Stream temperature science and management is rapidly shifting from single-metric driven approaches to multi-metric, thermal regime characterizations of streamscapes. Given considerable investments in recovery of cold-water fisheries (e.g., Pacific salmon and other declining native species), understanding where cold water is likely to persist, and how cold-water thermal regimes vary, is critical for conservation. California's unique position at the southern end of cold-water ecosystems in the northern hemisphere, variable geography and hydrology, and extensive flow regulation requires a systematic approach to thermal regime classification. We used publicly available, long-term (> 8 years) stream temperature data from 77 sites across California to model their thermal regimes, calculate three temperature metrics, and use the metrics to classify each regime with an agglomerative nesting algorithm. Then, we assessed the variation in each class and considered underlying physical or anthropogenic factors that could explain differences between classes. Finally, we considered how different classes might fit existing criteria for cool- or cold-water thermal regimes, and how those differences complicate efforts to manage stream temperature through regulation. Our results demonstrate that cool- and cold-water thermal regimes vary spatially across California. Several salient findings emerge from this study. Groundwater-dominated streams are a ubiquitous, but as yet, poorly explored class of thermal regimes. Further, flow regulation below dams imposes serial discontinuities, including artificial thermal regimes on downstream ecosystems. Finally, and contrary to what is often assumed, California reservoirs do not contain sufficient cold-water storage to replicate desirable, reach-scale thermal regimes. While barriers to cold-water conservation are considerable and the trajectory of cold-water species towards extinction is dire, protecting reaches that demonstrate resilience to climate warming remains worthwhile.