Customers

Community Cloud customers

Overlaps heavily with Vast's customer base:

• Indie ML developers and small teams.
• Researchers and academic users.
• Hobbyists (less so than Vast — RunPod's DX appeals more to slightly-more-serious users).
• ML consultancies running client workloads.
• Some Series-A/B AI startups using it for non-production experimentation.

The user who'd be on Vast but for the DX often ends up on RunPod. The 10-25% price premium is the cost of polish.

Secure Cloud customers

This is RunPod's enterprise channel:

• Series-A through Series-C AI startups running production inference or training.
• Mid-market enterprises with AI initiatives.
• Some enterprises piloting AI workloads before committing to larger reserved capacity elsewhere.
• Customers who need SOC 2 attestation but don't want hyperscaler pricing.

The customer mix tilts smaller than CoreWeave's enterprise base. RunPod doesn't have the multi-year Microsoft / Meta scale deals; it has lots of smaller annual commitments.

Serverless customers

Serverless attracts a specific profile:

• SaaS companies adding AI features to their products.
• Indie AI app builders deploying inference.
• Internal AI tools at non-AI-native companies.
• Workloads with sparse / spiky traffic.

The Serverless customer base is distinct from the Pod-based customer base — these are buyers who don't want to think about GPU operations at all.

Common workloads

• Fine-tuning small-to-mid open-source models (Llama family, Mistral, etc.).
• Image-generation training and inference (Stable Diffusion variants, ComfyUI workflows).
• LLM inference for product features.
• Computer-vision training (object detection, segmentation).
• Voice / audio model training and inference.
• Research prototyping and experimentation.

What's not common on RunPod: frontier-scale training, regulated-industry production deployments, multi-thousand-GPU runs. Those go elsewhere.

Who they win vs lose

Wins

• Users who want better DX than Vast offers.
• Users who want a managed inference flavor (Serverless) without going to Together.AI.
• Mid-market companies who need some reliability but don't want hyperscaler bills.

Loses

• Pure price optimizers (go to Vast).
• Largest enterprise deals with multi-year multi-thousand-GPU commitments (go to CoreWeave or Crusoe).
• Customers who need the broadest inference model selection (often go to Together.AI for that).
• Regulated industries with strict compliance (typically end up at hyperscalers or enterprise neoclouds with stronger compliance posture).

Takeaway

RunPod's customer base is bigger and more diverse than a pure marketplace's, smaller and less enterprise-heavy than a pure cloud's. The dual-product strategy lets RunPod capture both tails of demand. The next chapter looks at the strengths that follow from this position.