QumulusAI increased its number of deployed GPUs by over three times in the second quarter, as it brought contracted computing capacity into operation and secured new customer agreements worth $169.7 million. The company, which is based in Atlanta, had 183,088 GPUs deployed by the end of June, compared to 952 at the start of the quarter.
The revenue increased by more than two times compared to the previous year, reaching $6.7 million, with compute revenue making up approximately 84% of the total revenue, in contrast to 43% a year prior. In the quarter, QumulusAI secured 21 fresh direct AI compute deals, resulting in a cumulative signed customer contract value of $282.5 million across 40 long-term agreements.
During the quarter, $169.7 million was signed, which represents the total anticipated take-or-pay value. QumulusAI concluded the quarter with 83 MW of HPC capacity that has been allocated to clients through executed lease and colocation deals. The final GPU installations are currently being carried out.
The corporation anticipates that the entire 8 MW will produce income this year. ChatDolphin has agreed to obtain a maximum of 3.75 MW in the metropolitan Atlanta region, with plans to start operations in the final quarter. This is connected to the news that a California judge has mandated a comprehensive environmental assessment of a 330 MW data center.
According to CEO Michael Maniscalco, the speed of QumulusAI’s deployment is now reliant on identifying available, powered data center capacity that can be made operational according to customers’ schedules. Maniscalco explained that this is essentially a combination of power and the ability to deploy data center capacity. “The more challenging aspect is identifying available power sources capable of being operational on the required timelines of our clients.”.
The firm aims to focus on smaller chunks of capacity that can be utilized and enlarged as the demand increases.
