We address the long-term uranium supply from first principles, summarizing estimates of the abundance of uranium in the crust of the earth as a function of concentration and accessibility. Defining the supply curve as a functional relationship between the cumulative quantity of uranium extracted and the cost of extracting the next unit of uranium, we note that a supply curve requires a crustal abundance model plus a correlation between ore grade and extraction cost. Surveying a number of supply curves that appear in the literature, we observe that while estimates vary widely (we observe an order of magnitude difference in forecasts of the quantity of uranium available at $100/kg U or less), they generally reflect expectations that uranium availability will be significantly greater than the Red Book numbers imply. Furthermore, by comparison with historical data for more than 40 other minerals, we show that the assumption of time invariance when formulating a supply curve is not borne out by experience. In fact, the price of most other minerals has decreased with time as well as with cumulative quantity extracted. Neither the Red Book nor the other supply curves we survey explicitly accounts for the unit-based technological learning that fosters this behavior.