Cloud Engine-경제적인 AI GPU 렌탈 플랫폼

 A100 pcie40G - $ 200/card/month;BMS:8*Ascend 910B $ 2000/month  8*4090D - $ 800/month  8*4090 - $ 900/month  8*A100 pcie40G - $ 1400/month  8*A100 pcie80G - $ 3200/month  8*A100 nvlink80G - $ 3800/month  8*A800 nvlink80G - $ 3800/month  8*H20 - $ 4000/month  8*L20 - $ 1300/month  8*L40 - $ 1600/month  8*L40S - $ 2200/month  8*H100 - $ 8000/month  8*H200 - $ 9200/month  8*B200 - $ 13000/month 

Global GPU Computing News【20250518】

  1. Beijing Launches Subsidies for Domestic GPU Purchases
    Beijing introduced subsidies for enterprises purchasing domestically developed GPUs to build AI computing centers. Eligible companies can receive up to ¥50 million, contingent on achieving a localization rate of at least 60%.

  2. Huawei Ascend 910C Performance Breakthrough
    Huawei’s Ascend 910C chip demonstrated 82% of NVIDIA A100’s performance in tests. Its market share in Inner Mongolia’s Ulanqab computing hub surged from 10% to 58%, highlighting domestic substitution trends.

  3. Moore Threads Deploys Thousand-Card Clusters
    Moore Threads’ "Kuae" AI cluster solution has been deployed in three 1,000-GPU clusters, supporting training for trillion-parameter models. It claims 50% higher efficiency and CUDA ecosystem compatibility.

  4. China’s Telecom Giants Plan ¥70B Computing Investment
    China Mobile, China Unicom, and China Telecom announced a combined ¥70 billion investment in AI data centers for 2025, prioritizing regions like the Yangtze River Delta. Domestic server adoption is expected to exceed 80%.

  5. JD Cloud’s vGPU Platform Boosts Utilization by 70%
    JD Cloud released a vGPU resource pooling platform using fine-grained partitioning and heterogeneous management, improving GPU utilization by 70% for clients like financial institutions.

  6. National Computing Network Accelerates
    China’s National Data Bureau aims to establish a unified computing network by 2025, optimizing resource allocation across eastern, central, and western regions to enhance domestic GPU cluster efficiency.

  1. NVIDIA Launches Blackwell Ultra GPU
    NVIDIA unveiled its Blackwell Ultra chip, offering 14x performance gains over predecessors. Microsoft and Meta pre-ordered 3.6 million units, driving demand for optical modules.

  2. U.S. Tightens AI Chip Export Controls
    The Biden administration added China to its AI technology embargo list, restricting NVIDIA and others from exporting advanced GPUs. This policy accelerates domestic substitution efforts in China.

  3. Liquid Cooling Demand Soars
    ByteDance and Tencent reported 5x higher costs for liquid cooling retrofits. Telecom operators’ procurement of liquid cooling equipment surged 700% YoY, creating a ¥10B+ market.

  1. NVIDIA Partners with Chinese EV Makers
    At its developer conference, NVIDIA announced collaborations with BYD, GAC, and XPeng on autonomous driving. Its GB200 chip’s copper interconnect design also boosted related supply chain stocks.


Key Takeaways

This week’s GPU computing landscape reflects:

  • Domestic Focus: Policy-driven localization (subsidies, clusters) and infrastructure expansion (Huawei Ascend, Moore Threads).

  • Global Shifts: NVIDIA’s Blackwell launch and U.S. export controls intensify competition, while liquid cooling and optical modules see explosive demand.

  • Collaboration: JD Cloud and Tencent optimize efficiency via technical solutions, while NVIDIA deepens ties with automakers and hardware suppliers.

영업:+8619886543278(WhatsApp)
이메일:yqtxben@163.com
주소:중국 선전시 난산구 가든 시티 디지털 빌딩 A동 2층
연락처 정보