Oblivus

Oblivus is a small independent GPU cloud operator that markets on-demand and reserved access to NVIDIA accelerators for AI training, inference, and rendering workloads. The company runs a published fleet of "more than 18,000" GPUs across nine datacenter locations, with named sites in Montreal (MON1) and outside Oslo (OSL1) and seven additional single-tenant facilities the company has not publicly identified. Its listed inventory centers on last-generation and mid-tier silicon, H100 80GB, A100 80GB, L40, RTX A6000, and RTX 4090, rather than the H200 or Blackwell B200 parts that anchor the marketing of larger neoclouds.

The operating history is unusual for a company pitching itself as a serious AI infrastructure provider. Oblivus Limited was incorporated at a Covent Garden mail-forwarding address in March 2022 by three Turkish founders, Mehmet Ceylan, Doruk Alp Ulgen, and Anil Ipeksumer, all in their twenties at the time; Ipeksumer resigned in December 2023, and Companies House records show the UK entity was dissolved on 21 July 2026. Public disclosures of funding rounds, employee headcount, revenue, and any successor US entity are not available, and the company has not surfaced in coverage from Reuters, Bloomberg, The Information, or Data Center Dynamics.

Where Oblivus does compete is on price and long-tail customer mix. Its published H100 rate of $3.50 per hour on demand and $2.56 per hour reserved sits below the CoreWeave/Lambda tier, and its customer logos lean toward university research groups (Stanford, UCLA, Notre Dame, Sydney) and crypto- or decentralized-AI-adjacent projects (Bittensor, Corcel, Akash, ShadeForm) rather than named hyperscalers or frontier labs. It is best understood as a mid-market GPU rental brand rather than a peer of the funded, hyperscale neoclouds.