d-Matrix has announced plans to adopt NVIDIA NVLink Fusion for its new Raptor XPU and to join NVIDIA’s AI platform partner ecosystem. The company intends to use NVIDIA MGX infrastructure components, including standardized rack designs, power delivery and liquid cooling, for AI-inference deployments.
Google AI and Infrastructure executive Amin Vahdat said AI infrastructure bottlenecks can shift rapidly among compute, power and memory, requiring short-cycle adjustments to resource allocation. Speaking at SEMICON Taiwan 2026, he also highlighted workload-specific TPU designs, reduced data movement and dynamic cooling as areas of hardware development.
Huawei has demonstrated a 7.2Tbps near-packaged optics (NPO) solution at the China International Optoelectronic Exposition and said it intends to deploy the technology in its superPoD supercomputing systems. The company is also pursuing an NPO specification through the Optical Internetworking Forum, although the proposal has not yet become an official OIF standard.
NVIDIA has announced a framework with eight Australian cloud and data-center infrastructure operators to prepare up to 2 GW of AI infrastructure by 2027. The figure is a maximum programme target, not operational capacity or a confirmed construction pipeline.
US-Armenian startup Firebird AI has opened a compute data center in Hrazdan, Kotayk Province, Armenia, built on NVIDIA technologies. The sole provided source describes the facility as the region’s largest and says Firebird secured a US export licence for NVIDIA B200 accelerators destined for Armenia in November 2025.
The US House of Representatives is expected to consider the bipartisan Ratepayer Protection Act next week. The bill would require state utility regulators to consider rules under which data centers bear part of the cost of new generation needed to serve their electricity demand.
Japan’s Ministry of Economy, Trade and Industry has included Nanto, Toyama Prefecture, in the first group of GX Strategic Areas. The designation is linked to measures intended to encourage data-center investment, secure land and assess coexistence with the local community.
Nanto City in Toyama Prefecture, Japan, has been identified as a GX strategic region, with reported measures aimed at encouraging data center investment. The available source headline references land provision and an assessment of coexistence with the local community.
Nebius Group has reportedly notified customers that it will raise on-demand rates for selected Nvidia GPU resources, including H100, H200, B200 and B300, from 1 October. The reported increase is about 20% and would be the company’s second price rise in three months.
Sunly is seeking a construction permit for a data center adjacent to a solar park in Risti, Estonia. A source headline says the company has secured an investor to finance construction, although no transaction details were disclosed.
Arm chief executive Rene Haas said the company is more confident than it was in July that its new specialized data-center AI CPU can reach a $2 billion revenue target. The increased confidence is linked to improved supply visibility and work to expand manufacturing capacity, rather than disclosed shipment volumes or booked revenue.
At the AI Infra Summit in Silicon Valley, held from 15 to 17 September, participants identified memory, power and interconnects as core constraints on AI infrastructure. Google, Intel and Qualcomm executives focused in particular on memory availability, cost and bandwidth as limits on effective accelerator utilisation.
Google has announced plans to invest at least €13 billion in AI infrastructure in Finland over the next two years. The announcement has prompted Finnish opposition parties to call for a national approval process for new data-center investments, centred on generation adequacy, grid capacity and consumer power prices.
According to S1, seven Chinese government agencies have jointly issued a 2026–2030 digital-green transformation plan that calls for deploying 800V high-voltage DC power architectures in computing infrastructure and promoting large-scale practical use of racks above 100kW. The reported policy direction targets the power-delivery and cooling requirements of high-density AI workloads.
Investor Mark Cuban has published a 10-point set of recommendations for US municipalities considering their first AI data center project. The proposals call for enforceable limits and financial protections covering water use, noise, backup generation, infrastructure costs and eventual decommissioning.
Oracle reported more than $30 billion in new AI cloud orders for the first quarter of fiscal 2027, with remaining performance obligations reaching $664 billion. Cloud Infrastructure revenue rose 121% year on year to $7.4 billion, alongside quarterly capital expenditure of $28.5 billion.
Power semiconductor supplier PANJIT International reported August revenue of TWD 13.77 billion, up 29.82% year on year despite a 4.89% month-on-month decline. The company identifies AI data centers and AI servers among the demand drivers for its power-device portfolio.
JPMorgan has upgraded French semiconductor-materials supplier Soitec to Overweight from Neutral and raised its price target to €200 from €98, citing expected growth in silicon photonics. The bank expects the company’s silicon-photonics revenue to exceed €1 billion by fiscal 2030.
Global Information, the Japanese agent for 360iResearch, has announced the start of sales of a report covering the global data-center power market through 2032. The report projects the market will reach $38.92 billion by 2032, representing a 9.56% CAGR from 2026.
Thailand has temporarily halted 166 data-center projects while it reviews requirements covering power, water, siting, safety and project economics. The affected portfolio includes 117 projects that are either under construction or awaiting approval.