City-level CO2 emission accounting is necessary to identify the different energy circumstances among all cities. However, due to a lack of data, energy consumption and emission statistics are not well documented. Focusing on the industrial production using fossil fuels, our work provides the first detailed city-level estimation of production-based sectoral CO2 emissions in the Xinjiang Uygur Autonomous Region. In 2010, 15 cities in Xinjiang emitted a total of 304.06 million tonnes CO2, and 97.7% of those emissions were related to fossil fuel combustion (i.e. energy-related emissions), with the remaining 2.3% from industrial processes associated with the production of cement (i.e. process-related emissions). The consumption of raw coal and crude oil were the main sources of Xinjiang’s emissions (50.3% and 23.0%, respectively), whereas ‘petroleum processing and coking’ and ‘power and heat production’ were the two sectors that contributed the largest emissions at 32.6% and 27.9%, respectively. The cities in Xinjiang presented considerable variations in the total CO2 emissions and emissions per capita. The emissions intensity and emissions per capita shared similar distributions, and the emissions are significantly spatial autocorrelated. Cities whose economies relied on emission-intensive pillar industries and/or energy mainly sourced from raw coal tended to have high emissions per capita and high emissions intensities. Those cities included Altay, Changji, Hami and Shihezi. We also examined the time-series emissions of Urumqi, the largest city, from 2005 to 2015. Urumqi presented a generally rising trend in CO2 emissions over the decade, with emissions increasing by 324.2%. The major driving sector was ‘power and heat’, which showed increases in the total CO2 emissions and percentage of Xinjiang’s emissions. Based on the findings, policy recommendations for emission reductions and low-carbon development for the cities in Xinjiang are provided, including adjusting the energy structure and introducing multiple industries.