Yandex AI Studio launches on-premises with Selectel GPUs and two-week deployment
Three Russian vendors now deploy Yandex AI Studio inside customer data centers, pairing generative models and no-code agent builders with Selectel's HGX B300, H200, A100, and RTX PRO 6000 racks.
Yandex B2B Tech, Selectel, and MetaMentor have packaged on-premises AI infrastructure for organizations blocked from public cloud by data-residency or security rules. Customers lease a pre-configured rack that arrives at their own network perimeter within five days, with full operational readiness in roughly two weeks. Yandex AI Studio runs on Selectel hardware (HGX B300, H200, A100, RTX PRO 6000), while MetaMentor handles deployment, driver tuning, orchestration software, and storage-layer configuration—work that typically consumes weeks of in-house engineering.
The offering closes a real gap: between "hardware exists" and "models actually train on it" lies a month of infrastructure setup. Yandex AI Studio includes multimodal generative capabilities and a drag-and-drop agent builder; Selectel supplies GPU clusters and networking fabric; MetaMentor installs the InfiniBand or Lustre storage layers and support contracts. All three vendors are based in Russia, and workloads stay inside the customer's own infrastructure. The B300-based HGX nodes use Nvidia's latest Blackwell generation; H200 and A100 options cover a range of price and memory configurations. Yandex has not published per-rack pricing, but the two-week deployment window includes software licensing and cluster tuning.
