HPE unveiled AI networking updates, expanded private AI offerings, and a vision for the 'agentic enterprise' at its Discover 2026 event, emphasizing networking as the foundation of AI infrastructure following its $14 billion Juniper Networks acquisition.
Global server revenue reached $122.6 billion in Q1 2026, with non-x86 architectures accounting for 47.9% of the total, driven by AI infrastructure investment and supply constraints for traditional x86 systems.
North American data center operators are shifting to on-site power generation to overcome grid constraints, using natural gas turbines and reciprocating engines to align power delivery with construction timelines and AI-driven demand.
German operator firstcolo has begun construction on a 24MW data center in Rosbach vor der Höhe, designed for high-density AI and cloud workloads with liquid cooling and green electricity. The €250M project aims to address local demand for sovereign infrastructure while navigating power constraints and sustainability expectations.
Data center operators are shifting from sustainability discussions to operational efficiency, with power availability, cooling infrastructure, and utility timelines becoming critical factors in AI deployment decisions.
Brazos, Google's rack-mounted liquid cooling system, enables high-density AI and HPC deployments in existing air-cooled data centers with minimal infrastructure changes. The design will be open-sourced via the Open Compute Project.
firstcolo has broken ground on FRA7, a 24MW colocation data center in Hesse, Germany, designed for AI and high-performance computing workloads. The €250 million project targets 200kW rack densities, liquid cooling, and 100% certified green electricity.
Israel is launching a NIS 150 million photonics R&D infrastructure program and a quantum startup ecosystem initiative to bolster domestic hardware production amid global supply chain fragmentation.
Hewlett Packard Enterprise (HPE) is partnering with eight quantum technology companies to develop hybrid architectures combining supercomputers, AI accelerators, and quantum processors. The initiative aims to address infrastructure challenges like ultra-low latency interconnects and workload orchestration for future scientific and industrial workloads.
The lapse of the Federal Data Center Enhancement Act (FDCEA) could remove standardized requirements for federal datacenters, including uptime, sustainability, and security, as neither Congress nor the Trump administration has moved to extend or replace it.
Crusoe says contracted capacity across its data center projects and Crusoe Cloud has reached 4.9GW, with a broader development pipeline exceeding 40GW — numbers that highlight both the appetite for AI infrastructure and the execution challenges still ahead.
KKR has formally launched Helix Digital Infrastructure, an AI-focused infrastructure company led by former AWS CEO Adam Selipsky, with anchor investors including Kuwait Investment Authority, NVIDIA and Vistra and commitments exceeding $10 billion.
NEMA, ASHRAE, and PNNL have jointly released the AI Data Center Energy Performance Framework, a unified guide covering electrical systems, thermal management, and operational practices for power-intensive AI facilities.
A detailed look at where fuel cells, hydrogen engines, and renewable diesel stand as alternatives to diesel generators in data centers, with Bloom Energy and Oracle striking a 2.8 GW deal and Microsoft running multiple pilot programs.
New York is pursuing parallel strategies to manage surging data center power demand: a legislative moratorium on new builds above 20 MW and a Public Service Commission proceeding that could impose new cost-allocation rules and grid obligations on high-load facilities.
American utilities are restructuring their planning, financing, and procurement operations after discovering that only a fraction of AI-related electricity requests translate into signed agreements or energized facilities.
Synchronized pauses baked into modern AI training create sharp power-demand swings that operators mask with secondary workloads, inflating consumption, accelerating hardware wear, and lengthening grid interconnection timelines.