The integration of Virtual Reality (VR) technology in education is increasingly relevant in the emerging metaverse, a digital landscape that blurs the boundaries between physical and virtual interactions. While VR is well-received in STEM education, its incorporation into K–12 pedagogical frameworks and its impact on learning performance remain underexplored. By applying Cognitive Load Theory (CLT), which aligns with working memory management, problem-solving schema development, and instructional design—we specifically investigate how 2D and VR modalities shape intrinsic, extraneous, and germane cognitive load, revealing that VR fosters higher engagement through immersive interactions, while 2D remains more intuitive for first-time users. Using discourse analysis, we examined cognitive load in 12 students (Grades 6–8) engaging in structured learning tasks in an Indian K–12 STEM context, collecting data through transcripts, observations, think-aloud protocols, and pre/post-task interviews. Findings reveal that intrinsic load was higher in VR due to unfamiliarity but was offset by motivation as learners acclimated; germane load increased in VR owing to enhanced schema integration, whereas extraneous load surfaced from interface limitations in 2D and interaction challenges in VR. By integrating cognitive science with immersive learning design, this study highlights the importance of optimizing instructional strategies in both traditional and emerging environments to balance cognitive demands and improve learning outcomes in the evolving metaverse.