Orient: AI4S is empowering multiple disciplines and industries, from foundational models and computing power to practical applications in industries.

Orient's research report states that AI4S has currently covered fields such as life sciences, earth sciences, and energy materials, but the industry's penetration rate is relatively low. In the future, it is expected to enter a new stage of scaled value realization. AI4S is empowering multiple disciplines and industries, including biopharmaceuticals, materials energy, industrial simulation, meteorological environments, and agricultural breeding. The rapid development and iterative progress of AI4S have brought investment opportunities in scientific modeling, computational infrastructure construction and operation, as well as industrial applications. Orient's main points are as follows: AI4S is becoming the "next stop" for the fifth paradigm of scientific research and large model iteration. AI4S refers to a new model in which artificial intelligence aids or independently completes scientific research tasks, capable of reshaping the entire research system and breaking traditional bottlenecks. It is regarded as the fifth paradigm of research following experiments, theory, computation, and data science. In recent years, AI4S has driven accelerated innovation in numerous scientific research fields, and AlphaFold winning the Nobel Prize has further validated its hard-core scientific value. As a result, major global economies such as the US, Europe, and the UK have elevated AI4S to a national strategy, intensively introducing policies to bolster its layout. On the other hand, as general large model capabilities continue to enhance, finding higher value implementation scenarios and addressing more significant problems has also become the direction for model companies. Due to the stringent technical demands of scientific research scenarios and the high returns on output value, they have become a core choice for model companies. Currently, AI4S has covered areas such as life sciences, earth sciences, and energy materials, but the industry penetration rate remains low and is expected to enter a new phase of scaled value realization in the future. AI4S models are undergoing continuous iteration, moving from single scenario models to an integrated architecture of specialization and generalization. Compared to general large models, AI4S models possess three exclusive core capabilities: scientific mechanism constraints, unified representation of multiple modalities, and long-term sequential reasoning, making them better suited to the real-world laws of scientific research and complex scientific tasks. AI4S models are undergoing continuous iteration, evolving from early domain-specific models and tool-enhanced models to a mainstream architecture that integrates scientific foundational models with domain-specific models. Domestic institutions are continuously breaking through and launching integrated model architectures such as S1-Omni and SAGE, along with intelligent research platforms, solving the imbalance between generalization and specialization of single models, and promoting AI4S from isolated scientific computing to a full-process, closed-loop scientific intelligent innovation. AI4S requires a new type of heterogeneous computing power architecture that deeply integrates supercomputing and intelligent computing. The dual characteristics of AI4S, which combine mechanistic simulation with AI model coupling, raise new demands for computing power, necessitating mixed-precision computing, long-term stable operations, and a dual-ecosystem software stack. Traditional supercomputing and general intelligent computing clusters are unable to meet these requirements. Therefore, the industry proposes a new architecture for heterogeneous computing power that deeply integrates supercomputing and intelligent computing to achieve comprehensive compatibility in computing power precision, hardware interconnection, storage, and software ecology. The leading domestic computing power company, Dawning Information Industry, has launched a domestically produced Shuguang 8000 cluster with over 100,000 cores, creating a fully self-developed foundation for supercomputing-intelligent computing integrated AI4S computing power. This platform has enabled world-class scientific breakthroughs such as trillion-atom simulations and accelerated protein folding, pushing domestically produced high-end computing power from demonstration applications towards large-scale replication. AI4S is blooming in multiple fields, bridging the gap between basic research and industrial applications. AI4S is empowering multiple disciplines and industries, including biopharmaceuticals, materials energy, industrial simulation, meteorological environments, and agricultural breeding. AI in drug development covers the entire process from target discovery and molecular generation to clinical trials, significantly compressing research and development cycles and costs. In the industrial sector, a "data + physics" dual-driven simulation paradigm has emerged, revolutionizing CAD/CAE/EDA toolchains. In the meteorological field, AI models have achieved operational forecasting, enabling power predictions for new energy sources. Agricultural breeding leverages multi-omics data analysis to help shorten the breeding cycle for quality varieties. AI4S is gradually moving beyond a tool-assisted approach, promoting a new research and development model across industries that prioritizes prediction over trial-and-error experiments, bridging the complete chain from scientific discovery to engineering commercialization. Risk warnings: Risks of dynamic adjustments in regulatory policies; risks of technological development and conversion falling short of expectations; risks associated with increasing competition in various sectors.