NVIDIA·Datacenter
A100
NVIDIA's flagship data center GPU for AI training and HPC. Available in 40GB and 80GB HBM2e variants, widely used in cloud rental markets.
Cloud rental pricing
Live rental pricing is not available yet. This GPU is tracked — data will appear after the next sync from Vast.ai, Lambda Labs, or RunPod.
- MSRP
- $10,000
- Availability
- in stock
Hardware specifications
- Architecture
- Ampere
- Release
- May 2020
- Memory
- 40 / 80 GB
- Memory type
- HBM2e
- CUDA cores
- 6,912
- TDP
- 400W
Features
- Tensor Cores
- NVLink
- MIG partitioning
- FP64
- Multi-instance GPU
Use cases
- LLM training
- HPC
- Scientific simulation
- Large-scale inference
Strengths
- • Industry standard for AI
- • Massive memory bandwidth
- • Mature software stack
Considerations
- • High rental cost
- • Power hungry
- • Superseded by H100 for newest workloads
Price summary
No price history recorded yet.