Digital Infrastructure Monthly Insights

Uztelecom has commissioned two Tier III design-certified data centres in Bukhara and Kokand, expanding its UzCloud platform with AI computing, GPUaaS, cloud services and geographically distributed infrastructure.

BDx Data Centers initiated construction of a 640MW AI-focused data center campus in West Java, expanding hyperscale compute capacity in Indonesia.

Telxius has deployed a carrier-neutral Landing Gateway and Tier III Edge Data Center in Cancún integrated with its global backbone. The site will serve as Mexico’s landing point for the Tikal submarine cable, features N+1/2N power and cooling redundancy and advanced security, and enables multi-operator/cloud interconnection. Tikal is slated for H1 2027 with an initial 380 Tbps trunk, providing a diverse high-capacity route to the United States and alternative routing toward Santiago de Querétaro.

VAST Data introduced DataEnclave, a confidential AI capability integrated with VAST DataEngine, enabling proprietary and open AI models to run in customer-controlled environments (on-prem or trusted cloud). The solution uses hardware-isolated execution with encryption of guest memory, GPU memory, and NVLink traffic; cryptographic attestation including NVIDIA GPU attestation; and independent key control via BYOK KMS. It supports connected or fully air-gapped deployments with attestation services based on CNCF Trustee or via Fortanix. Previewed now; general availability targeted for Q1 2027 through partners including Cisco and Supermicro.

Mitsubishi Electric has introduced a chip-to-grid infrastructure blueprint designed for AI factories powered by NVIDIA accelerated computing. The approach brings together electrical systems, cooling, automation and energy management to support the growing power density, efficiency and reliability requirements of large-scale AI infrastructure.

STL Networks approved the formation of a wholly owned subsidiary to build and own digital infrastructure, obtain telecom licenses, and deliver data-center connectivity services for hyperscalers, data center operators, and enterprises in India. The new unit will provide an integrated infrastructure-plus-connectivity offering and assume greater ownership across the delivery lifecycle.

Fujitsu will start global sales in November of its next-generation Fujitsu-Monaka CPU, offered separately to cloud and data centre operators and server vendors. A Monaka Server optimized for AI inference and sovereign AI infrastructure will be available initially in Japan and Europe.

d-Matrix will integrate its next-generation Raptor XPUs with NVIDIA's AI infrastructure using NVLink Fusion, enabling high-bandwidth, low-latency scale-up networking within NVIDIA MGX rack architectures. The deployment plan includes NVIDIA Spectrum-X Ethernet for scale-out, Vera CPUs, ConnectX-9 SuperNICs, and BlueField-4 DPUs, with interoperability alongside NVIDIA Vera Rubin NVL72 systems for disaggregated inference. The approach leverages NVIDIA’s validated rack designs, liquid cooling, and supply chain to reduce integration time and risk for large-scale inference deployments.

HCLTech has launched a new semiconductor laboratory in Bengaluru as part of a ₹185 crore investment to strengthen its engineering and R&D capabilities. The facility will support semiconductor design, validation, testing and advanced technology development for global clients across the chip ecosystem.

Equinix introduced Fabric One, a managed service to automate connectivity between enterprise sites, clouds and AI environments, and unveiled Inference Exchange, a distributed AI inference program developed with NVIDIA and Together AI to run inference across Equinix data center and network footprints.

Swisscom deployed NVIDIA DGX SuperPOD clusters as part of a CHF 100m Swiss AI Platform program, adding on-premises GPU compute capacity in Switzerland.

SK Telecom has put Rebellions NPUs into production to operate four in-house AI services, indicating on-prem inference capacity scaling in South Korea.

ETDatacenters reports that average data center rack loads doubled from 12 kW in 2024 to 26 kW by 2026, driven by AI workloads and space constraints. Designs in some cases target up to 1 MW per rack. Supporting this shift requires re-architected high-amp redundant power delivery, broader adoption of liquid cooling (direct-to-chip, rear-door heat exchangers, immersion), and structural upgrades to handle increased floor loads and service complexity.

At a UK event hosted by King Charles III on 17 September, CEO Jensen Huang said Nvidia expects chip sales to double in 2027, citing accelerating AI adoption across industries. Bloomberg reported the remarks. Nvidia remains the leading supplier of AI accelerators and does not disclose chip unit sales.

Vendors flagged supply bottlenecks impacting AI data center builds and telecom network upgrades, noting risks across cloud and broadband infrastructure in the domestic market.

A government-backed plan to build an 800-trillion-won semiconductor cluster in the Honam region faces delays due to power and water capacity constraints and labor-relations uncertainty following the revised Trade Union and Labor Relations Adjustment Act. The government granted a preliminary feasibility exemption for water infrastructure and is seeking a similar waiver for power. It is pushing an early start to site construction near Gwangju military airport with potential temporary relocation by mid-2028. The Ministry of Employment and Labor will issue non-binding administrative guidelines to limit legitimate labor disputes to restructuring-driven personnel transfers. Unions at Samsung Electronics (84% opposing in a vote) and SK Hynix are mobilizing against relocations, and industry argues that non-enforceable guidance will not curb bargaining or disputes over site assignments.

Meta signed an agreement to use Firmus' GPU computing power in Southeast Asia to support AI research and model training.

Charter Communications activated its Spectrum Edge Compute Infrastructure leveraging NVIDIA platforms and announced new collaborations with Cast AI and HP to support edge AI workloads.

Anthropic agreed to an $11.6 billion, seven-year contract for computing capacity from Akamai, focused on CPU-based resources to support AI services. The agreement includes a warrant enabling Anthropic to purchase Akamai Series B shares at $111.33 per share.

Indosat Ooredoo is collaborating on Zankore, a dedicated AI compute and neocloud platform targeted at the Indonesian market. The announcement references an AI-focused cloud offering but does not provide technical specifications, capacity, or launch timelines.

Qualcomm and Amazon are partnering to develop customized silicon for AWS’s AI infrastructure, featuring optical connectivity up to 1.6 Tbit/s.

Nscale has signed a multi-year strategic partnership with humanoid robotics company Figure, becoming its preferred compute provider and supporting development of Figure’s Helix AI models. The agreement includes an initial $3.5 billion compute commitment, potential expansion beyond $6 billion, and possible deployment of up to 100,000 NVIDIA GPUs beginning in 2027. Light Reading

Nvidia deepened its partnership with MediaTek by bringing MediaTek into Nvidia’s data center ecosystem and committing to purchase $3.5B of MediaTek-issued bonds.

AMD agreed to acquire AI research firm World Labs for approximately $8.2B in an all-stock deal, adding spatial intelligence models that generate, reconstruct, and simulate interactive 3D environments from text, images, and video, along with robotic learning technologies.

Nvidia confirmed a $13B acquisition of Hugging Face, aiming to scale the platform, strengthen infrastructure, and expand access for developers and institutions. Hugging Face reports over 18M users sharing 3M+ models, 500K datasets, and about 1M applications.

UK-based AI infrastructure company Nscale raised $3.4B in pre-IPO financing via a convertible note offering led by Third Point, with participation from Nvidia, Apollo, Citadel, Hudson Bay Capital, Abu Dhabi Investment Council, and 8090 Industries. The subscription agreement was signed on 15 September. The company is preparing for a US public listing, following earlier reports it sought up to $3.5B before flotation.

Microsoft will deploy more than $10B through 2030 to expand Azure cloud, regional data centres and AI infrastructure across the UAE, Saudi Arabia, Qatar and Kuwait, combining new and previously announced funding, with an added focus on improving service resiliency in the region.

NTT DATA is committing US$1.5 billion to build AI-ready data center infrastructure in India with an emphasis on sovereign compute and secure data residency. The program supports physical AI by enabling on-site edge inference and integrating private 5G for low-latency, secure operations. NTT DATA highlights work with Ericsson to combine private 5G and edge platforms with edge AI services for real-time autonomous decision-making.

Fujifilm will establish a greenfield semiconductor materials manufacturing facility in India through a two-phase investment of about ₹800 crore. Phase one is expected to break ground in FY2026, with commercial production targeted for FY2028, strengthening India’s domestic semiconductor supply chain.

Philippines announces a $34B national AI plan targeting a 30-fold increase in AI data center capacity and nationwide broadband improvements.

Tata Consultancy Services’ subsidiary HyperVault secured 264 acres in Hyderabad to develop a 1GW AI data centre campus in phases, budgeting up to INR700 billion ($7.4B). The build will align with customer demand and technology requirements and is expected to be among India’s largest AI infrastructure facilities once complete.

Sydney-headquartered, Nasdaq-listed cloud operator Iren proposed a $10 billion data centre project in Bundey, South Australia, intended to supply AI compute largely for U.S. technology companies, aligning with a broader national build-out that positions Australia to export AI capacity.

The Philippines’ three major telecom operators have allocated more than $2.2 billion in 2026 capital spending, with PLDT, Globe Telecom and Converge ICT investing in network expansion and digital infrastructure. The spending supports wider fiber deployment, the National Fiber Backbone and broader efforts to improve connectivity, resilience and digital services across the country.

Andreessen Horowitz launched a $1.1 billion Machine Age Fund to accelerate development of the physical infrastructure required to support scaling AI workloads as systems advance from chat interfaces to complex reasoning, coding, and autonomy. The firm highlights constraints in physics, computer science, and legacy supply chains as key bottlenecks.