Sources & methodology
Every number in our calculator and articles is verifiable. Marketplace rental rates are dynamic (they move with demand), so we link the live pages where you can check current values yourself. We are independent and not affiliated with Vast.ai.
1. Platform fee and payouts
Vast.ai. The host sets the prices for three components: GPU rental ($/hr) + storage ($/TB·mo) + bandwidth ($/TB) — and receives them in full: nothing is deducted from the host’s price. Vast.ai’s margin is a markup charged to the client and is not taken out of the host’s income. Our own measurement (August 2026, our own machine) puts the markup at ×1.33: a host price of $0.21/hr shows as $0.28/hr to the client, and the same ratio applies to storage and traffic ($3/TB → $4/TB). So the client pays about 33% more, and the platform keeps 25% of the client price. (Vast’s own wording differs: docs — “no markup on host-set prices”, article — “prices ~25% above host earnings”; either way the host keeps their full price.) Billing is per-second; payouts are weekly (Fridays, 03:00 MSK / 00:00 UTC) once your balance reaches $20, minus the payment provider’s transfer fee (Stripe 0.25–1% / Wise 1% / PayPal 3%).
2. Rental rates ($/GPU-hour)
The default rates in the calculators are what the host receives: the median offer price on vast.ai/pricing (snapshot of 14 Sep 2026, on-demand) without the platform markup, i.e. ÷1.33. They change with demand, so check the live links below and enter your own. The card price is a 2026 second-hand market reference and is editable too. Defaults:
| GPU | VRAM | Host rate, $/hr | Power, W | Card price, $ |
|---|---|---|---|---|
| RTX 2060 | 6 GB | $0.05 | 160 | 110 |
| RTX 2060 SUPER | 8 GB | $0.05 | 175 | 130 |
| RTX 2070 | 8 GB | $0.07 | 175 | 140 |
| RTX 2070 SUPER | 8 GB | $0.07 | 215 | 160 |
| RTX 2080 Ti | 11 GB | $0.10 | 250 | 260 |
| Titan RTX | 24 GB | $0.11 | 280 | 600 |
| RTX 3060 | 12 GB | $0.05 | 170 | 180 |
| RTX 3060 Ti | 8 GB | $0.05 | 200 | 200 |
| RTX 3070 | 8 GB | $0.06 | 220 | 230 |
| RTX 3070 Ti | 8 GB | $0.08 | 290 | 260 |
| RTX 3080 | 10 GB | $0.08 | 320 | 320 |
| RTX 3080 Ti | 12 GB | $0.10 | 350 | 400 |
| RTX 3090 | 24 GB | $0.12 | 350 | 900 |
| RTX 3090 Ti | 24 GB | $0.14 | 450 | 1000 |
| RTX 4060 | 8 GB | $0.05 | 115 | 250 |
| RTX 4060 Ti | 16 GB | $0.08 | 165 | 350 |
| RTX 4070 | 12 GB | $0.08 | 200 | 450 |
| RTX 4070 SUPER | 12 GB | $0.07 | 220 | 520 |
| RTX 4070 Ti | 12 GB | $0.09 | 285 | 600 |
| RTX 4070 Ti SUPER | 16 GB | $0.10 | 285 | 700 |
| RTX 4080 | 16 GB | $0.14 | 320 | 900 |
| RTX 4080 SUPER | 16 GB | $0.14 | 320 | 1100 |
| RTX 4090 | 24 GB | $0.37 | 450 | 2200 |
| RTX 5060 | 8 GB | $0.07 | 145 | 300 |
| RTX 5060 Ti | 16 GB | $0.11 | 180 | 430 |
| RTX 5070 | 12 GB | $0.10 | 250 | 550 |
| RTX 5070 Ti | 16 GB | $0.18 | 300 | 750 |
| RTX 5080 | 16 GB | $0.25 | 360 | 1000 |
| RTX 5090 | 32 GB | $0.50 | 575 | 2500 |
| RTX A4000 | 16 GB | $0.08 | 140 | 550 |
| RTX A5000 | 24 GB | $0.15 | 230 | 1600 |
| RTX A6000 | 48 GB | $0.27 | 300 | 3200 |
| RTX 4000 Ada | 20 GB | $0.12 | 130 | 1200 |
| RTX 4500 Ada | 24 GB | $0.38 | 210 | 2200 |
| RTX 5000 Ada | 32 GB | $0.48 | 250 | 4000 |
| RTX 6000 Ada | 48 GB | $0.60 | 300 | 6500 |
| RTX PRO 4000 Blackwell | 24 GB | $0.20 | 140 | 1500 |
| RTX PRO 4500 Blackwell | 32 GB | $0.26 | 200 | 2300 |
| RTX PRO 5000 Blackwell | 48 GB | $0.70 | 300 | 4500 |
| RTX PRO 6000 Blackwell | 96 GB | $1.10 | 600 | 8500 |
| Tesla T4 | 16 GB | $0.11 | 70 | 600 |
| A10 | 24 GB | $0.18 | 150 | 1800 |
| L4 | 24 GB | $0.25 | 72 | 2200 |
| A40 | 48 GB | $0.35 | 300 | 3500 |
| L40 | 48 GB | $0.35 | 300 | 6000 |
| L40S | 48 GB | $0.59 | 350 | 7000 |
| A100 PCIe 80GB | 80 GB | $0.53 | 300 | 8000 |
| A100 SXM 80GB | 80 GB | $0.60 | 400 | 9000 |
| H100 PCIe 80GB | 80 GB | $2.05 | 350 | 22000 |
| H100 SXM 80GB | 80 GB | $2.24 | 700 | 28000 |
| H100 NVL 94GB | 94 GB | $2.15 | 400 | 26000 |
| H200 141GB | 141 GB | $3.26 | 700 | 32000 |
| B200 192GB | 192 GB | $6.33 | 1000 | 45000 |
| B300 288GB | 288 GB | $7.00 | 1100 | 55000 |
RUB reference rate: 92 ₽/$. All maths is in dollars, electricity included (cents per kWh); ruble figures are for reference only.
3. Utilization
Utilization is not a fixed fact but an assumption: it depends on demand, GPU class, price, uptime and host rating. The income calculator starts at 50% — a starting point, replace it with your real figure. A large, long-running marketplace (Vast.ai since 2019) usually sees higher AI demand, but there are no guarantees.
4. Calculator methodology
Income calculator (month = 730 hours):
effUtil = min(1, util × ramFactor × cpuFactor) # ramFactor: ≥16 ГБ/GPU → 1, ≥8 → 0.9, иначе 0.75; cpuFactor 0.92 / 1.00 / 1.04
grossHour = rate × effUtil × count # $/час, ставка хоста — без удержаний
gross = grossHour × 730
fee = gross × 0.01 # комиссия при выводе (Wise, Stripe)
gpuW = TDP × (0.8 × effUtil + 0.1 × (1 − effUtil))
kWh = 730 × (gpuW × count + cpuW + ram × 0.4) / 1000 / 0.92 # cpuW 65 / 120 / 200, КПД БП 0.92
elec = kWh × cents / 100 # центы США за кВт·ч → $
net = gross − fee − elec − 8 # $8/мес — VPS для туннеля
invest = card × count + server + ram × 3 # сервер без карт: 500 / 1000 / 2000 по умолчанию
payback = invest / net # месяцев5. All sources
- vast.ai/hosting — GPU hosting on Vast.ai (overview)
- cloud.vast.ai/earnings — host earnings calculator (live market data)
- vast.ai/article — how much you can earn — per-GPU rates and “prices ≈25% above host earnings”
- docs.vast.ai — pricing — Vast.ai pricing docs