Cómo se calcula
Owning a GPU costs two things each month: depreciation — the purchase price minus what you can sell it for later, spread over three years — and electricity for the hours it actually runs (board power × hours × your rate). Renting costs the hourly price times the hours you use, and nothing when it's idle.
That makes utilization the deciding number. The breakeven is the monthly usage where both cost the same: fixed depreciation ÷ (rental rate − power cost per hour). Below it, renting wins because you only pay while working; above it, owning wins because the card is paid for whether it's busy or not. A GPU that runs a few evenings a week almost never pays back; one that trains or serves around the clock often does within a year.
What this leaves out, deliberately: the host machine (CPU, RAM, PSU, cooling — often 30–60% on top of a consumer card), your time spent on maintenance, and the value of instant access to bigger GPUs when a job needs them. Rental prices also fall as new generations launch, which shortens the payback window for buyers. Launch MSRPs are pre-filled for consumer cards only; datacenter GPUs are sold through vendors without public list prices, so enter a quote you actually have.