Empirically observed intertemporal choices about money have long been thought to exhibit present bias, i.e. higher short-term compared to long-term discount rates. Recently, this view has been called into question on both empirical and theoretical grounds, and a spate of recent findings suggest that present bias for money is minimal or non-existent when one allows for curvature in the utility function and transaction costs are tightly controlled. However, an alternative interpretation of many of these findings is that, in the interest of equalizing transaction costs across earlier and later payments, small delays were introduced between the time of the experiment and the soonest payment. As such, these tests of present-bias may not be sufficiently powerful. We conduct a laboratory experiment in Kenya in which we elicit time and risk preference parameters from 291 participants, using convex time budgets and tightly controlling for transaction costs. Importantly, we make the soonest payments truly immediate, using the Kenyan mobile money system M-Pesa to make real-time transfers to subjects’ phones. We find strong evidence of present bias, with estimates of the present bias parameter ranging from 0.901 to 0.937. This result suggests that present bias for money does in fact exist, but only for truly immediate payments.