Runpod
Our sponsor- On-Demand
- from $5.49
- Reserved
- on request
Air-cooled CDNA 4 accelerator for standard data center racks.
Weekly median price per GPU per hour · Get the data
By provider shows one card per company, placed where its best offer ranked. By configuration lists every offer.
What you can rent comes first: in stock, then waitlist, then not reported, then out of stock. Priced offers rank ahead of quote-only, and on-demand ahead of other billing types.
Within a group, five factors set the order:
We match the provider, its country, GPU model, form factor, billing type, availability and instance name. Matching is partial and case-insensitive. Everyday words work too, so "interruptible" finds spot and "sold out" finds out of stock.
No providers match .
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Verify before provisioning. More on how we price.
One MI350X has 288 GB of VRAM. In practice, that's enough memory for roughly 460B parameters at 4-bit or 125B at 16-bit, assuming a 32K context. Below are some open-weight LLMs, with the estimated memory and GPUs each one needs.
| Model | Memory (INT4 / FP4) | MI350Xs needed | Cost /hr | Cost /mo |
|---|---|---|---|---|
|
|
18 GB
|
1
|
–
|
–
|
|
|
22 GB
|
1
|
–
|
–
|
|
|
67 GB
|
1
|
–
|
–
|
|
|
178 GB
|
1
|
–
|
–
|
|
|
239 GB
|
1
|
–
|
–
|
|
|
306 GB
|
8
|
–
|
–
|
|
|
1,544 GB
|
8
|
–
|
–
|
Estimates based on the median on-demand rate. Memory is weights plus FP8 KV cache at 32K context per request (FP16/BF16 in the 16-bit column). GPU counts assume 90% of advertised VRAM is usable. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.
AMD MI350X · Per GPU
| Compute · dense | |
|---|---|
| FP4 | 9,227.5 TFLOPS |
| FP8 | 4,614 TFLOPS 9,227.4 with sparsity |
| FP16 / BF16 | 2,309.6 TFLOPS 4,619.2 with sparsity |
| INT8 | 4,614 TOPS 9,227.4 with sparsity |
| FP32 | 144.2 TFLOPS |
| FP64 | 72.1 TFLOPS |
| Precision support | FP4FP6FP8FP16BF16FP32FP64INT8 |
| Memory | |
|---|---|
| Capacity | 288 GB HBM3e |
| Bandwidth | 8,000 GB/s |
| Bus width | 8,192-bit |
| ECC | Yes |
| Silicon | |
|---|---|
| Architecture | CDNA 4 |
| Process | TSMC 3nm / 6nm FinFET |
| Transistors | 185 billion |
| Shader cores | 16,384 Stream processors |
| Matrix cores | 1,024 |
| Compute units | 256 |
| Fabric and host | |
|---|---|
| GPU interconnect | Infinity Fabric 1,075 GB/s |
| Host interface | PCIe 5.0 x16 |
| Power | |
|---|---|
| Board power | 1,000 W |
| Cooling | Passive |
| Platform | |
|---|---|
| Partitioning | Up to 8 partitions |
Source: official AMD MI350X datasheet.
As of November 19, 2026, we track 8 configs from 5 providers. Prices are per GPU per hour.
| Billing type | Configs | Cheapest |
|---|---|---|
On-demand | 4 | $5.49 (in stock, Runpod) |
Reserved | 3 | $3.89 (not verified in stock, Cirrascale) |
Spot | 1 | $3.00 (not verified in stock, DigitalOcean) |
No median for on-demand, reserved and spot: we only show this when at least 3 providers list the GPU on that billing type.
The cheapest verified in-stock on-demand MI350X tracked by GetDeploying is $5.49 per GPU per hour from Runpod. Other billing types: reserved $3.89 (not verified in stock) from Cirrascale and spot $3.00 (not verified in stock) from DigitalOcean.
The cheapest verified in-stock estimate is $3,953 per month on-demand.
GetDeploying currently tracks MI350X configs from 5 providers. Runpod has verified in-stock on-demand configs. See the full price comparison above for every provider and config.
One MI350X runs models up to roughly 460B parameters at 4-bit quantization or 125B at 16-bit, assuming a 32K context. Larger models run across multiple GPUs: the model table above shows the estimated memory and GPU count for popular open-weight LLMs.
288GB HBM3e with MXFP6 and MXFP4 support on CDNA 4, in a passive OAM module at 1000W that drops into air-cooled racks. The same memory and bandwidth as the MI355X without a liquid loop.
Clocked lower and held to a smaller power envelope than the MI355X, which returns more throughput per GPU where direct liquid cooling is available. The ROCm software stack, while rapidly improving, is narrower than CUDA, so check framework support first.
Last updated