Contrary to the theoretical foundations of the asset-growth factor in both the five-factor model of Fama and French (2015) and the q-factor model of Hou, Xue, and Zhang (2015), I find that investment level is not the reason why asset growth is negatively associated with future stock returns. Instead, the source is investment structure – the division of total investments between those on and off the balance sheet. In particular, (1) asset growth (∆A/LagA) is not associated with future stock returns among ∆A < 0 firms, and (2) decomposing asset growth of ∆A > 0 firms into two components − investment structure (∆A/TINV, increase in assets divided by total investments) and investment level (TINV/LagA, total investments divided by one-year-lagged assets) − reveals that the component negatively associated with future stock returns is not investment level but rather investment structure. Furthermore, the associations of asset growth with both systematic risk and fundamental uncertainty exhibit similar patterns to its association with future stock returns, as described in (1) and (2).