NVIDIA A2 vs NVIDIA L40S — GPU Comparison (Apr 2026)
NVIDIA A2 (16GB GDDR6, 18 TFLOPS FP16, Ampere) vs NVIDIA L40S (48GB GDDR6, 366 TFLOPS FP16, Ada Lovelace). Cloud pricing: NVIDIA A2 from $0.22/hr, NVIDIA L40S from $0.55/hr. Compare specs, VRAM, performance, and pricing across 8 cloud providers to find the best GPU for your AI workload.
|
NVIDIA A2
16GB GDDR6 · Ampere
|
NVIDIA L40S
48GB GDDR6 · Ada Lovelace
|
||
|---|---|---|---|
| Specifications | |||
| Manufacturer | NVIDIA | NVIDIA | |
| Architecture | Ampere | Ada Lovelace | |
| VRAM | 16 GB GDDR6 | 48 GB GDDR6 | |
| Memory Bandwidth | 200 GB/s | 864 GB/s | |
| FP16 (Tensor) | 18.0 TFLOPS | 366.0 TFLOPS | |
| FP32 | 4.5 TFLOPS | 91.6 TFLOPS | |
| TDP | 60W | 350W | |
| Release Year | 2021 | 2023 | |
| Segment | Data center | Data center | |
| Best For | Edge inference entry-level AI | Inference fine-tuning video processing AI-accelerated graphics | |
| Cloud Pricing | |||
| Cheapest On-Demand | $0.22/hr | $0.55/hr | |
| Cheapest Spot | — | $0.79/hr | |
| Providers | 1 | 7 | |
| Provider Pricing (On-Demand) | |||
|
$0.22/hr | N/A | |
|
N/A | $0.55/hr | |
|
N/A | $0.55/hr | |
|
N/A | $0.61/hr | |
|
N/A | $0.86/hr | |
|
N/A | $0.86/hr | |
|
N/A | $1.57/hr | |
|
N/A | $1.67/hr | |
NVIDIA A2
NVIDIA L40S
Related GPU Comparisons
Top Providers for NVIDIA A2 and NVIDIA L40S
These 3 providers offer both NVIDIA A2 and NVIDIA L40S. Full head-to-head comparison of GPU models, pricing, infrastructure, and developer tools.
Cherry Servers vs Vast.ai vs Novita AI - GPU Provider Comparison (April 2026)
Side-by-side comparison of Cherry Servers vs Vast.ai vs Novita AI. Quickly compare GPU models, hourly pricing, spot instances, billing granularity, VRAM, regions, developer tools, Kubernetes support, and compliance to narrow down your cloud GPU provider shortlist. Data updated April 2026.
|
Cherry Servers
Bare metal GPU servers with 24 years of hosting experience and full hardware-level control.
|
Vast.ai
Instant GPUs. Transparent Pricing.
|
Novita AI
AI & Agent Cloud platform with 200+ model APIs, GPU instances, and serverless inference at scale.
|
|
|---|---|---|---|
| Overview | |||
| Trustpilot Rating | 4.6 | 4.4 | 3.3 |
| Headquarters | Lithuania | United States | United States |
| Provider Type | N/A | GPU Marketplace | GPU-Focused |
| Best For | AI training inference fine-tuning rendering research HPC generative AI deep learning | AI training inference fine-tuning Stable Diffusion batch processing research LLM serving generative AI | AI training inference fine-tuning generative AI research LLM serving Stable Diffusion |
| GPU Hardware | |||
| GPU Models | A100 A40 A16 A10 A2 Tesla P4 | B200 H200 H100 SXM H100 NVL A100 SXM A100 PCIe RTX 5090 RTX 5080 RTX 5070 Ti RTX 6000 Pro RTX 6000 Ada RTX 4500 Ada RTX A6000 RTX A5000 RTX A4000 L40S L40 A40 A10 RTX 4090 RTX 4080 RTX 4070 Ti RTX 4070 RTX 4060 Ti RTX 4060 RTX 3090 Ti RTX 3090 RTX 3080 Ti RTX 3080 RTX 3070 Ti RTX 3070 Tesla V100 Tesla T4 A2 GTX 1080 | H100 SXM A100 SXM L40S RTX 4090 RTX 6000 Ada RTX 5090 RTX 3090 |
| Max VRAM (GB) | 80 | 192 | 80 |
| Max GPUs/Instance | 2 | 8 | 8 |
| Interconnect | PCIe | NVLink, InfiniBand | NVLink |
| Pricing | |||
| Starting Price ($/hr) | $0.16/hr | $0.06/hr | $0.11/hr |
| Billing Granularity | Per-hour | Per-second | Per-second |
| Spot/Preemptible | No | Yes | Yes |
| Reserved Discounts | N/A | Up to 50% (1-6 month reserved) | N/A |
| Free Credits | None | Small test credit on signup | Up to $10,000 for startups |
| Egress Fees | N/A | Varies by host ($/TB) | None (Free) |
| Storage | NVMe SSD, Elastic Block Storage ($0.071/GB/mo) | Varies by host ($/GB/hr, charged while instance exists) | Container disk (60GB free), volume disk, network volumes |
| Infrastructure | |||
| Regions | Lithuania, Netherlands, Germany, Sweden, US, Singapore (6 locations) | 500+ locations, 40+ data centers | US, EU, APAC, South America, Africa, Middle East (20+ locations) |
| Uptime SLA | 99.97% | No formal SLA (host reliability scores visible) | 99.9% |
| Developer Experience | |||
| Frameworks | PyTorch TensorFlow CUDA (bare metal — full stack control) | PyTorch TensorFlow CUDA vLLM ComfyUI | PyTorch TensorFlow CUDA cuDNN TensorRT |
| Docker Support | Yes | Yes | Yes |
| SSH Access | Yes | Yes | Yes |
| Jupyter Notebooks | No | Yes | Yes |
| API / CLI | Yes | Yes | Yes |
| Setup Time | Minutes | Seconds | N/A |
| Kubernetes Support | Yes | No | No |
| Business Terms | |||
| Min Commitment | None | None | None |
| Compliance | ISO 27001 ISO 20000-1 GDPR PCI DSS | SOC 2 Type 2 HIPAA GDPR CCPA | SOC 2 |
Cherry Servers
Vast.ai
Novita AI
Build your own comparison
Select any 2-6 firms from this guide and open them in the full comparison table.
Tip: if you do not select any firms we will start with the top 2 from this guide.