Background and aims The positive effects of cereal-legume intercropping can occur through complementarity and facilitation for the acquisition of soil nitrogen (N) and phosphorus (P). However, the magnitude of the benefits depends on the availability of such resources. To improve our understanding and design best-performing systems, we evaluated the performance and underlying mechanisms of a cereal-legume mixture on two crossed gradients of N and P. Methods In a pot experiment, wheat (Triticum aestivum L.) and white lupin (Lupinus albus L.) were grown separately or together on four levels of N combined with four levels of P. We compared biomass, N content, P content, shoot-to-root biomass ratios and white lupin production of cluster roots and nodules of sole crops with those of intercrops. We used the relative interaction intensity index to determine how the nature of the interaction between the two species changed along the two resource gradients. Results Regardless of N and P supply, combined biomass and N content of intercropped wheat and white lupin were 29% and 23% greater than those of sole crops, respectively. Intercropping increased P content by 34%, on average, particularly at the lowest P supply. While wheat benefited from intercropping in all treatments, white lupin's performances decreased with increasing N and P supply, resulting in a shift from mutualism to competition along the two gradients and compensation mechanisms between the two species. Conclusion Nutrient availability negatively influenced the nature and mechanisms of wheat-lupin interaction. However, we only observed mutualistic interactions under low resource supply, which led to the lowest production, while competition was associated with the most productive situations.