Low academic achievement is common and associated with poorer social and employment outcomes. Working memory (the ability to temporarily store and manipulate information) is vital for general academic skills. Cognitive training can improve working memory, but the translation to longer-term academic outcomes and cost-effectiveness of intervention is unknown. We aim to determine the cost-effectiveness of providing a computerized working memory training program to children with low working memory in Grade 1 (aged 6-7 years), compared with receiving usual classroom teaching, in terms of academic achievement 1 and 2 years later. Methods Economic evaluation of Memory Maestros, a randomised controlled trial run across 44 primary schools in Victoria, Australia from 2012 to 2015. This population-based selective prevention trial assessed the effects of a computerized working memory training program on working memory, academic outcomes and costs of health and education service use. Costs were estimated using resource use data from trial records and parent-report. Primary outcome was academic achievement (WRAT-4) at 2 years. Secondary outcomes included child quality of life (CHU-9D) and national school test scores (NAPLAN). Results Of 1723 children screened, 226 were randomised to each arm, with 90% retention at 1 year and 88% at 2 years. 90.3% of children in the intervention arm received the program in full. Visuo-spatial short-term memory was improved (effect size at 12 months, 0.49 [95% CI 0.28-0.70]) but there were no benefits to academic outcomes. Intervention costs were A$1035 per child. Conclusions This working memory intervention was not cost-effective in improving academic outcomes. Keywords: Child Health; Evaluation; Health Economics Key messages 1. Previous studies suggest adaptive working memory training interventions can improve working memory outcomes for children at risk of poor academic achievement. 2. Our population-level randomized controlled trial found that despite there being short term benefits on some domains of working memory, there was no transfer benefit for academic outcomes at 1 and 2 years post-randomisation. 3. The working memory intervention had significant cost and was not cost-effective.