One ton of refrigeration is exactly 3.5168528 kilowatts. Despite the name it is not a mass at all — it is a rate of cooling, defined by how much heat is absorbed melting a short ton of ice in a day.
About the units
Before mechanical refrigeration, cooling was sold as ice. When compressors arrived, their capacity was naturally described by how much ice they replaced: a machine rated at one ton could do the work of a ton of ice melting over 24 hours. The latent heat of fusion of water is 333.55 kJ/kg, so a short ton of 907.18474 kg absorbs 302.6 megajoules, and spread across 86,400 seconds that is about 3,502 watts. The defined value is slightly higher, 3,516.85 W, because the definition was fixed using the older figure of 144 BTU per pound of ice — 334.94 kJ/kg — rather than the modern measurement. The unit is therefore a definition, not a result you can reach from the constants in use now. The unit is standard in American commercial HVAC and in much of Asia, while European practice uses kilowatts directly.
The exact factor
The exact value is 3.5168528 kW, equal to 12,000 BTU per hour. That last equivalence is the one worth memorising, because it links the two American cooling units: a 3-ton system is 36,000 BTU/h is 10.55 kW. Residential air conditioners run from 1.5 to 5 tons, a small office system 10 to 25 tons, and a large commercial chiller several hundred. Note that the ton of refrigeration describes cooling output, not electrical input — a 3-ton unit with a coefficient of performance of 4 draws about 2.6 kW of electricity.
Where you meet this conversion
Commercial HVAC specification and chiller selection, particularly in the United States, the Middle East and parts of Asia. Data centre cooling capacity is often quoted in tons alongside the IT load in kilowatts, and the two have to be brought into the same unit before a design can be checked. Cold storage and process cooling use it. Anyone reading an American HVAC drawing against a European equipment catalogue needs both this factor and the BTU/h relationship.