Overview
- Palantir designated Nebius Group its preferred sovereign AI infrastructure partner on Tuesday, Sept. 8, and said Nebius compute and inference endpoints will be brought inside Palantir’s enterprise perimeter after an integration period.
- The integration will let eligible Palantir customers deploy open models on Nebius systems and adapt those models with proprietary data while keeping control of compute, model weights and sensitive information.
- Nebius has been rapidly scaling to support the work, reporting strong Q2 revenue growth and large financings that it says will fund modular data centers, GPU purchases and targets of 800 MW to 1 GW connected power by year‑end.
- Separately, Palantir and NVIDIA on Thursday, Sept. 10, published results from a supply‑chain pilot showing a fine‑tuned 30B Nemotron model scored 86.7% on an allocation task versus 55.5% for a 550B model, underscoring that smaller, specialized open models can outperform much larger general models on domain tasks.
- Key unknowns remain: companies have not disclosed integration timing, initial capacity, pricing or contract terms, so customer rollout scale, revenue impact and whether compute supply or margins will constrain expansion are the next things to watch.