The core idea behind government debt is to fund fiscal deficit, which is anticipated to drive economic investments. To a larger extent, this is not the case, as evidenced in the Nigerian context, where debt has risen so high with investment levels declining, thus questioning the government's ability to manage and sustain its debt to pursue vital investment needs. This study aimed to investigate the threshold effect of debt sustainability on investments amidst macroeconomic swings from 1981 to 2020. In this regard, the threshold autoregressive regression (TAR) was used because it gave information on the optimal threshold of debt sustainability that would attract investments. Also, the Granger causality test was carried out to show the direction of causality among the variables. This paper concentrated on debt service to revenue and total debt stock to GDP as debt sustainability measures while investment was decomposed into public, private, and foreign investments. The paper yields that, based on the multivariate TAR analyses, the main threshold variables, that is, debt service to revenue and total debt stock to GDP, had a non-linear relationship with public, private, and foreign direct investments amidst changes in macroeconomic variables such as exchange rate, inflation, and monetary policy rate. The threshold coefficient of debt service to revenue indicated that public and foreign direct investments declined during low thresholds while private investment increased. However, the opposite prevailed when debt service to revenue exceeded the threshold values. However, the Granger causality test showed that debt service to revenue Granger caused total debt stock to GDP and exchange rate Granger caused debt service to revenue ratio, implying that exchange rate swings could affect the government's ability to service debt which in turn explains the non-linear relationship between debt sustainability and investments. Hence, it was concluded that Nigeria's lack of debt sustainability was associated with revenue generation, which explains why the models did not follow a linear path.