its $2B deal with Nvidia, Marvell is getting interest from Google, which wants to add Marvell as another major custom chip partner to design specialized hardware that complements its current TPUs.
RCRTV Pulse episode: Inference economics pulls edge computing back into play, says AvidThink Founder Roy Chua, who says companies are carrying placement logic to AI, and more complicated operations getting passed up to the edge.
FutureNet AI takeaways: At FutureNet World in London, BT, Orange, Vodafone, and Telefonica talked of incremental steps with AI, and the reasons sovereign cloud and AI without sovereign networks “doesn’t make a lot of sense.”
AI Today: What You Need to Know
Amazon – Anthropic partnership: Amazon will invest $5 billion now, and up to $20 billion more in the future, for Anthropic, which will secure up to 5 GWs of current and future generations of Trainium chips to train and power advanced AI models.
Meta DC in Tulsa: Tulsa, OK will be home to Meta’s newest AI data center, located at the Fair Oaks 2,000- acre industrial innovation park. It will span more than 2 million sq.ft. and will spur an addition $25 million for roads and water systems.
Digital Shovel AP Pods: Digital Shovel premiered Modular AI Pods at Data Center World, showcasing pre-terminated fiber assemblies and factory-integrated power to reduce on-site installation from months to just days or weeks, with $2MM/M pricing.
Ampace AI-ready battery: Ampace introduced the PU200 at Data Center World, a semi-solid battery for UPS applications, engineered to absorb rapid fluctuations, and support uninterrupted operation under highly variable AI workloads.
Chatsworth edge computing: Chatsworth Products unveiled a new VersaEdge™ Wall-Mount Cabinet at Data Center World, and showcased its eConnect® Universal Input PDUs and eConnect® Sensor Array for hi-density environments.
Cloud Next ’26: Google Cloud announced a $750 million fund to deliver new resources and incentives to partners in its 120,000-member partner ecosystem to help accelerate joint customers’ transformations with agentic AI.
Google TPU8i: Additionally at Google Cloud Next, Google announced that in partnership with DeepMind, its TPU 8i would be “coming soon,” custom built for training and inference, with a focus on scale across training and agentic workloads.
RCR Events
Defense Communications Forum, April 28thJoin defense leaders, technology innovators and system integrators and explore how commercial network solutions can be adapted to meet defense-grade requirements, enabling secure, resilient, AI-driven, multi-bearer networks. Test & Measurement Forum, May 19thJoin the annual Test & Measurement Forum where industry leaders, innovators, and engineers come together to tackle the most pressing test and measurement challenges in telco today. Register now
Industry Resources
Webinar, April 21st: Scaling AIOPs from insight to action
Webinar, May 21st: Securing telecom infrastructure for the quantum era
Report: AI in testing: Developing trust, delivering results
Whitepaper: Powering sovereign AI at scale
Report: Networks at the frontline: How 6G, ISAC and NTNs will redefine defense technologies
Summit Access: GSMA Device Enablement Summit: How operators can fix device-network fragmentation