The June 1912 eruption of Novarupta filled nearby glacial valleys on the Alaska Peninsula with ash-flow tuff (ignimbrite), and post-eruption observations of thousands of steaming fumaroles led to the name ‘Valley of Ten Thousand Smokes’ (VTTS). By the late 1980s most fumarolic activity had ceased, but the discovery of thermal springs in mid-valley in 1987 suggested continued cooling of the ash-flow sheet. Data collected at the mid-valley springs between 1987 and 2001 show a statistically significant correlation between maximum observed chloride (Cl) concentration and temperature. These data also show a statistically significant decline in the maximum Cl concentration. The observed variation in stream chemistry across the sheet strongly implies that most solutes, including Cl, originate within the area of the VTTS occupied by the 1912 deposits. Numerous measurements of Cl flux in the Ukak River just below the ash-flow sheet suggest an ongoing heat loss of ∼250 MW. This represents one of the largest hydrothermal heat discharges in North America. Other hydrothermal discharges of comparable magnitude are related to heat obtained from silicic magma bodies at depth, and are quasi-steady on a multidecadal time scale. However, the VTTS hydrothermal flux is not obviously related to a magma body and is clearly declining. Available data provide reasonable boundary and initial conditions for simple transient modeling. Both an analytical, conduction-only model and a numerical model predict large rates of heat loss from the sheet 90 years after deposition.