[SMM Computing Power News] A drone company has disclosed that AI model training is mostly outsourced to pod equipment suppliers, and its demand for self-built computing power is limited.

Published: Jun 26, 2026 15:51
A drone company employee stated that the training of its drone AI image recognition models is primarily handled by pod equipment suppliers, while drone manufacturers directly integrate pre-trained model APIs without investing their own computing resources. Previously, SMM Telecom learned that some enterprises use consumer-grade GPUs such as 4090 and 5090 for model training; short-term training needs are met through server rentals or direct token purchases. The person mentioned a pod company in Chengdu as a representative enterprise in the industry, responsible for model training and equipment integration. Currently, AI training computing power in the small and medium-sized drone sector is predominantly outsourced, and the demand for self-built computing power on the drone manufacturer side is not yet significant.

Data Source Statement: Except for publicly available information, all other data are processed by SMM based on publicly available information, market communication, and relying on SMM's internal database model. They are for reference only and do not constitute decision-making recommendations.

For any inquiries or for more information, please contact: lemonzhao@smm.cn
For more information on how to access our research reports, please contact:service.en@smm.cn
Related News
[Semiconductor demand robust, South Korea's June exports expected to see stronger growth]
11 mins ago
[Semiconductor demand robust, South Korea's June exports expected to see stronger growth]
Read More
[Semiconductor demand robust, South Korea's June exports expected to see stronger growth]
[Semiconductor demand robust, South Korea's June exports expected to see stronger growth]
Thanks to robust semiconductor demand, South Korea’s June exports are expected to post even stronger growth, with the monthly trade surplus likely to surpass $30 billion for the first time. According to a median forecast from a Wall Street Journal survey of nine economists, exports are expected to grow 57.3% YoY in June, up from a revised 53.4% in May, while imports are projected to rise 23.5% YoY, driving the trade surplus for the month to $32 billion. Seunghoon Stephen Lee, chief economist at Meritz Securities, said chips and computers are seen as the key drivers of export growth, with the pace of export growth expected to remain above 50% in H2.
11 mins ago
[Institution: DDR2 Contract Price Estimated to Rise Approximately 55-60% in Q2, with a Further Increase of 35-40% in Q3]
12 mins ago
[Institution: DDR2 Contract Price Estimated to Rise Approximately 55-60% in Q2, with a Further Increase of 35-40% in Q3]
Read More
[Institution: DDR2 Contract Price Estimated to Rise Approximately 55-60% in Q2, with a Further Increase of 35-40% in Q3]
[Institution: DDR2 Contract Price Estimated to Rise Approximately 55-60% in Q2, with a Further Increase of 35-40% in Q3]
According to the latest research by TrendForce, due to the structural tightness of mature-node DRAM supply, Consumer DRAM buyers are being forced to adopt older-generation products to secure larger DRAM supply quotas, triggering a new wave of recent procurement demand for older-generation Consumer DRAM chips in the industry. This has caused the contract prices of Consumer DRAM chips of generations including DDR2 and DDR3 to extend the upward momentum from the first quarter of 2026. It is estimated that DDR2 contract prices will increase by approximately 55–60% in the second quarter, with a further 35–40% increase projected for the third quarter.
12 mins ago
[7 National Standards from the “Artificial Intelligence Agent Interconnection” Series Published]
13 mins ago
[7 National Standards from the “Artificial Intelligence Agent Interconnection” Series Published]
Read More
[7 National Standards from the “Artificial Intelligence Agent Interconnection” Series Published]
[7 National Standards from the “Artificial Intelligence Agent Interconnection” Series Published]
It was learned from a press conference held by the State Administration for Market Regulation that the national standards series on *Artificial Intelligence Agent Interconnection* has been officially released. The seven national standards in this series comprehensively cover core areas such as overall architecture, identity codes, identity management, agent description, agent discovery, agent interaction, and agent tool invocation, systematically building a full-coverage, closed-loop standards and specifications system spanning "identity identification — capability description — supply-demand discovery — collaborative interaction — tool invocation." This effectively fills the standards gap in this field. With the unified architecture and interaction rules enabled by these standards, enterprises can reuse standard components, reduce custom development, and shorten product time-to-market; simultaneously, the establishment of unified identity authentication and end-to-end traceability mechanisms reinforces the institutional foundation for cross-domain trustworthy and secure interactions.
13 mins ago