Estimate GPU rental income
Net income
—
The card draws power whether or not anyone rents it. Idle draw is why low utilization so often turns a good hourly rate into a loss.
How the math works
Gross income is the hourly rate times the hours in a month, times the share of those hours someone actually rents your card. The platform takes its percentage. Then the line everyone forgets: your rig draws power around the clock — idle or rented — so electricity is charged on all 730 hours of the month, not just the rented ones.
The formula: net per month = rate × 730 × utilization × (1 − fee) − watts ÷ 1000 × 730 × your rate. Break-even is the card cost divided by monthly net.
Where the defaults come from
Rates come from Vast.ai's public per-GPU pricing feed (September 2026), discounted about 25% from median listing prices to reflect what Vast's own article says hosts typically realize. The default 30% utilization reflects idle time on an open marketplace — not the near-100% the listing page implies. Average power draw sits between idle and full load because a shared rig bounces between both. The electricity default of 18.3¢ is the EIA's US residential average from July 2026; your real rate is on your bill, and under about 10¢ it changes every conclusion on this page.
What the calculator does not include
Vast.ai also bills storage while an instance exists (even when stopped) and bandwidth per byte in both directions — small on a single card, real on a busy one. Payouts carry a $20 minimum. Hardware risk is unmodeled: a card running hot AI jobs around the clock ages fans faster. And demand risk dominates everything: rates fell hard the last time rented supply flooded in, and they can fall again. Treat every output here as this month's snapshot, not a lease.
Next steps
The full platform comparison, utilization reality check, and host reports are in the GPU rental income guide. To compare against mining the same card, run the GPU mining calculator; for the big-iron version of the same trade, see the Antminer profitability calculator.
Frequently asked questions
How much money can you make renting out your GPU?
Vast.ai's public pricing feed (September 2026) shows median listing rates of about $0.15 per hour for a used RTX 3090 and $0.44 for an RTX 5090 — but Vast's own article says hosts typically realize roughly 25% less than the listing price, and open-marketplace cards sit idle most hours. At a realistic 30% utilization, a 5090 nets roughly $30 a month after electricity at the US average rate, and a 3090 roughly breaks even. One card is pocket money, not income.
Is renting out a GPU profitable?
With an average electricity rate, barely — power is the swing factor. A 3090-class rig drawing 200 W around the clock costs about $27 a month at the 18.3¢ US average, which eats a 3090's entire realistic rental income. It turns profitable with cheap power (under about 10¢), higher-demand cards (4090/5090), or a card you already own. Vast's own "$700 to $1,400 a month" example requires a four-5090 rig at 80% utilization — $8,000-plus in cards, not a one-GPU outcome.
Which platforms pay you to share your GPU?
Vast.ai is the main open marketplace for home hosts: you set your own price, payouts arrive weekly with a $20 minimum via PayPal, Wise or Stripe, and it publishes live per-GPU pricing. Salad runs Windows consumer PCs but pays in app balance redeemable for PayPal or gift cards. RunPod's Community Cloud and io.net also connect home GPUs, with stricter eligibility. Read the current host docs before buying hardware — terms move.
Why does my rig earn less than the rate times 24 hours?
Three haircuts. Utilization: most hours, nobody rents your card — 30% is a realistic default. The listing-to-earnings gap: Vast's own article puts typical host earnings about 25% below live listing prices. And 24/7 power: the card draws electricity whether rented or idle, which is why low utilization turns a good hourly rate into a loss.
