Back to briefings
Download PDF

January 31, 2026

Global

Hyperscalers step up AI data-centre capex guidance Mississippi emerges as a major US AI/data-centre investment hub Project finance tightens; customer-funded models gain attention European grid constraints push development to new regions Heat reuse and efficiency measures highlighted in Europe Battery storage fundraising links directly to AI/data-centre deployments

Top news (global)

  • Hyperscalers signaled another step-up in AI data-centre spend: Microsoft and Meta earnings reveal AI data center buildout highlighted Microsoft’s US$37.5bn quarterly capex (two-thirds toward chips) and Meta’s US$72.22bn 2025 capex guidance and US$115–135bn for 2026, with constraints called out around power, silicon, supply chains, and financing/ownership models.

  • A major US AI-campus-scale build was disclosed in Mississippi: xAI to build $20 billion MACROHARDRR data center in Southaven described >$20bn of corporate investment, a retrofit of a purchased building near a newly acquired Southaven power plant site, and a target to boost computing power to nearly 2 GW.

  • Bank commentary pointed to tightening capital availability for data-centre builds: Oracle may cut jobs, sell Cerner amid financing strain said TD Cowen estimates US lenders have retreated from data-centre project financing (raising borrowing costs), with Oracle exploring customer-funded models (e.g., 40% upfront deposits and BYOC) amid an estimated US$156bn infrastructure capex requirement.


Key deals & projects

North America

  • Mississippi (US): supply-chain buildout around large campuses

  • Michigan (US): proposed HPC/data centre with unusually detailed utility footprint disclosures

  • Canada: rural connectivity investment (relevant to edge/last-mile enablement)

Europe

  • France (Paris): portfolio M&A and AI-density positioning
    • nLighten acquires Émerainville Paris data center from oXya: nLighten acquired an Émerainville facility near Paris (about 1 km from PAR1), adding its 8th French site and taking the platform to 30+ data centres across seven markets. Site to keep serving oXya under a long-term master services agreement; described as designed for high-density, AI-ready configurations.

Power, grid capacity, and energy infrastructure

Grid constraints are changing where capacity gets built (Europe)

  • Power constraints drive European data centre growth to new regions flagged grid capacity constraints pushing expansion toward the Nordics, Spain, and Portugal (cooler climates and renewables).
    • Power-demand forecast cited: 96 TWh (2024) → 168 TWh (2030).
    • Heat reuse example cited: atNorth supplying surplus heat from its Espoo data centre to Kesko, cutting >200 tCO2 annually.

US system-level demand outlook and annual build requirement (policy research)

Storage financing signal relevant to power-constrained AI campuses

  • Redwood raises US$425m to scale battery recycling and storage: Redwood Materials closed a US$425m Series E (Google joined investors led by Eclipse; participation from Capricorn, Goldman Sachs Alternatives, and NVentures).
    • Proceeds intended to accelerate Redwood Energy grid-scale storage deployments, including a cited 12 MW / 63 MWh system deployed for Crusoe, and to scale Pack Manager-based BESS positioned for data centres and AI infrastructure.

Policy and regulation

Broadband / digital infrastructure policy (US)

EU agenda items with potential downstream relevance to data-centre power and networks


Operators, platforms, and “AI factory” positioning (market structure signals)

  • Neocloud positioning for dedicated AI training/inference

  • Developer/operator business-model separation


Two-line close

Capital intensity is still rising, but financing structures and power availability are increasingly shaping where and how new AI capacity gets built.
Expect more experimentation in customer-funded models and storage-backed power solutions as grid constraints bite.

Subscribe to Data Centres Briefings

Get AI-powered briefings delivered to your inbox

Region