Hyperstack

Hyperstack is the on-demand GPU cloud brand of NexGen Cloud, a UK-owned AI infrastructure company founded in 2020 and headquartered in London with a secondary office in Marbella, Spain. Launched as a self-serve platform in 2023, Hyperstack targets the layer between hyperscaler AI clouds and bare-metal broker marketplaces: developers rent NVIDIA GPUs by the hour with published prices, no egress fees, and no committed spend, while enterprise customers move to the sister "Secure Private Cloud" product for single-tenant reserved clusters.

The company positions itself as an EU-sovereign, NVIDIA-partnered alternative to AWS/Azure/GCP, with infrastructure distributed across six deployments in Norway, Sweden, Canada and the United States and a marketing claim that most of its capacity runs on renewable power. Its GPU menu spans the current NVIDIA training and inference stack (H100, H200, A100, L40, plus Blackwell B200 available and B300 slated for late-2026 delivery) at on-demand list prices around $3.20/hr for H100 SXM and $3.99/hr for H200 SXM, undercutting hyperscaler on-demand rates but pricing above the cheapest broker marketplaces.

Distinct from pure-play training clouds such as CoreWeave or Crusoe, NexGen splits the business into three lanes: Hyperstack (self-serve GPUaaS), Secure Private Cloud (reserved enterprise clusters) and Infrahub Compute (hardware financing and deployment). The company has not disclosed a total funding figure, GPU inventory count, or megawatt capacity, and remains privately held with no publicly named lead investors, which places it in the mid-tier of the neocloud pack in terms of externally verifiable scale.