No, it reports 50 million detections on Windows endpoints, not that 50 million Windows endpoints were compromised. It reports an average of 5.8 detections per Windows endpoint, from which we can compute that fewer than 10 million Windows endpoints have MalwareBytes installed but not how many have detected malware (except that the total install base is an upper bound).

A more meaningful statistic they could report is the percentage of endpoints with MalwareBytes installed that have detected malware.