Compact single-slot professional GPU for entry-level AI and visualization.
Barriers to entry are low here. 10 cloud providers list the A4000 with prices starting at just $0.08/hr per GPU (on-demand), making it accessible for testing without a large commitment. Spot instances start lower, at $0.03/hr per GPU.
USA 3 configs 1x-8x
USA 10 configs 1x-6x
USA 1 config 1x
UAE 6 configs 1x-10x
France 3 configs 1x-4x
USA 1 config 1x
UK 2 configs 1x
USA 1 config 1x
USA 8 configs 1x
USA 4 configs 8x
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.
16GB GDDR6, compact single-slot professional GPU. Low power consumption. Suitable for entry-level AI development and visualization.
No NVLink. 16GB VRAM limits model sizes. For dedicated cloud AI inference, the L4 offers better throughput at lower cost.
The median on-demand price across providers has risen about 5% since August 2025, from $0.20 to $0.21/hr per GPU. This reflects the market-wide median, which moves both when providers change prices and when lower or higher priced offerings enter the market.
With 16GB of VRAM, the A4000 is best for 7B-class models in 4-bit or 8-bit quantized form, and smaller models in FP16.
The A4000 has 16GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.
The A4000 has 448 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.
The A4000 supports 6 precision formats. Training: BF16, FP16, TF32, FP32. Inference: INT4, INT8.
No. The A4000 is a PCIe-only GPU with no NVLink, so it is better suited to single-GPU inference and smaller-scale workloads than large distributed training jobs.
A4000 pricing currently ranges from $0.03/hr to $0.88/hr per GPU, depending on the provider, instance type, and billing model.
At 720 hours per month, one A4000 can cost between $22.34 to $633.60 per month, depending on the provider. Reserved and spot pricing can lower that further.
The A4000 is available from 10 cloud providers, including Cirrascale, Runpod, Sesterce. Pricing and availability vary by region and billing model.
Yes. We currently track 39 A4000 listings across 10 cloud providers:
| Billing type | Listings | Avg $/GPU/hr |
|---|---|---|
| On-demand | 19 | $0.32/hr |
| Reserved | 16 | $0.22/hr |
| Spot | 4 | $0.05/hr |
| GPU Memory | 16GB GDDR6 |
| Memory Interface | 256-bit |
| Memory Bandwidth | 448 GB/s |
| Error-Correcting Code (ECC) | Yes |
| NVIDIA Ampere Architecture-based CUDA Cores | 6,144 |
| NVIDIA Third-Generation Tensor Cores | 192 |
| NVIDIA Second-Generation RT Cores | 48 |
| Single-Precision Performance | 19.2 TFLOPS |
| RT Core Performance | 37.4 TFLOPS |
| Tensor Performance | 153.4 TFLOPS |
| System Interface | PCIe 4.0 x16 |
| Power Consumption | Total board power: 140 W |
| Thermal Solution | Active |
| Form Factor | 4.4” H x 9.5” L, single slot |
| Display Connectors | 4x DisplayPort 1.4a |
| Max Simultaneous Displays | 4x 4096 x 2160 @ 120 Hz, 4x 5120 x 2880 @ 60 Hz, 2x 7680 x 4320 @ 60 Hz |
| Power Connector | 1x 6-pin PCIe |
| Encode/Decode Engines | 1x encode, 1x decode (+AV1 decode) |
| VR Ready | Yes |
| Graphics APIs | DirectX 12 Ultimate, Shader Model 6.6, OpenGL 4.6, Vulkan 1.3 |
| Compute APIs | CUDA 11.6, DirectCompute, OpenCL 3.0 |
Source: official Nvidia A4000 datasheet.
Higher-end with 50% more memory. From $0.11/hr per GPU across 11 providers.
Datacenter inference GPU. From $0.06/hr per GPU across 9 providers.
Single-slot Ada professional card for desktop AI and rendering. From $0.16/hr per GPU across 8 providers.
Ampere professional card for visualization and AI development. From $0.19/hr per GPU across 2 providers.
Last updated
