Setup
One API key. Full Vast.ai visibility.
Connect your Vast.ai account with a read-only API key. DoiT ingests your rental, storage, and bandwidth charges automatically. No agents, no changes to your training or inference workloads. You're looking at unified GPU cost reports within hours of connecting.
What you get
Built for the realities of renting GPUs on a marketplace
The things AI and FinOps teams actually ask us for when their GPU spend moves to marketplace pricing on Vast.ai.
GPU spend reporting
Slice Vast.ai spend by GPU type, instance, project, or team without exporting invoices into spreadsheets.
Real-time anomalies
Get alerted when a training run or forgotten instance spikes your bill.
Idle GPU detection
Spot rented instances sitting idle and stop paying for unused GPU hours.
Multi-cloud context
View Vast.ai spend next to your other cloud providers in one report.
Storage and bandwidth visibility
Separate GPU rental charges from the storage and bandwidth costs that pile up around training datasets and checkpoints.
Governance and budgets
Set budgets per team or project before experiments turn into surprises.
The billing page tells you what you spent. Cloud Intelligence™ helps you do something about it.
Beyond the Vast.ai billing page
Rollups by GPU type and project
Consolidated views across every rented instance, with drilldown into any GPU model, project, or team.
Real-time anomaly alerts
Machine-learning detection on instance, GPU type, and project dimensions, routed to Slack or email.
Forecasting and budgets
Project where marketplace-priced GPU spend is heading and set budgets before training costs run away.
Cost allocation hygiene
Attribute GPU hours to teams and projects, and split shared training infrastructure the way finance expects.
Multi-cloud AI cost view
Combine Vast.ai rentals with GPU spend on AWS or Google Cloud for one picture of your AI compute costs.
Forward Deployed Engineers
World-class cloud architects who work as an extension of your team to implement optimizations.
Attribute + AI compute
Map AI compute spend to the work that drives it
Attribute's runtime sensor covers infrastructure and AI inference, mapping costs to customers and workloads without relying on a tagging program. Paired with Vast.ai spend data in DoiT, you can see what each customer and workload actually costs. Best for: AI teams that want infrastructure and inference cost attribution in one view.
Compute coverage
Attribute costs across containerized and VM-based AI workloads.
Data service attribution
Cover the databases, object storage, caches, and queues feeding your pipelines.
Self-managed workloads
Attribute Kafka, Elasticsearch, and RabbitMQ costs supporting training and serving.
Full inference API coverage
Capture model invocation and streaming response calls across your serving stack.
Model-level costs
Break down inference spend by model across your AI stack.
Customer and workload views
Show which customer and workload is responsible for each cost.
Infrastructure and AI anomalies
Detect unusual spend across infrastructure and model inference.
Fast-growing companies run on Cloud Intelligence™
Avg. savings within first 90 days
Avg implementation time
“DoiT's focus on reliability, mixed with the system's flexibility, helps us safely optimize our Amazon EKS workloads with zero-touch from our engineers.”
Oren Ashkenazy
Director of DevOps and Cloud at Fiverr
Ready to connect your Vast.ai account?
Put your Vast.ai GPU spend in focus.
Frequently asked
questions
Can I see Vast.ai spend alongside my other cloud costs?
Yes. Cloud Intelligence™ normalizes Vast.ai rental, storage, and bandwidth charges into the same reporting layer as your other cloud providers, so you can compare GPU costs across platforms and roll everything up by team or project in one view.
What does connecting Vast.ai to Cloud Intelligence™ involve?
Connect with a read-only API key from your Vast.ai account. DoiT handles ingestion and normalization of your billing and usage data, with no agents to install and no changes to your training or inference workloads. Most teams are live within a day.
Which GPU types or instances account for the bulk of my spend?
Reports let you drill from total Vast.ai spend down to a specific instance or GPU model, and filter by project, team, or time range. You can see exactly which training runs and serving workloads are driving costs without exporting invoices.
Will I be alerted if a training run or forgotten instance spikes my bill?
Anomaly detection runs continuously across your Vast.ai spend. When a runaway training job or an instance left running overnight pushes costs outside the expected pattern, you get a Slack or email alert with the likely cause.
Can DoiT help me find idle GPU instances?
Yes. Because you pay for Vast.ai rentals by the hour whether GPUs are busy or not, idle instances are the most common source of waste. Cloud Intelligence™ surfaces spend on instances with no meaningful activity so you can stop them before the hours add up.
How does Vast.ai's marketplace pricing affect cost reporting?
Vast.ai hosts set their own prices, so your effective rates change with supply and demand. Cloud Intelligence™ reports on your actual billed charges over time, making it easy to see how rate changes, instance choices, and usage each contribute to your bill.
Vast.ai already shows me my billing. What does Cloud Intelligence™ add?
The Vast.ai billing view shows what you spent on that one platform. Cloud Intelligence™ adds multi-cloud rollups, anomaly detection, budgets and forecasting, cost allocation by team and project, and access to forward deployed engineers who help you act on what the data shows.
What access does Cloud Intelligence™ get to my Vast.ai account, and how is it protected?
Cloud Intelligence™ uses read-only API access with least-privilege scope. We never start, stop, or modify your instances without your approval, and the platform is SOC 2 Type II certified.