Voltage Park Unveils $1 Billion Cloud Infrastructure to Bridge the ML Compute Gap in AI Development

Voltage Park introduced a groundbreaking cloud infrastructure for AI development, addressing the significant scarcity of Machine Learning (ML) compute resources in the market, obstructing innovation and progress in the field.
This infrastructure, valued at $1 billion, encompasses 24,000 NVIDIA H100 GPUs, offering large-scale users bare-metal access initially and aiming to introduce short-term leases and hourly billing, making cutting-edge hardware more accessible.

Voltage Park, under the leadership of CEO Eric Park, unveiled an ambitious initiative on October 29, 2023, to alleviate the severe shortage of advanced ML computing resources that has been stifling innovation in the field of AI. Eric Park highlighted the challenges faced by ML teams and AI founders due to the scarcity, emphasizing the months-long wait and the exorbitant costs they have to bear to access cutting-edge hardware for training their models.

This newly introduced cloud infrastructure by Voltage Park boasts a staggering value of $1 billion and includes approximately 24,000 NVIDIA H100 GPUs, positioning itself as one of the largest ML compute clouds globally. Initially providing bare-metal access to high-performance clusters, the plan is to extend support for short-term leases and hourly billing, catering to the needs of diverse users, primarily smaller entities seeking flexibility.

Empowering AI Innovators

Voltage Park’s infrastructure, fully interconnected with 3.2T InfiniBand, not only promises accessibility but also integrates tools like Slurm, Kubernetes, and Mosaic, aiming to assimilate with existing training frameworks seamlessly. The initiative has already commenced services to prominent AI firms such as Imbue and is finalizing agreements with other industry leaders like Character.ai and Atomic AI. The full operational capacity is expected to be ready by early next year.

Kanjun Qiu, CEO of Imbue, praised Voltage Park for significantly enhancing their model training performance by facilitating quicker access to crucial computing resources. Voltage Park actively seeks feedback from potential customers to tailor their infrastructure, catering to a wide range of use cases, from experimentation to training and inference.