In December 2023, The AMD Instinct MI300X was released as the "H100 killer" for its massive 192 GB of VRAM. However, its actual value depends on your project.
This guide compares current MI300X on-demand pricing reviewed weekly.
Takeaways
- TensorWave advertises MI300X starting from $1.71/hr, but the real rate depends on a sales quote.
- Runpod and DigitalOcean are the next lowest listed options at $2.39/hr and $2.59/hr.
- At $3.20/hr, Thunder Compute H100s remain a strong CUDA-first alternative.
Current on-demand MI300X prices (per GPU)
| Provider | GPU / Instance | On-Demand $/GPU-hr* | Notes |
|---|---|---|---|
| TensorWave | N/A | $1.71 | Quote-only. Normalized from an 8-GPU node. |
| Runpod | MI300X | $2.39 | |
| DigitalOcean | MI300X | $2.59 | |
| Crusoe Cloud | MI300X | $3.45 | |
| Vultr | MI300X | $3.99 | |
| Azure | Standard_ND96isr_MI300X_v5 | $6.00 | Normalized from an 8-GPU node. Cheapest US region (eastus2); ranges $6.00-$7.86/GPU-hr across 6 US options. |
| Oracle Cloud | BM.GPU.MI300X.8 | $6.00 | |
| Last reviewed on September 11, 2026. | |||
Methodology
Why you can trust these numbers:
- On-demand only - no reservations or prepaid contracts.
- Same silicon - every figure refers to AMD MI300X (192 GB) SKUs.
- Public price lists - pulled in September 2026 from each provider's published pricing pages.
- U.S. regions, USD - prices elsewhere vary by 5-20 %.
Azure ND MI300X Instances
Through the ND_MI300X_v5-series, Azure offers hyper-scale virtual machines designed for high-end Deep Learning training and tightly coupled, scale-up and scale-out Generative AI or HPC workloads.
Powered by AMD Instinct MI300X accelerators and 4th Gen Intel Xeon Scalable processors, these instances provide immense high-bandwidth memory (HBM) capacity, making it possible to fit larger LLMs into a single node.
| Spec | Standard_ND96isr_MI300X_v5 |
|---|---|
| Provider | Azure |
| MI300X GPUs | 8 |
| vCPUs | 96 |
| RAM | 1,850 GiB |
| Hourly price | $48.00 |
| Price per-GPU | $6.00 |
Why this matters for developers
- More memory, new math - MI300X packs 192 GB HBM3 and FP8 support, ideal for 70B+ parameter models.
- H100 availability still matters - Unless your model truly needs 192 GB or AMD's MI300X architecture, an available H100 80 GB at $3.20/hr can still be the better CUDA-first workflow.
- Bookmark this page - we review the numbers weekly so you don't have to.
Get the full picture in our AI GPU rental market trends overview.
What is the cheapest AMD MI300X price per hour?
The cheapest AMD MI300X GPU pricing in this comparison is $1.71/hr from TensorWave, followed by Runpod at $2.39 per GPU-hour and DigitalOcean at $2.59 per GPU-hour.
How much does Azure charge for MI300X GPUs?
Azure charges about $6.00/hr per MI300X GPU for its ND MI300X v5 instances when normalized from the 8-GPU VM.
How much memory does the AMD MI300X GPU have?
The AMD MI300X has 192 GB of HBM3 memory, making it ideal for running large language models with 70 billion or more parameters. This high memory capacity is one of its key advantages for AI and machine learning workloads.
Which cloud providers offer AMD MI300X GPUs?
AMD MI300X GPUs are available from cloud providers including TensorWave ($1.71/hr), Runpod ($2.39/hr), DigitalOcean ($2.59/hr), Crusoe Cloud ($3.45/hr), Vultr ($3.99/hr), Azure ($6.00/hr), and Oracle Cloud ($6.00/hr).
What is the cost to run MI300X for 2 hours?
Running an AMD MI300X GPU for 2 hours costs anywhere from $3.42 (TensorWave at $1.71/hr) to $12.00 (Azure or Oracle Cloud at $6.00/hr). These are on-demand rates without any reservations or prepaid contracts