Peatlands contain a large belowground carbon (C) stock in the biosphere, and their dynamics have important implications for the global carbon cycle. However, there are still large uncertainties in C stock estimates and poor understanding of C dynamics across timescales. Here I review different approaches and associated uncertainties of C stock estimates in the literature, and on the basis of the literature review my best estimate of C stocks and uncertainty is 500 &plusmn; 100 (approximate range) gigatons of C (Gt C) in northern peatlands. The greatest source of uncertainty for all the approaches is the lack or insufficient representation of data, including depth, bulk density and carbon accumulation data, especially from the world's large peatlands. Several ways to improve estimates of peat carbon stocks are also discussed in this paper, including the estimates of C stocks by regions and further utilizations of widely available basal peat ages. <br><br> Changes in peatland carbon stocks over time, estimated using <i>Sphagnum</i> (peat moss) spore data and down-core peat accumulation records, show different patterns during the Holocene, and I argue that spore-based approach underestimates the abundance of peatlands in their early histories. Considering long-term peat decomposition using peat accumulation data allows estimates of net carbon sequestration rates by peatlands, or net (ecosystem) carbon balance (NECB), which indicates more than half of peat carbon (> 270 Gt C) was sequestrated before 7000 yr ago during the Holocene. Contemporary carbon flux studies at 5 peatland sites show much larger NECB during the last decade (32 &plusmn; 7.8 (S.E.) g C m⁻² yr^&ndash;1) than during the last 7000 yr (&sim; 11 g C m⁻² yr^&ndash;1), as modeled from peat records across northern peatlands. This discrepancy highlights the urgent need for carbon accumulation data and process understanding, especially at decadal and centennial timescales, that would bridge current knowledge gaps and facilitate comparisons of NECB across all timescales.