Calculate the monthly cost of running cloud GPU instances continuously or on demand.
Monthly cloud GPU cost is the product of hourly rate, hours of daily utilization, GPU count, and days in the billing period. Running GPUs 24/7 (hoursPerDay = 24) reflects always-on training clusters, while lower daily hours model bursty or scheduled workloads that spin down during idle periods to save cost.
monthly_cost = hourly_rate × hours_per_day × gpu_count × days_per_month
Reducing daily usage from 24 to, say, 10 hours cuts cost proportionally — about 58% savings — assuming your workload can tolerate scheduled downtime.
Spot instances offer lower hourly rates (often 50-70% off) in exchange for the risk of interruption; substitute a lower gpuTypeRate to model spot pricing.
This calculator adds a daily-hours and days-per-month dimension for monthly budgeting; the GPU Cost Estimator compares GPU types directly against a fixed hours total.