NVIDIA — Partner Intelligence

Current state (as of 2026-05-17)

NIM microservices (NVIDIA Inference Microservices) are the default inference layer for the NVIDIA AI Enterprise stack — containerized, GPU-optimized, multi-cloud-deployable. NeMo continues to anchor the LLM build + finetune surface. DGX Cloud (managed GPU clusters via cloud partners — Oracle, AWS, Azure, GCP) is the enterprise-tier compute lane. The Oracle × NVIDIA AI accelerator pack wave is moving through EMEA channel right now, with Arrow ECS Germany as a primary distribution surface. NVIDIA Inception (startup program) operates a nominee pathway — entry typically through a corporate or VC nomination (to verify current rev). Munich EBC (Executive Briefing Center) is active for partner-event work.

Product map

• NIM (NVIDIA Inference Microservices) — containerized inference for production LLMs and multimodal models. GPU-optimized, multi-cloud-deployable, OpenAI-compatible API surface.
• NeMo — framework + microservices for LLM build, finetune, retrieval, and guardrails. Includes NeMo Curator, NeMo Customizer, NeMo Evaluator, NeMo Guardrails, NeMo Retriever.
• NVIDIA AI Enterprise — the licensed enterprise stack. NIM + NeMo + AI Workbench + management + support, deployable on certified infrastructure or DGX Cloud.
• DGX Cloud — managed GPU clusters delivered through cloud partners (Oracle, AWS, Azure, GCP).
• AI Workbench — local + cloud GPU development environment.
• NVIDIA Blueprints — opinionated reference architectures (RAG, video search, drug discovery, virtual screening, etc.).
• CUDA / cuDNN / TensorRT / TensorRT-LLM — the silicon-layer SDKs.
• NVIDIA Inception — startup program. Free tier of credits + technical resources + go-to-market visibility. Nominee pathway. (Entry criteria to verify against current program page.)
• Partner Network (NPN) — channel + technology partner tiers for resellers, distributors, OEMs, ISVs.
• Munich EBC — Executive Briefing Center for European partner work.
• Open-source artifacts — NeMo (Apache 2.0), TensorRT-LLM (Apache 2.0), CUDA-Q (Apache 2.0), many reference blueprints.