GPU Rental Cost Calculator
Renting a GPU in the cloud costs as little as $0.40/hr for an RTX 4090 or up to $2.50/hr for an H100, but buying hardware has a high upfront cost. This calculator compares your monthly rental bill against owning your own GPU — factoring in purchase price, amortization, and electricity — so you can find the exact month when owning becomes cheaper.
When to use this calculator
- ML engineers deciding whether to rent cloud GPUs or build a local workstation
- Researchers estimating monthly compute budgets for training experiments
- Startups comparing Vast.ai, RunPod, and Lambda pricing for inference workloads
- Hobbyists calculating if a used RTX 4090 pays off faster than a cloud subscription
- Data scientists optimizing spot-instance hours vs. reserved GPU time
- Teams planning GPU infrastructure spend for 2026 project budgets
How it works
2 min readWhat is GPU rental cost?
GPU rental cost is the hourly or monthly fee charged by cloud providers like Vast.ai and RunPod for accessing graphics processors remotely. Prices range from $0.40/hour for an RTX 4090 to $2.50/hour for an H100. Renting eliminates upfront hardware investment, making it cost-effective for short-term projects or variable workloads.
How It Works
This calculator compares two cost models for GPU compute: cloud rental (pay-per-hour on platforms like Vast.ai, RunPod, or Lambda Labs) versus hardware ownership (purchase + electricity + overhead amortized over a set period).
---
Formulas
// Rental side
hours_per_month = hours_per_day × days_per_month
monthly_rental = hours_per_month × rental_rate_per_hour
// Ownership side
monthly_amortization = purchase_price / amortization_months
monthly_electricity = (tdp_watts / 1000) × hours_per_month × electricity_rate
monthly_overhead = (purchase_price × overhead_fraction) / 12
monthly_ownership = monthly_amortization + monthly_electricity + monthly_overhead
// Break-even
// At month N: cumulative_rental(N) = purchase_price + cumulative_ownership_opex(N)
// cumulative_rental(N) = N × monthly_rental
// cumulative_ownership_opex(N) = N × (monthly_electricity + monthly_overhead)
// purchase_price = N × (monthly_rental - monthly_electricity - monthly_overhead)
// breakeven_month = purchase_price / (monthly_rental - monthly_electricity - monthly_overhead)The break-even formula assumes you pay the full purchase price upfront and your only ongoing ownership costs are electricity and overhead (no amortization debt once you own the hardware).
---
GPU Reference Rates (2026 spot pricing, platforms vary)
| GPU | Rental $/hr | TDP | New/Used Price |
|---|---|---|---|
| RTX 4090 | ~$0.40 | 450 W | ~$1,800 new |
| A100 40GB | ~$1.10 | 300 W | ~$10,000 |
| A100 80GB | ~$1.40 | 400 W | ~$12,000 |
| H100 SXM | ~$2.50 | 700 W | ~$30,000 |
| RTX 3090 | ~$0.20 | 350 W | ~$700 used |
Spot prices fluctuate. Vast.ai prices can be 30–50% lower during off-peak hours.
---
Worked Example
Scenario: RTX 4090, 8 hrs/day, 22 days/month, $0.12/kWh, $1,800 purchase, 24-month amortization, 10% overhead.
In this scenario, renting is cheaper unless you plan to use the GPU for 39+ months continuously.
---
Limitations
Frequently asked questions
What is the cheapest GPU to rent for AI inference in 2026?
The RTX 4090 offers the best cost-per-TFLOP for FP16 inference at approximately $0.40/hr on Vast.ai spot instances. For FP8 workloads requiring larger VRAM, an A100 80GB at ~$1.40/hr may be more cost-effective per token generated.
Does the break-even calculation include the cost of the host machine?
No. This calculator only models the GPU itself. If you are building a dedicated workstation, add $500–$2,000 for a PCIe platform. If using an existing machine, that cost is already sunk and can be ignored.
How does Vast.ai pricing compare to RunPod or Lambda Labs?
Vast.ai typically offers the lowest spot prices (peer-to-peer marketplace) but with variable uptime. RunPod offers a balance of reliability and cost. Lambda Labs targets enterprise reliability with slightly higher rates. All three are significantly cheaper than AWS or GCP GPU instances.
What overhead percentage should I use for a home setup?
For a home desktop workstation, 5–10% is reasonable (covers thermal paste, dust cleaning, minor repairs). For a rack-mounted server, use 15–20% to account for colocation fees, redundant cooling, and maintenance contracts.
Why is the A100's TDP listed as 300W when the spec sheet says 400W?
The A100 40GB PCIe variant has a 250–300W TDP. The SXM4 (server) variant reaches 400W. The calculator uses conservative PCIe figures for the 40GB model. Adjust TDP manually if using SXM4 infrastructure.
Should I use amortization period = 24 months or 36 months?
For fast-depreciating consumer GPUs (RTX 4090), 24 months aligns with the typical useful-life cycle before a next-generation card cuts resale value significantly. For datacenter GPUs (A100, H100) with longer enterprise lifespans, 36–48 months is more accurate.
Does electricity rate include cooling overhead?
No. The formula uses raw GPU TDP only. A typical home PUE (Power Usage Effectiveness) of 1.1–1.3 means actual electricity draw is 10–30% higher. Multiply your electricity rate by your PUE factor to get a more accurate figure.
Can I use this calculator for multi-GPU setups?
Yes — multiply your hours/day by the number of GPUs, or equivalently multiply the rental rate and purchase price by the number of GPUs. The per-unit math is linear.
What if my rental rate is much lower than the defaults shown?
Select 'Custom / Other GPU' and enter your actual negotiated or spot rate. Vast.ai spot prices for RTX 4090s have been seen as low as $0.18/hr during off-peak periods, which dramatically extends the break-even timeline for ownership.