Metal additive printing technology facilitates unprecedented freedom in the design of mechanical components and has mostly been used to improve their performance. However, the prediction of fabrication failure is equally important in metal additive manufacturing. To address this challenge, this paper presents an integrated strategy that demonstrates the process from design to fabrication of a cooler. This study adopted a multi-fidelity surrogate model and a supporting effect-based prediction algorithm to analyze the geometry of coolers for additive manufacturing. The original geometry was modified based on the prediction results and fabricated using the laser powder bed fusion method. The internal space of the modified cooler design was verified using computed tomography scanning, which revealed no defects. Additionally, the fabricated cooler demonstrated a notable improvement, exceeding 60%. This study demonstrated an effective approach for designing and fabricating metal-additive-printed components, considering both functional requirements and fabrication constraints, which can be widely applied in various fields where performance and reliability are paramount.