Digital Infrastructure Monthly Insights

Relativity Networks closed a $22M funding round with investors including Rhapsody Venture Partners, Bell Ventures, and Faster Than Glass LLC, and received a $40M purchase order from an unnamed hyperscaler for hollow-core fiber.

Marvell signed a commercial agreement with Google on July 29 and issued a warrant on August 18 granting Google the right to purchase up to 58.97 million Marvell shares at $206.58 per share, exercisable until August 2033. The warrant is tied to an expanded collaboration to develop custom chips for Google.

IBM will provide infrastructure under a $240M agreement to Together AI, supporting AI workload capacity on an AI-optimized cloud.

SK Telecom announced the launch of an AI-focused data center under the SK Horizon initiative to support AI workloads. The source snippet does not provide specifications such as capacity, location, or vendor partners.

Zayo and Nvidia announced a collaboration focused on expanding network capacity and connectivity to better support AI-driven traffic and data center interconnect needs.

Microsoft commissioned a new data center in Hyderabad, activating its fourth India cloud region (with Pune, Chennai, Mumbai). The facility is designed for AI workloads, supports sovereign cloud requirements, and integrates with Microsoft’s global cloud. Part of a >$20B India commitment, including $17.5B (2026–2029); Hyderabad region planned with three availability zones and positioned as the company’s largest hyperscale region in India.

Anthropic confirmed formation of an internal silicon design team to develop custom AI chips co-designed with its Claude model, targeting faster and more efficient operation at customer scale and recruiting engineers across the stack.

OpenAI released benchmark data for its Jalapeno custom inference chip co-developed with Broadcom, indicating higher speed and power efficiency than Nvidia’s Blackwell Ultra-based GB300. Reported package power: 700W for Jalapeno vs 1,400W for GB300. Tests used SemiAnalysis’ InferenceX end-to-end AI request-serving benchmark.

AM Intelligence placed an order for approximately 9,000 NVIDIA Rubin GPUs to build one of Asia’s first Vera Rubin clusters using NVIDIA Vera Rubin NVL72 rack-scale systems at a 30MW AI factory in Hyderabad, India. The facility, slated for Q1 2027, is designed to run trillion-parameter models and agentic AI applications.

SK Telecom launched SK Hyper as a dedicated AI data centre unit, targeting 5 GW capacity by 2029 and 15 GW by 2035, with 750 billion won investment by 2030. It is co-developing an AI data centre in Ulsan with AWS and aligning group capabilities (SK Hynix, SK Broadband). Technically, SKT is integrating Arm AGI CPU and Rebellions' Rebellcard AI inference accelerator on a common server platform, and developing a CXL-based architecture with Panmnesia to pool CPU/GPU/memory resources for higher utilisation. For power and cooling efficiency, it is collaborating with Supermicro and Schneider Electric.

GCT progressed development and certification in Q2 2026 for 4G/5G chipsets and user equipment enabling direct-to-device and hybrid satellite–cellular connectivity, working with Globalstar and another satellite provider. Earlier in 2026, it licensed its 4G/5G chipsets to a satellite communications provider for next‑generation user equipment, partnered with Skylo for joint chip/module certification for satellite fallback, and signed a reference-platform supply agreement for 4G/5G chipsets.

KT is expected to reintegrate its spun-off KT Cloud under a reported $12.6B investment plan to become an AX platform company, while LG Uplus is committing nearly $1B in additional capex to expand AI data center capacity. Both operators are reallocating focus from saturated mobile services to higher-margin AI infrastructure.

Cisco announced a commercial alliance with Supermicro to offer Supermicro liquid- and air-cooled rack-scale systems packaged with Cisco Secure AI Factory and Nvidia products. Availability begins in October.

Elon Musk stated that SpaceX will begin deploying orbital AI data centers in Q4 2027 using a design powered by Nvidia chips, with operations reaching significant scale in 2028.

Qualcomm is establishing the Qualcomm Japan Robotics Center to collaborate with Japanese technology companies on robotics and industrial automation, with a focus on open ecosystems and edge AI R&D.

Zayo extended its supply agreement with Corning to lock in a significant share of fiber optic cable needs through the rest of the decade, providing supply assurance for one of its largest network expansions as AI infrastructure growth tightens material availability.

Anthropic formed a strategic partnership with Macquarie Asset Management and GIC to develop AI computing sites in the United States using the Theseus Infrastructure platform. Macquarie and GIC will fund most project equity, while Anthropic will lease the facilities long term and assume exposure to consumer electricity price increases linked to the sites.

SK Group and Nvidia signed Letters of Intent for a USD 500B+ partnership to build large-scale AI factories and secure next-generation AI memory supply. SK Telecom will deploy a 2 GW AI Cloud in South Korea using Nvidia DSX and Nvidia Vera Rubin accelerated computing with SK hynix HBM4, with the first AI factory targeted for 2027. SK hynix and Nvidia will co-develop and optimize HBM and other AI memory for workloads spanning large language model training, agentic AI, and physical AI.

According to The Information, Nvidia has agreed to acquire open-source AI platform Hugging Face for approximately $12.9 billion, gaining control of a large repository of AI models and datasets. The report cites Hugging Face’s annualised revenue at around $150 million.

Bloomberg reports Nvidia met Rebellions at Nvidia’s Santa Clara headquarters to discuss potential collaboration options, including a technical partnership, equity investment, or full acquisition; no terms or product details disclosed.

India introduced a comprehensive semiconductor scheme covering six segments: design by Indian firms, establishing units for chip-production capital equipment, building semiconductor fabs, and scaling assembly, packaging, and testing capabilities.

Huawei proposed an Egypt government AI data center using 2,008 Ascend accelerators (1,408 Ascend 950-series for training and ~600 additional Ascend chips, potentially 910B, for inference) on a 12-month build schedule. The US State Department is engaging Nvidia, AMD and Microsoft to form a competing offer, positioning an alternative AI hardware/software stack for Egypt’s national AI infrastructure.

Across the Asia-Pacific, governments have implemented resource caps and efficiency measures for data centers, which has reduced the likelihood of US-style public and political pushback against AI-related facilities.

Micron launched Micron Research Labs in Boise, Idaho as a long-horizon R&D hub, committing $10B over 10 years to advance memory and compute for AI. The initiative brings together customers, academia, government, start-ups, and the broader semiconductor ecosystem to pursue innovations beyond current technology roadmaps, with a focus on critical memory technologies.

Velaura AI secured $110M in Series A financing led by Seligman Ventures, valuing the chip designer at over $1B. Funds will accelerate development and deployment of low-power AI data center chips and the Titan Core chip design platform, expand software for data center and physical AI workloads, and support hiring in engineering and customer-facing roles.

Nvidia signed memorandums of understanding with major investment firms to establish global AI compute infrastructure financing platforms aimed at mobilizing over $500 billion in third-party capital to fund frontier AI laboratories, cloud providers, and enterprises.

SpaceX outlined the first phase of its Terafab project to build a large semiconductor fabrication facility in Texas, committing over $16.8B to produce AI chips for its satellite and space businesses. The initiative is supported by a $30M Texas Enterprise Fund grant, is eligible for additional state incentives, and is projected to create 3,000 jobs.

Elon Musk indicated the company intends to build AI chips at scale for terrestrial and space deployments, implying in-house compute for satellites and ground systems. No technical specifications, timelines, or partners were disclosed.

AI agents are being operationalized in semiconductor workflows. SK hynix is piloting agent-based automation in back-end manufacturing (assembly/packaging/test). Samsung reports reducing SoC verification cycles to two days using AI-driven verification agents integrated with EDA flows.