Distant long-period comet C/2017 K2 has been outside the planetary region of the solar system for 3 Myr, negating the possibility that heat retained from the previous perihelion could be responsible for its activity. This inbound comet is also too cold for water ice to sublimate and too cold for amorphous water ice, if present, to crystallize. C/2017 K2 thus presents an ideal target in which to investigate the mechanisms responsible for activity in distant comets. We have used Hubble Space Telescope to study the comet in the pre-perihelion distance range 13.8 to 15.9 AU. The coma maintains a logarithmic surface brightness gradient $m = -1.010pm$0.004, consistent with steady-state mass loss. The absence of a radiation pressure swept tail indicates that the effective particle size is large (0.1 mm) and the mass loss rate is $sim$200 kg s$^{-1}$, remarkable for a comet still beyond the orbit of Saturn. Extrapolation of the photometry indicates that activity began in 2012.1, at 25.9$pm$0.9 AU, where the blackbody temperature is only 55 K. This large distance and low temperature suggest that cometary activity is driven by the sublimation of a super-volatile ice (e.g.~CO), presumably preserved by K2s long-term residence in the Oort cloud. The mass loss rate can be sustained by CO sublimation from an area $lesssim 2$ km$^2$, if located near the hot sub-solar point on the nucleus. However, while the drag force from sublimated CO is sufficient to lift millimeter sized particles against the gravity of the cometary nucleus, it is 10$^2$ to 10$^3$ times too small to eject these particles against inter-particle cohesion. Our observations thus require either a new understanding of the physics of inter-particle cohesion or the introduction of another mechanism to drive distant cometary mass loss. We suggest thermal fracture and electrostatic supercharging in this context.