The use of automated blood cultures system, such as MB/BacT, has provided a novel opportunity for laboratories to diagnose mycobacteremia in HIV-infected patients. However, no such study has been carried out in India so far. This prospective study was conducted on 52 HIV-positive patients with suspected tuberculosis who were referred to our tertiary care hospital in New Delhi. In these patients, the prevalence of mycobacteremia was evaluated using the MB/BacT automated culture system (bioMérieux, France). Twenty-seven HIV-negative but suspected tuberculosis patients were also included for comparison. Mycobacteria could be isolated from sputa or fecal samples of 20 HIV-positive patients (38.4%), and in nine (45%) of these 20 cases, mycobacteria could also be isolated simultaneously from their blood specimens. In the remaining 32 patients, all relevant non-hematological clinical samples remained negative for mycobacteria, but the pathogen could be detected from the blood samples of seven (21.87%) of these 32 patients. Therefore, only 25 (48%) clinically suspected patients remained negative in both Löwenstein-Jensen (L-J) and MB/BacT culture methods, and 12 of these responded to anti-tubercular treatment, while in the rest either non-tubercular diagnosis was established or they were lost to follow-up. The study revealed that low Symbol counts and poor or no reactivity to purified protein derivative (PPD) were the best clinical predictors for the occurrence of mycobacteremia in HIV-positive patients. Of the 16 isolates from blood, 13 were diagnosed as Mycobacterium tuberculosis and one each were identified as M. avium, M. kansasii, and a mixed infection of M. tuberculosis and M. avium complex. The prevalence rate of mycobacteremia was significantly low (11.1%) in HIV-negative patients. In conclusion, this study showed that blood culture could be an important adjunct investigation for confirming the clinical diagnosis of tuberculosis in HIV-positive patients.