Abstract: Agricultural big data is one of the most important driving elements in smart agriculture. Its service platforms can serve as the primary carriers of big data. This study aims to review agricultural big data and service platforms of crop production as representative scenario. (1) The key technical architecture of these platforms was summarized to classify five modules: a) Data acquisition and transmission with Internet of Things (IoT) sensors, unmanned aerial vehicles (UAVs), and satellite remote sensing; b) Data storage and preprocessing for massive throughput and multi-source heterogeneity; c) Data security and privacy protection using federated learning and cryptographic techniques; d) Data mining and modeling using machine learning and deep learning algorithms; e) and data visualization and sharing providing multi-stakeholder access and interoperability. A single mode evolved into differentiated systems for perception, data transmission, computation, and decision-making tailored to various agricultural scenarios, such as field, greenhouse, and orchard. For instance, field applications prioritize large-scale monitoring over extensive areas and medium-to-long-term early warning; greenhouses focus on minute-level real-time control in enclosed spaces; While orchards emphasize precision management of individual fruit trees. (2) Current situations were reviewed. Developed countries in Europe and North America also established an integrated equipment–data–service ecosystem for data-driven services throughout the entire agricultural production. Whereas, in China, despite the progress made in multiple provinces, the challenges remained, including rural infrastructure, weak independent innovation in key technologies, fragmented data resources between different institutions, and a shortage of interdisciplinary talent. (3) According to current research status, the differences between agricultural big data platforms were proposed in six dimensions: a) In terms of federated sharing, AgReFed platform abroad adhered to FAIR principles for cross-institutional interoperability, while China Qilu Agricultural Cloud relied on an administrative framework in the entire region, but lacking in standards and severe data silos; b) Regarding to personalized services, New Zealand's AgYields system was mature and versatile, whereas Zhejiang's platform focused on specialty crops, but its mode was difficult to replicate; c) In multi-dimensional collaboration, Spain's Farmdata boasted a well-established five-party coordination mechanism, while Sinochem MAP integrated full-chain resources, but lacking in the participation from market entities; d) In precision farming, Germany's xarvio technology was maturely implemented with a complete commercial ecosystem, while unmanned farms in South China significantly increased yields, though mostly as demonstration projects; e) In intelligent decision-making, America's Climate FieldView established a closed-loop data-driven decision service, while Shandong's smart farms achieved in intelligent management, but heavily dependent on core technologies; f) In intelligent management, America's John Deere exhibited high equipment integration, and Jilin's "Jinnong Cloud" monitoring system was robust, but prioritized presentation over practical effectiveness. (4) Finally, the existing agricultural big data and service platforms were analyzed to propose recommendations from the perspectives of policy, technology, management, promotion, application, and talent cultivation. This finding can provide a strong reference to promote the iterative optimization of big data and service platforms in sustainable agriculture.