Core Viewpoint - The marine sector is undergoing a transformative shift due to the integration of high-performance computing and artificial intelligence, enhancing marine safety, resource development, and environmental governance [1] Group 1: Industry Challenges - The marine field faces inherent barriers in technology integration, as AI relies on mathematical logic while marine science emphasizes precise physical phenomena [1] - The demand for computational power in marine big models for forecasting and environmental analysis is increasing, necessitating tailored computational service models [1][3] Group 2: MaaS Solution - The MaaS (Model as a Service) solution by the company addresses industry pain points by standardizing AI models into cloud services, allowing users to access models via API without the need for training or maintenance [2] - The MaaS platform integrates various leading models to support marine forecasting, with expectations for the deployment of numerous specialized big models in the future [2] Group 3: Computational Support - The company has established a robust computational network with over 45 intelligent computing centers and 15 supercomputing centers, boasting a total scheduling capacity exceeding 2 million CPU cores and over 50,000 GPUs [3] - Customized computational solutions are provided based on the specific characteristics of different marine scenarios, ensuring optimal cost-effectiveness for users [3] Group 4: Market Growth Potential - The market potential for MaaS in the marine sector is significant, with IDC predicting that the volume of public cloud big model calls in China will reach 536.7 trillion tokens by the first half of 2025 [4] - The integration of AI is expected to drive a high growth rate in the marine MaaS business over the next 3-5 years, emphasizing cost reduction and efficiency improvement [4] Group 5: Future Developments - The company plans to enhance the ecological architecture of its new computational base, optimizing resource scheduling and compliance detection across various layers [5] - A hybrid deployment model combining public and private clouds will be developed to ensure effectiveness, cost optimization, and data security [5]
并行科技李津宇： MaaS算力形态为海洋人工智能多元发展提供新范式