Skip to main content

What is AI Streams?

AI Streams is a GPU cluster monitoring and management platform built on the same Streams infrastructure as Hash Streams. Where Hash Streams is designed around ASIC miners, AI Streams is purpose-built for GPU nodes: it monitors cluster health, manages workload scheduling, optimizes power draw, and provides usage analytics for AI/ML training runs, inference deployments, rendering farms, and GPU mining operations. Both products share the same agent architecture, the same data pipeline, and the same platform account. If you run both ASIC miners and GPU clusters you manage them from a single dashboard.

Key features

Cluster monitoring

Live visibility into GPU utilization, VRAM usage, temperature, power draw, fan speed, and NVLink / PCIe health for every node in your cluster. Dashboard refreshes on the same 50-second polling cycle as Hash Streams.

Workload management

Track active jobs, queued workloads, and job history. Associate resource consumption with specific training runs or inference deployments for accurate cost attribution.

Power optimization

Apply power limits per GPU or per node based on grid conditions. Integrates with the same curtailment signal pipeline as Hash Streams — a single demand response event can curtail both your ASIC miners and your GPU cluster simultaneously.

Node scaling

Monitor cluster capacity and track node availability over time. Flag nodes that are offline, degraded, or suspended, and view uptime trends for capacity planning.

Job queue

View the current job queue, identify bottlenecks, and see throughput metrics per GPU model or node. Useful for identifying underutilized hardware.

Usage analytics

Per-node and per-GPU usage reports with energy consumption, average utilization, and cost attribution. Export data for billing or infrastructure planning.

Use cases

Monitor long-running training jobs across multi-GPU and multi-node setups. Track GPU utilization, VRAM headroom, and power draw in real time. Set up event rules to alert you if a training job stalls or a GPU goes offline mid-run.
Track GPU utilization and latency for inference workloads. Identify nodes that are overloaded or underutilized and rebalance accordingly.
Monitor GPU utilization across render farm nodes. Track job completion times and hardware health to minimize render farm downtime.
Use AI Streams to monitor GPU-based mining rigs the same way Hash Streams monitors ASIC fleets — with live hashrate, temperature, power draw, and automated curtailment when power prices spike.

How AI Streams relates to Hash Streams

Both products are part of the Dune Digital Streams platform:
Hash StreamsAI Streams
Compute typeASIC miners (SHA-256, Scrypt, etc.)GPU nodes (NVIDIA, AMD)
Primary useBitcoin / altcoin mining operationsAI/ML, rendering, GPU mining
AgentSame binary, configured for ASIC scanningSame binary, configured for GPU/CUDA discovery
CurtailmentStops or throttles ASICsThrottles GPU power limits or pauses jobs
DashboardMining control, PDU control, earningsCluster monitor, job queue, usage analytics
PricingPay As You Go or BusinessPay As You Go or Business
If you already use Hash Streams, you can add AI Streams to the same account and manage both from the same organization and billing setup.

Getting started with AI Streams

The setup process mirrors Hash Streams:
1

Create a site (or reuse an existing one)

AI Streams sites work the same as Hash Streams sites. If your GPU cluster is co-located with your ASIC miners, you can use the same site and deploy a second agent configured for GPU discovery.
2

Download and install the agent

Download the agent from Infrastructure → Agents → Download agent in your dashboard. The same agent binary supports both ASIC and GPU monitoring — the mode is configured through the agent’s network segment settings.
3

Register the agent

Generate a registration token from Infrastructure → Agents, same as Hash Streams. The token binds the agent to your organization and site.
4

Configure GPU network segments

Add network segments pointing at the subnets where your GPU nodes are reachable. The agent will discover nodes and begin reporting telemetry.

Pricing

AI Streams uses the same usage-based pricing as Hash Streams — you pay for the telemetry data your agent uploads.
PlanBase priceData rate
Pay As You Go$0 / month$0.018 / MB
Business$149 / month$0.015 / MB
Both plans include 200 MB of welcome credits. The Business plan unlocks automated curtailment (including GPU power throttling), unlimited agents, advanced analytics, and priority support.
AI Streams is currently in active development. Some features listed here may be in preview. Contact sales for early access or to discuss enterprise deployments.