Infrastructure
Data centers & powerData center construction, power, cooling, and the physical backbone of AI compute.
AI-generated editorial illustration.Shantou Brings Ascend Compute, Green Power and Submarine Cables Together
A new AI computing cluster in Shantou combines Huawei Ascend hardware, China Telecom data centers, offshore wind power and international submarine cable links to serve overseas token demand. The project is presented as China’s first Ascend 950DT cluster built for token exports.
Natural Resources AI Plan Moves Approvals Onto National Compute Network
China’s natural resources authorities plan to lease capacity from the national computing network and direct suitable workloads to East Data West Computing hubs. The policy also introduces AI into permitting, planning and disaster-response processes while leaving accountability and sensitive-data safeguards to later implementation.
U.S. Data Center Moratoriums Threaten to Cap AI Compute Growth by 2028
A wave of state-level data center moratoriums is raising the prospect that U.S. AI compute capacity could plateau near 2028. We examine what an extreme freeze on all unstarted construction would mean for AI training and deployment.
Substations, Turning Radius and 300MW Load Tests: The Physical Limits of AI Data Centers
At a Kuala Lumpur data center trade show, engineers described the unglamorous constraints throttling AI factory buildouts: mid-construction switches from liquid to air cooling, confidential rack dimensions and turning radius limits, and campuses past 300MW where no load bank on the market is big enough to test them. Sungrow's AIDC head also argued the legacy 480V power architecture has to be rebuilt.
Wuxi Domestic AI Data Center Discloses First Operating Data as Neolix Adds 1 Billion Yuan Project
Moore Threads' Wuxi (Huishan) domestic AI data center disclosed its first operating data: 2,112 MTT S5000 cards across 264 servers at 92.9% utilization, with only 59.7% of that compute paid for and training making up more than 62% of workloads. Neolix also landed an East China headquarters project nearby with 1 billion yuan of planned investment.
Wuhan Switches On 1,024-Card Ascend 950DT Super-Node: 1 EFLOPS FP8, 256TB Unified Memory
China's first 1,024-card Ascend 950DT AI computing cluster, funded by Hubei Science and Technology Investment Group, entered service in Wuhan on October 29, 2026, delivering 1 EFLOPS of FP8 compute and 256TB of global unified memory. The liquid-cooled super-node extends cooling to interconnect cabinets and went from project kickoff to switch-on in a single month.
China's Compute Exodus: Billions Spent Overseas, Blocked by Power and Policy
Chinese cloud providers, colocation operators and equipment vendors are spending billions on data centers in Southeast Asia and the Middle East, pushed out by misallocated domestic capacity, chip supply limits and customers that have already moved. Once there, power, water, land and shifting host-country rules become the binding constraints.
Xinjiang's AI Computing Power Nears 30,000 P, But Who Pays for the Capacity?
Compiled public data puts Xinjiang's AI computing power at close to 30,000 P as of mid-2026, with Karamay alone accounting for more than 17,000 P across six data centers at about 72% utilization. The region's cheap green power at around 0.3 yuan per kWh covers the cost side, but the article argues sustained customer orders, not construction milestones, determine whether these projects pay off.
Container AI Factories: NVIDIA's New POD Meets a Decade of Mining Prefab
At GTC Berlin, NVIDIA unveiled the DSX POD, a prefabricated AI factory in a shipping container that cuts deployment from about six months to as little as 14 days. Bitcoin mining vendors like Bitmain and FBOX have been shipping containerized, liquid-cooled compute for years, and are now turning that playbook toward AI data centers.
North American AI Data Centers Turn to On-Site Gas as Grid Queues Stretch 5 to 10 Years
A cluster of mid-October deals and filings shows North American AI data centers shifting to behind-the-meter natural gas as grid interconnection queues stretch five to ten years, with equipment backlogs booked out to 2028. Oracle has gone as far as trucking compressed natural gas to keep a Utah site running.
Ulanqab Tracks 33 Data Center Projects, With 43.88 Billion Yuan Filed but Only Five Building
IDC Circle's tally of 33 Ulanqab data center projects as of October 19 finds 18 at the filing or early-stage mark with 43.882 billion yuan in disclosed investment, against just five under construction. Filed projects account for roughly 280,300 planned racks, with four mega-campuses from Chindata, Volcano Engine and Jinkai New Energy making up 87.2% of that total.
OCP 2026: Vera Rubin Becomes the Industry's Reference Point
At OCP 2026, NVIDIA contributed its Vera Rubin rack and compute tray designs to the Open Compute Project through the open MGX architecture, releasing the MGX 1.2 specification with support for 250kW racks, blind-mate liquid cooling and a 5,000A busbar. The summit's agenda widened from server racks to data center templates, 800VDC power and grid resilience, with OEMs and the Taiwanese supply chain aligning around the platform.
China Mobile's Changping Node Adds 1 EFLOPS Ascend SuperPod, Backed by Nine-Chip Compatibility Layer
China Mobile says a 1,024-card Huawei Ascend 950DT SuperPod delivering 1 EFLOPS (FP8) has gone live at its Mobile Cloud Beijing node in Changping. The carrier also says Mobile Cloud has completed compatibility work for nine domestic AI chips and that its MoMA platform now hosts more than 300 models.
Data Centre World Asia Fills Four Floors as Power and Cooling Dominate the Floor
Data Centre World Asia in Singapore drew more than 600 exhibitors across four floors of Sands Expo, with attendance up 25% year on year. On the floor, the conversation had shifted from site selection to power availability, cooling and delivery speed.
IDC and Inspur: China's Intelligent Computing Power to Hit 2,576.5 EFLOPS in 2026
A joint IDC and Inspur Information report projects China's intelligent computing power will reach 2,576.5 EFLOPS in 2026, up 87.9% year on year, and 10,524.3 EFLOPS by 2030. Beijing, Hangzhou, Shenzhen, Shanghai and Guangzhou lead the city rankings, which for the first time weigh agent deployment and inference infrastructure rather than data center counts alone.
Report on the Development of Generative AI Applications (2026)" Released: China's Generative AI Users Top 700 Million, Intelligent Computing Power Reaches 2,185 EFLOPS
A new CNNIC report says China's generative AI user base passed 700 million in the first half of 2026, while intelligent computing power reached 2,185 EFLOPS, up 177% year over year. The report also flags the launch of the country's first all-domestic 100,000-chip AI supercluster and 250.14 billion yuan of AI funding in H1.
Chinese Academy of Sciences Launches Frontier Science Breakthrough Program With First 10 Projects
The Chinese Academy of Sciences launched its Frontier Science Breakthrough Program in Beijing on September 29, with the first 10 projects focused on the "four extremes": new element synthesis, dark matter and dark energy, microscopic measurement of chemical bonds, and artificially synthesized cells. The program pairs shared major research facilities with AI platforms and pilots evaluation rules that de-emphasize paper counts and short-term output.
China's Generative AI Users Pass 700 Million as Intelligent Computing Hits 2,185 EFLOPS
A CNNIC report says China's generative AI users topped 700 million in the first half of 2026, with intelligent computing power reaching 2,185 EFLOPS, up 177% year over year. It also says the country's first fully domestic 100,000-accelerator AI supercluster is now running.
WorldSkills Shanghai: Skilled Workers Say Adaptability Still Beats AI
At the 48th WorldSkills Competition in Shanghai, 1,385 competitors from 68 countries and regions are competing across 64 skill categories as AI reshapes vocational work. Competitors and delegation heads argue human adaptability, judgment and care work remain hard to automate, with lifelong learning the key skill going forward.
Huawei Upgrades Xinghe AI Data Center Network Solution at HUAWEI CONNECT 2026
At HUAWEI CONNECT 2026, Huawei announced a comprehensive upgrade to its Xinghe AI Data Center Network solution, positioning it as a high-efficiency production network for AI compute centers. No article body was provided, so specifics of the upgrade remain unconfirmed.
Two Ministries Explain the 15th Five-Year Plan for the Electronic Information Manufacturing Industry: Revenue at Above 30 Trillion Yuan by 2030
The Ministry of Industry and Information Technology and the National Development and Reform Commission have jointly issued the 15th Five-Year Plan for the Development of the Electronic Information Manufacturing Industry. The plan calls for annual revenue at above 30 trillion yuan for large-scale electronic information manufacturing companies by 2030, R&D spending to reach 3.5% of revenue, and 17 key initiatives spanning integrated circuits, advanced computing, AI chips and terminals, among other areas.
Yigeyun Raises Nearly 100 Million Yuan in Series B+ Funding to Advance Enterprise AI Agent Governance Platform
Yigeyun announced on September 14 that it had raised nearly 100 million yuan in a Series B+ round co-invested by MTR Lab and Northern Light Venture Capital. The funds will support further development of its enterprise governance platform for employees and AI agents, as well as overseas expansion. The company aims to manage enterprise agents through asset discovery, permission controls, identity assignment, behavioral oversight and cost measurement.
Beijing-Tianjin-Hebei Releases AI Computing Supply and Demand Lists, Accelerating an AI Data Center Corridor Around the Capital
The 2026 China Computing Power Conference runs from September 11 to 13 at the Beijing-Tianjin-Hebei Big Data Innovation and Application Center in Langfang. As part of the program, the Beijing-Tianjin-Hebei Forum on AI Computing and Advanced Manufacturing was held on September 11. The forum saw the official release of the region's AI computing supply and demand lists, designed to connect buyers and sellers of computing power directly and to speed up the convergence of AI computing with advanced manufacturing across the three jurisdictions.
Rack Supply Tightens at China’s Western Computing Hubs
Data center construction is accelerating in Qingyang and Zhongwei as operators seek to meet strong demand for high-density racks. Builders still face thin leasing margins, rising infrastructure costs and estimated investment payback periods of eight to 10 years.
China Issues National Standard for Cold-Plate Data Center Cooling
China’s GB/T 48023—2026 standard establishes technical and testing requirements for cold-plate liquid cooling systems across data center design, deployment and operations. The standard is scheduled to take effect on February 1, 2027.
DeepSeek Pursues Maximum “Intelligence Density
DeepSeek is testing whether its new Flash model can replace Pro and will cut Flash prices on September 10. Following an earlier R1 upgrade, AI model companies are reshuffling their premium capabilities. DeepSeek is pursuing efficiency on several fronts, though the results remain to be seen.
China’s AI Compute Prices Rise as Demand Outpaces Hardware Supply
Some Chinese cloud and AI service providers have reportedly raised compute prices by more than 30% as inference demand grows and supplies of high-end chips and memory remain constrained. The effects are spreading across data centers, hardware suppliers, compute lessors and AI application developers.
Tencent Opens WorkBuddy Agent Platform to Developers and Hardware Makers
Tencent has launched the WorkBuddy open agent platform for developers, application providers and hardware manufacturers. The company says the ecosystem has more than 100 initial partners, nine co-branded hardware products and over 30 applications across more than 20 fields.
H3C Details Open Supernode and High-Speed Interconnect Plans at ODX 2026
H3C showcased supernode interfaces, high-speed interconnects, 800V HVDC power, liquid cooling and AI data center networking solutions at ODX 2026 in Beijing. Several performance figures and interoperability capabilities remain company claims or future development targets pending real-world validation.
Envicool Showcases Full-Rack Liquid Cooling Solution at ODX 2026, Covering Cold Plates, Piping and Cooling Systems
Envicool showcased a full-chain liquid cooling portfolio for high-density AI servers and data centers at Beijing’s ODX 2026, including cold plates, connectors, manifolds and prefabricated piping.
China's Compute-Leasing Boom Faces a Reality Check in 2026 Interim Results
A midyear review examines the performance of major players in China’s AI compute-rental market. The article body is unavailable, so specific companies, figures and results could not be assessed.
China's Chip-Equipment Makers Push Deeper Into Advanced Production
Chinese semiconductor-equipment makers are expanding from etching into deposition, cleaning, polishing and process-control tools as chipmakers seek more domestic suppliers. Progress remains uneven, especially in advanced lithography, but growing fab investment and AI demand are creating new opportunities for local equipment companies.
China AI Data Centers Move Toward Gigawatt Scale
China’s AI data center buildout is moving into gigawatt-scale projects, led by major hubs such as Ulanqab and Zhongwei. The article says demand is splitting into tiers, with large training campuses for national and internet companies and smaller local facilities for inference and everyday use.
Musk’s SpaceX Has Yet to Send Data Centers Into Space. China’s GobiX Has Already Landed in the Gobi
By Ma Shiqing
The Last Line of Defense for the Tech Bull Market Is Starting to Give Way
Capital spending was raised, yet the stock fell.
Google Says Gemini Prompt Uses About 9 Seconds of TV Power, but Experts Dispute the Math
Google published new figures on the energy, water and carbon footprint of a median Gemini text prompt, saying it uses 0.24 watt-hours of electricity and 0.26 milliliters of water. Experts say the report is incomplete, citing missing indirect water use, market-based carbon accounting and limited methodological detail.
The World’s Largest Data Center Falls Through
This article is from the WeChat public account Fenghuang Finance, by Fengbaoyan.
How Did AI Become a Business About Steel, Power and Land?
This article is from the WeChat public account Wang Zhiyuan, by Wang Zhiyuan. Original title: “AI Faith Has Turned Into FOMO, and Capital’s Patience Is Now on the Clock.” Header image: AI-generated.
Tokens Go Global, Selling China’s Electricity to the World
This article is from the WeChat public account Deep Tide TechFlow, written by Black Lobster. The header image is from Visual China Group.
Moonshot AI Denies Kimi K3 Is a Distilled Copy as Viral Model Demand Surges
Moonshot AI says Kimi K3 is based on original architectural work, not distillation, after the model’s strong benchmark results and rapid user demand drew praise from Elon Musk and scrutiny from rivals. The company also signaled that full weights for its open-source path are planned for release before July 27, 2026.