IndiaMicrosoft announced that its India South Central region in Hyderabad is live, giving it four Indian cloud regions alongside Pune, Chennai and Mumbai. The Hyderabad deployment uses a three-zone architecture, while Microsoft says its Indian regions now contain AI-capable hardware for advanced workloads. The facility uses high-efficiency mechanical cooling designed for effectively zero water use for cooling. The development forms part of Microsoft's previously announced $20.5 billion India commitment through 2030; Microsoft did not disclose the Hyderabad region's MW capacity.
IndiaAdaniConneX subsidiaries Giridhari Build Estate and Granthik Realtors secured an $800 million bridge financing facility from a syndicate including Natixis, Rabobank, SMBC, DBS, MUFG, ING and Intesa Sanpaolo. The financing supports construction, operation and maintenance of five Hyderabad projects: HYD22 — 80MW; HYD23 — 30MW; HYD24 — 80MW; HYD25 — 80MW; HYD26 — 40MW, totaling 310MW. The $800m is financing rather than capex already deployed, while 310MW represents stated project capacity rather than operational capacity.
IndiaAdani Group announced plans to invest ₹1 trillion, approximately $10.4 billion, in West Bengal through 2035 across data centres, clean energy, logistics, transport and other infrastructure. The announcement signals additional potential digital-infrastructure investment in eastern India, but Adani did not disclose how much of the ₹1tn is allocated to data centres, nor associated MW. The amount therefore cannot be counted as dedicated data-centre capex.
GCCMicrosoft announced plans to invest more than $10 billion through 2030 across the UAE, Saudi Arabia, Qatar and Kuwait, focused partly on expanding cloud and AI infrastructure and improving digital resilience. Microsoft also plans more than $400 million of investment in subsea and terrestrial connectivity across the Middle East. Partnerships cited include G42 in the UAE, Humain in Saudi Arabia and Qai in Qatar. Microsoft did not disclose the country-level allocation or MW attached to the $10bn programme.
QatarMEEZA awarded Elegancia Contracting & Industries a turnkey contract worth more than QAR500 million (~$137 million) for a new AI-ready data centre. Construction will take place in two phases, covering civil, MEP, ICT, controls, security, testing and commissioning. The Qatar Science & Technology Park facility is intended to expand Qatar's ability to host hyperscale cloud and high-density AI workloads. Neither MW capacity nor completion date was disclosed.
QatarQatar's Ministry of Communications and Information Technology and Google Cloud launched an Innovation Lab designed to give developers, researchers and businesses infrastructure for developing and testing AI applications. The initiative expands Qatar's government-cloud and AI ecosystem around Google Cloud's existing Doha region and includes further cloud-skilling activity. No additional physical MW, GPU count or new cloud region was announced, so this is primarily a sovereign/government-cloud capability development rather than a new capacity build.
United StatesCipher Digital secured a further 10-year lease for its Barber Lake HPC development in Texas from an unnamed leading AI lab following Fluidstack's initial term, taking the aggregate contracted period to 20 years. The new lease is expected to add approximately $5.2 billion of contracted revenue, lifting total contracted facility revenue above $9 billion. Cipher lists Barber Lake at 300MW of contracted capacity, with individual data halls expected to be delivered between Q4 2026 and Q1 2027. Cipher will bear the first $359.3m of costs above the initial budget, with the tenant reimbursing 50% of further excess costs over the contract term. The 300MW is contracted/developing capacity, not operational capacity.
United StatesVolato Group's Alignment Engine subsidiary entered a Master Services Agreement covering two dedicated single-tenant GPU deployments with an aggregate contractual value of approximately $1.17 billion. The first cluster will use AMD MI355X GPUs for 48 months starting contractually on 31 December 2026; the second will use next-generation AMD MI455X GPUs for 48 months beginning 30 June 2027. The infrastructure will be deployed at Alignment Engine's Ohio AI campus. The SEC filing establishes that the MSA was entered on 22 September, although the public press release came on Sept. 28. No GPU count or MW was disclosed.
United StatesAtlas Energy Solutions disclosed purchase commitments covering 283MW of Caterpillar natural-gas generation equipment for a dedicated power project. An unnamed leading frontier AI lab is supporting the equipment purchase through a reimbursement arrangement and is intended to become the project's power off-taker. The agreement is significant because an AI compute buyer is directly helping underwrite generation infrastructure to overcome grid constraints. The 283MW represents generation equipment committed, not operational generation.
United StatesOracle issued a force-majeure notice related to the New Mexico Project Jupiter AI data-centre development because of potential delays in securing power. The project has approximately $18 billion of loans and around $3 billion of Blue Owl equity, and is linked to infrastructure serving OpenAI. Oracle is seeking protection against higher rent commencing before the site can meet delivery requirements; Oracle nevertheless says the project remains on schedule for its targeted 2028 operation. The event demonstrates how power availability is now affecting financing and lease economics, not merely construction schedules.
United StatesThe US Department of Energy announced its intention to help fund 31 grid-improvement projects across 26 states under its Speed to Power programme. The programme covers transmission upgrades, advanced reconductoring and grid-enhancing technologies intended to improve reliability and accelerate connection of large new loads, including data centres. The announcement represents proposed/selected grid investment rather than completed infrastructure; individual projects should therefore not be counted as commissioned power.
United StatesCalifornia Governor Gavin Newsom signed seven bills governing data-centre electricity, water, land use and local oversight: AB 1577, AB 2383, AB 2469, AB 2619, SB 886, SB 887 and SB 1168. The package requires additional reporting of electricity and water use, seeks to prevent grid-upgrade costs from being shifted to ordinary ratepayers, requires applicable facilities to meet energy-procurement requirements and adds water-infrastructure and environmental-review provisions. Data centres can be required to pay for grid and water upgrades associated with their projects.
European UnionThe European Commission moved forward with a common sustainability rating framework covering individual data centres with installed IT power demand above 500kW. Measures include PUE, WUE, clean-energy sourcing, grid flexibility and waste-heat reuse. EU data-centre electricity use was approximately 68TWh in 2024 and is projected to reach about 114TWh by 2030, equivalent to roughly 3.2% of EU electricity consumption. Initial labels are expected from 2027. This is a rating/disclosure regime rather than an electricity-consumption cap.
European UnionThe European Commission opened a 12-week consultation on minimum environmental and energy-performance standards for EU data centres, with submissions due by 14 December 2026. The work could result in standards addressing energy consumption, water use, waste-heat reuse and integration with electricity grids. A legislative proposal is planned for Q2 2027. No actual minimum performance thresholds have yet been adopted.
United KingdomVIRTUS Data Centres secured a £2.45 billion financing package, including a £1.2 billion green capex facility, to support continued UK and European expansion. The transaction supplies capital for future data-centre development rather than representing £2.45bn already deployed. At a rounded late-September GBP/USD rate of roughly $1.34/£, the package is approximately $3.3 billion.
United KingdomUK-headquartered Nscale raised $3.36 billion through convertible loan notes, led by Third Point and supported by investors including Nvidia, Apollo-managed funds and Abu Dhabi Investment Council. Nscale reports more than $103 billion in total contracted value and says the capital will accelerate its vertically integrated infrastructure, including behind-the-meter power plants, liquid-cooled AI data centres and large-scale GPU clusters. The $103bn is company-reported TCV rather than recognized revenue, while $3.36bn represents financing rather than installed infrastructure.
GermanyFrux Capital provided financing connected with FREO's development work on Schwarz Group's large data-centre project in Mecklenburg-Vorpommern, Germany. The previously announced project's first phase targets 240MW of computing capacity, while the site could ultimately obtain up to 1GW of power capacity; Schwarz has previously outlined approximately €5.6 billion of investment through 2033. The financing amount itself was not disclosed. The 240MW, 1GW and €5.6bn figures are background project metrics and should not be represented as newly announced capacity this week.
GermanyAWS made EC2 C8i and C8i-flex instances available in the AWS European Sovereign Cloud Germany region. The instances use custom Intel Xeon 6 processors and AWS claims up to 20% higher performance, 2.5× memory bandwidth and up to 15% better price-performance versus relevant previous-generation infrastructure, with workload-specific gains including up to 40% for AI deep-learning recommendation models. The development expands operational compute options inside AWS's sovereignty-focused European environment but does not represent a new physical cloud region or disclosed incremental MW.
ChinaAlibaba set a target for Alibaba Cloud-operated global data-centre capacity to exceed 20GW by 2032. Alibaba Cloud also announced plans to establish its first cloud regions in Türkiye, Finland and the Netherlands over the following 12 months while expanding data-centre footprints in Malaysia, Germany, UAE, France and Hong Kong. The company currently operates across 31 cloud regions and 107 availability zones. The >20GW figure is a future company target, not current operational or contracted capacity.
South KoreaHyundai E&C began fast-track construction of the 40MW Pohang AI Data Center in South Korea, targeting completion in November 2027 on a 16-month construction schedule. The facility is designed specifically for AI workloads and incorporates Nvidia GPU-based systems, liquid cooling and enhanced energy-efficiency systems. Hyundai is using precast construction, preconstruction services and fast-track delivery to compress the build schedule. The 40MW is planned/project capacity under construction rather than operational capacity.
JapanSoftBank Group raised $11.1 billion through dollar- and euro-denominated senior bonds in what Reuters described as the world's largest high-yield corporate bond sale. Yields ranged from 7.125% to 9.75%. The financing supports SoftBank's large AI investment programme, which includes a $64.6 billion commitment to OpenAI, alongside other investments such as its DigitalBridge acquisition. The $11.1bn should be treated as corporate financing supporting the AI investment cycle, not direct data-centre capex.
AustraliaVocus began early development of a new approximately 4,000km ducted fibre corridor connecting Brisbane, Townsville and Darwin under its Australian Digital Infrastructure Platform. The network is designed to accommodate up to 3,456 fibre strands and is targeted to enter service in 2030. Vocus links the route to increasing hyperscaler, neocloud and AI/cloud infrastructure demand. Development and design have begun, but the full route is not yet operational.
AustraliaVictoria introduced its Sustainable Data Centre Action Plan with stronger planning and infrastructure conditions for future developments. New facilities are expected to support additional renewable generation, contribute to storage and relevant electricity-network upgrades, while the planning framework includes a 150-metre separation requirement between data-centre buildings and homes. The development indicates that Australian jurisdictions are increasingly requiring developers to internalize grid and community impacts rather than treating power connection as purely a utility issue.
GlobalSupermicro announced that it has begun shipping Nvidia Vera Rubin NVL72 racks integrated with its Data Center Building Block Solutions and direct liquid-cooling stack. Supermicro's infrastructure blueprints range from 5MW to gigawatt-scale deployments, with each Scalable Unit containing 1,152 GPUs and 331TB of HBM4. Its integrated cooling stack extends from cold plates through CDUs to cooling towers. The announcement marks product availability rather than proof that the stated GPU or MW quantities have already been deployed at customer sites.
GlobalVertiv announced that its 2.3MW CoolChip coolant distribution unit had become the first CDU specifically qualified as Nvidia DSX Ready. The system supports direct-to-chip and rear-door heat-exchanger applications and is designed for edge, colocation and hyperscale AI environments. A 2.3MW thermal rating illustrates the increasing scale of cooling infrastructure required for high-density AI compute. The figure represents CDU cooling capacity, not newly commissioned data-centre IT load.
