How much does an idle NVIDIA T4 cost in the cloud?
The lowest per-GPU rate for the NVIDIA T4 is on g4dn.xlarge on AWS (us-east-2), at $0.53/GPU-hr on-demand. That whole box left running around the clock costs about $384/mo. Scheduling it off outside business hours saves up to $270/mo (est.).
Cheapest per GPU
$0.53/GPU-hr
g4dn.xlarge · us-east-2
Whole instance 24/7
$384/mo
730-hour month
Saved with a schedule (up to, est.)
$270/mo
up to 70% off the 24/7 bill (est.)
Instances carrying the NVIDIA T4
| Instance | Provider | GPUs | vCPU / Memory | $/hr | $/GPU-hr | Spot $/GPU-hr |
|---|---|---|---|---|---|---|
| g4dn.xlarge | AWS | 1× GPU | 4 vCPU · 16 GB | $0.526/hr us-east-2 | $0.53 | $0.17 |
| g4dn.2xlarge | AWS | 1× GPU | 8 vCPU · 32 GB | $0.752/hr us-east-2 | $0.75 | $0.24 |
| g4dn.12xlarge | AWS | 4× GPU | 48 vCPU · 192 GB | $3.912/hr us-east-2 | $0.98 | $0.28 |
| g4dn.4xlarge | AWS | 1× GPU | 16 vCPU · 64 GB | $1.204/hr us-east-1 | $1.20 | $0.39 |
| g4dn.8xlarge | AWS | 1× GPU | 32 vCPU · 128 GB | $2.176/hr us-east-1 | $2.18 | $0.65 |
| g4dn.16xlarge | AWS | 1× GPU | 64 vCPU · 256 GB | $4.352/hr us-east-2 | $4.35 | $0.96 |
About the NVIDIA T4
The T4 is the fleet's elder statesman of cheap GPU compute: 16 GB of VRAM, available on AWS as g4dn, and for years the box teams reached for when the job needed a GPU but not a big one. A large installed base of inference services, video pipelines, and dev boxes still runs on it, often provisioned years ago.
Age is the tell. T4 boxes predate most teams' current cost discipline, and many run 24/7 simply because they always have. For workloads still fitting the T4, the cheapest upgrade is usually not newer silicon — it is fewer idle hours on the silicon already provisioned. For workloads that have outgrown it, the L4 tier offers a meaningful step up in efficiency.
Common questions
Should old T4 workloads migrate to the L4?
If the workload is throughput-bound or memory-tight, the L4 delivers considerably more per dollar and the migration usually pays back. If the T4 still handles the workload comfortably, migration buys little — an off-hours schedule on the existing box is the higher-leverage change.
Why do T4 instances so often run around the clock?
Inertia: they were provisioned before anyone was watching, they are cheap enough to escape per-box scrutiny, and nothing forces a decision. That combination is exactly what an automatic lease expiry fixes — the default flips from "runs forever" to "runs while someone is actually using it".
Other GPU tiers
Stop paying for idle cloud GPU servers
Idlefy runs a free audit of your fleet, then flips dev servers to stopped-by-default: lease one when you need it, and
when the lease ends it shuts itself off — stop, not terminate, persistent disks intact. Timers are set by your team, never
guessed, and Idlefy only manages instances tagged idlefy=enabled — on AWS that boundary is enforced by the IAM policy itself.
Prices are on-demand list price for Linux (AWS) / standard (GCP) instances and exclude storage, network, data transfer, and any negotiated or volume discounts. Actual bills vary by usage and account-level pricing. Spot figures are indicative — AWS values are derived from AWS's published typical savings vs on-demand for the region; GCP values are the listed Spot price. Spot capacity can be reclaimed by the provider. Stopping preserves EBS and persistent-disk volumes; data on local instance-store or local-SSD drives does not survive a stop. Prices as of 2026-08-31.