How they use Nebius AI Cloud

Shopify uses Nebius H200 clusters for model training. Its January 2026 engineering report describes persistent Kubernetes clusters, InfiniBand for distributed training and SkyPilot routing H200 jobs to Nebius. Training datasets are replicated across clouds and loaded from volumes within the Nebius environment. Shopify uses Nebius alongside Google Cloud so its teams can move training jobs between providers. Nebius reports reliable instances for Shopify’s large multi-node AI experiments and cross-cloud bursting through SkyPilot without changing established workflows. SemiAnalysis also describes Shopify’s work with SkyPilot and dstack sky on Nebius for dynamic research workloads. Shopify also lists Nebius B.V. as a subprocessor for optimizing computational workloads and hosting technical environments. The listing identifies all platform data as data that may be processed.

In a later talk, a Shopify speaker described foundation-model pre-training, fine-tuning and experiments on Nebius. Nebius also describes dedicated AI Cloud capacity for training Shopify’s foundation models at production scale to understand merchant and buyer journeys. Shopify Principal Engineer Javier Moreno says the SkyPilot integration lets the team execute jobs across GPU providers without disrupting internal processes. In its Q2 2025 earnings call, Nebius said Shopify uses its AI infrastructure alongside Toloka training data to optimize the merchant journey. In its Q3 2025 earnings call, Nebius reported expansion with Shopify as an existing customer. An announcement for an upcoming GTC Berlin session says Shopify fine-tunes open-source models and reduces inference costs while scaling on Nebius AI Cloud.

The team listed fine-tuning smaller SimGym models on 16 Nebius H200 GPUs as a next step in February 2026, with Qwen3-32B identified as a candidate.