GPU MODELS / MI350X
MI350X cloud GPU pricing
1 listings across 1 providers. Cheapest published rate: $4.000/GPU/hr at CloudRift. Sorted by per-GPU hourly price. Every row links to the provider's own published price.
WHAT IS THE MI350X
The MI350X is AMD's fourth-generation Instinct flagship: 288 GB of HBM3E at 8 TB/s with FP16 throughput at 2.3 PFLOPS, matching the MI355X's compute at a lower 1,000 W air-cooled power envelope. It is the direct competitor to NVIDIA's B200 for large-model training and inference, with AMD's usual pitch: more memory per GPU than the NVIDIA alternative, so fewer GPUs per model replica.
TENSOR COMPUTE
2.3 PFLOPS FP16/BF16 matrix (4.6 PFLOPS FP8, 9.2 PFLOPS MXFP4)
POWER DRAW
1,000 W (OAM, air-cooled; MI355X 1,400 W liquid)
INTERCONNECT
OAM (UBB), Infinity Fabric xGMI
WHAT IT IS GOOD AT
LLM training
The design target. 2.3 PFLOPS FP16 and 288 GB per GPU put it in the same class as the B200 for frontier-scale training, with ROCm as the software question rather than the hardware.
Large-model inference
Excellent fit - 288 GB holds a 70B model in FP8 with room to spare, or a 405B-class model across fewer GPUs than H100-class silicon.
HPC
Strong: 72.1 TFLOPS FP64 keeps it credible for double-precision workloads most AI SKUs abandoned.
WHO SHOULD NOT RENT THE MI350X
CUDA-locked software stacks that cannot move to ROCm - the hardware competes with B200, the ecosystem porting cost is the real line item.
Small or interactive workloads: a 1,000 W OAM module is rack-scale infrastructure, not a dev box.
| Provider | Config | VRAM | Location | $/hr | $/GPU/hr | Source |
|---|
 CloudRift | MI350X × 1 | 288 GB | Not published | $4.000 | $4.000 | receipt |
THE MI350X MARKET, BY THE NUMBERS
1 providers publish 1 MI350X configurations on the panel today. Published per-GPU hourly rates run from $4.000 (CloudRift) to $4.00 (CloudRift), a 1x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $4.00/GPU/hr. Rows are published catalog rates; stock flags appear per row where the provider reports them. All rates are on-demand - committed-use and reserved pricing is excluded.
MI350X PRICE TREND, DAILY PANEL SNAPSHOTS
Across every MI350X row on the panel, the daily floor moved from $3.650 to $4.000/GPU/hr over 10 days of tracking (+9.6%). Snapshots run daily; the full history is public in the repo.
MI350X PROVIDER NOTES
CloudRift · 1 configSets the panel floor for this model.
from $4.000/hrBUYING MI350X: WHAT THE PANEL SAYS
At the current floor, one MI350X running around the clock costs about $2,920 a month (730 hours of on-demand arithmetic). Compare the alternatives below before committing - the same budget often buys more than one chip class.
MI350X VS THE ALTERNATIVES
The generational jump: 288 GB vs 192 GB, 8 vs 5.3 TB/s, and roughly 1.8x the FP16 matrix throughput (2.3 vs 1.3 PFLOPS class). Where MI350X rents near MI300X on the panel, it is the clear pick; the older chip wins only on availability and price.
Same silicon, same compute figures - the MI355X is the 1,400 W direct-liquid-cooled variant that sustains higher clocks. Air-cooled clusters take the MI350X; dense liquid-cooled builds take the MI355X.
Compare live rates: MI350X vs MI300X · MI350X vs MI355X · MI350X vs B200 · MI350X vs MI325X
MI350X PRICING FAQ
What is the cheapest MI350X cloud GPU right now?
The cheapest published MI350X rate on the panel is $4.000 per GPU-hour at CloudRift (MI350X x 1, Not published). Every price links to the provider's own published rate.
How much does a MI350X cost per hour?
Across 1 providers tracked today, published per-GPU hourly rates for MI350X run from $4.00 to $4.00, with a median of $4.00. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent MI350X GPUs?
1 providers publish 1 distinct MI350X configurations on Compute Cafe. The panel shows published rates; stock flags appear per row where the provider reports them.
Are MI350X prices going up or down?
Over the 10 days Compute Cafe has tracked the panel, the MI350X floor is rising: $3.650 to $4.000/GPU/hr (+9.6%). Daily snapshots are public in the repo, so the trend is auditable.
What does a MI350X cost per month?
At the current floor of $4.000/GPU/hr, a MI350X running 24/7 costs about $2,920 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
How does the MI350X compare to the B200?
On paper it is a direct competitor: 288 GB HBM3E versus 192 GB, 8 TB/s versus 8 TB/s, and FP8 throughput in the same 4.5-4.6 PFLOPS dense class. The practical difference is software - CUDA versus ROCm - and rental availability, which is why the panel price matters more than the spec sheet.
MI350X or MI355X?
Same GPU, different cooling envelope. The MI350X runs at 1,000 W air-cooled, the MI355X at 1,400 W direct-liquid-cooled for sustained clocks. Choose by datacenter cooling, not by spec sheet.
See also alternatives to the MI350X · cheapest GPU-hour per model · inference chips · training chips · by location