Since Bridgman's discovery of five solid water (H2O) ice phases1 in 1912, studies on the extraordinary polymorphism of H2O have documented more than seventeen crystalline and several amorphous ice structures2,3, as well as rich metastability and kinetic effects4,5. This unique behaviour is due in part to the geometrical frustration of the weak intermolecular hydrogen bonds and the sizeable quantum motion of the light hydrogen ions (protons). Particularly intriguing is the prediction that H2O becomes superionic6-12-with liquid-like protons diffusing through the solid lattice of oxygen-when subjected to extreme pressures exceeding 100 gigapascals and high temperatures above 2,000 kelvin. Numerical simulations suggest that the characteristic diffusion of the protons through the empty sites of the oxygen solid lattice (1) gives rise to a surprisingly high ionic conductivity above 100 Siemens per centimetre, that is, almost as high as typical metallic (electronic) conductivity, (2) greatly increases the ice melting temperature7-13 to several thousand kelvin, and (3) favours new ice structures with a close-packed oxygen lattice13-15. Because confining such hot and dense H2O in the laboratory is extremely challenging, experimental data are scarce. Recent optical measurements along the Hugoniot curve (locus of shock states) of water ice VII showed evidence of superionic conduction and thermodynamic signatures for melting16, but did not confirm the microscopic structure of superionic ice. Here we use laser-driven shockwaves to simultaneously compress and heat liquid water samples to 100-400 gigapascals and 2,000-3,000 kelvin. In situ X-ray diffraction measurements show that under these conditions, water solidifies within a few nanoseconds into nanometre-sized ice grains that exhibit unambiguous evidence for the crystalline oxygen lattice of superionic water ice. The X-ray diffraction data also allow us to document the compressibility of ice at these extreme conditions and a temperature- and pressure-induced phase transformation from a body-centred-cubic ice phase (probably ice X) to a novel face-centred-cubic, superionic ice phase, which we name ice XVIII2,17.