The number of satellites in non-geostationary orbits (NGSOs) is planned to grow dramatically, from about a thousand in orbit today to more than 15,000. Most of the new constellations will provide broadband communications, sharing spectrum in Ku/Ka-band and V-band. While not all the constellations will be launched, the potential for harmful interference is obvious. However, the actual risk of interference is not known. We perform a quantitative risk assessment for NGSO-NGSO interference to inform regulatory and operational decision making. To our knowledge, this is the first publicly available probabilistic NGSO-NGSO risk assessment for multi-constellation configurations. Unlike previous studies that only consider RF signal levels, we characterize interference impacts by the degradation in data throughput. We use Monte Carlo modeling to calculate the risk of co-channel interference between four planned NGSO constellations operating in Ka-band (LeoSat, O3b, OneWeb’s MEO, and Telesat) and three in V-band (OneWeb’s MEO, SpaceX’s LEO, and Telesat). We consider the interference risk to communication between an NGSO satellite and a user terminal. (The interference risk is greater for user terminals than gateways.) Engineering risk is the combination of likelihood and consequence. Our consequence metric is the percentage degradation in throughput relative to a fixed reference value (%DTp). Our baseline condition excludes interference hazards but includes poor atmospheric transmission so that %DTp can be non-zero even for baseline. Our primary result is that NGSO-NGSO interference risk in Ka-band and V-band is low, even without mitigation. Under reasonable assumptions, we find that the mean value of %DTp for the baseline condition can be up to 12 percentage points (pp); interference increases this by 3pp or less. In our most conservative scenario (all satellites in an interfering constellation form co-channel, co-polarization links with earth stations in a small area), the increase in mean %DTp due to interference can reach 18pp. The large constellations—OneWeb and SpaceX—are most likely to cause interference to others. Our results suggest that the need for interference mitigation will be limited. In cases where mitigation is needed, we find that look-aside mitigation based on ephemeris information—which does not require coordination between operators—can benefit large constellations but can harm small constellations. We believe that mitigation will have to include coordination. This raises questions about how to achieve coordination using operational information, e.g., which satellite is serving a particular earth station on which channel.