MI300X cloud GPU pricing

Q: What is the cheapest MI300X cloud GPU right now?

The cheapest published MI300X rate on the panel is $0.500 per GPU-hour at RunPod (MI300X x 1, Global). Every price links to the provider's own published rate.

Q: How much does a MI300X cost per hour?

Across 13 providers tracked today, published per-GPU hourly rates for MI300X run from $0.50 to $12.00, with a median of $7.20. Committed-use and reserved rates are excluded; everything shown is on-demand.

Q: How many providers rent MI300X GPUs?

13 providers publish 48 distinct MI300X configurations on Compute Cafe. 44 of those rows were reported in stock by the provider's live endpoint at last observation.

Q: Where is MI300X cheapest?

By published per-GPU hourly rate, the cheapest geography for MI300X today is italy at $2.68/hr (Seeweb). The spread across 14 tracked geographies runs to $12.00/hr in brazil.

Q: Are MI300X prices going up or down?

Over the 11 days Compute Cafe has tracked the panel, the MI300X floor is holding flat: $0.500 to $0.500/GPU/hr (+0.0%). The median continuously-tracked listing moved from $3.45 to $3.45 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.

Q: What does a MI300X cost per month?

At the current floor of $0.500/GPU/hr, a MI300X running 24/7 costs about $365 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.

Q: Is a full 8-GPU MI300X node cheaper than a single card?

On today's panel: the single-GPU floor is $0.500/GPU/hr and the 8-GPU node floor is $1.750/GPU/hr - more expensive per GPU on a full node (+250.0%). Check the table for the exact configs behind each floor.

Q: Is ROCm ready for production?

For inference on vLLM/PyTorch and standard training frameworks, yes - major providers run MI300X fleets in production. Custom CUDA kernels and niche libraries are the gaps; verify your specific stack.

Q: Why is MI300X often cheaper than H100?

Supply and software moat: AMD prices for adoption, and buyers pay a premium for CUDA certainty. If your workload runs on ROCm, that discount is real savings, not a catch.

48 listings across 13 providers. Cheapest published rate: $0.500/GPU/hr at RunPod. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE MI300X

The MI300X is AMD's serious answer to Hopper: more memory (192 GB) and more bandwidth (5.3 TB/s) than any NVIDIA part before Blackwell, at panel prices that frequently undercut H100. ROCm maturity is the real question - for well-supported frameworks (vLLM, PyTorch) it serves large models beautifully; for exotic CUDA kernels, check before you commit.

MAKER

AMD

ARCHITECTURE

CDNA 3

LAUNCHED

2023

MEMORY

192 GB HBM3

MEMORY BANDWIDTH

5.3 TB/s

TENSOR COMPUTE

1,307 TFLOPS FP16/BF16 dense (2,614 sparse); 2,614 TFLOPS FP8 dense

POWER DRAW

750 W

INTERCONNECT

Infinity Fabric 896 GB/s

WHAT IT IS GOOD AT

Large-model inference

The design case: 192 GB serves BF16 70B with room, or quantized 180B-class models, per GPU. vLLM support is production-grade.

Memory-bound training

5.3 TB/s and 192 GB help anywhere HBM capacity is the constraint.

Cost-optimized clusters

Where ROCm covers your stack, per-dollar memory is the best on the panel.

WHO SHOULD NOT RENT THE MI300X

Stacks with custom CUDA kernels or libraries without ROCm ports.

Multi-node training where NVLink-class fabric maturity matters most.

Provider	Config	VRAM	Location	$/hr	$/GPU/hr	Source
RunPod	MI300X × 1	192 GB	Global	$0.500	$0.500	receipt
SpotGPUs	MI300X × 1	192 GB	Not published	$1.450	$1.450	receipt
Vultr	MI300X (bare metal) × 8	192 GB	Not published	$14.000	$1.750	receipt
Vultr	MI300X × 8	192 GB	Not published	$14.800	$1.850	receipt
DigitalOcean	MI300X × 1	192 GB	Not published	$2.590	$2.590	receipt
Seeweb	MI300X × 1	192 GB	Italy	$2.678	$2.678	receipt
M Machine0	MI300X × 1	192 GB	Not published	$2.849	$2.849	receipt
Cyfuture	MI300X × 1	192 GB	India	$2.858	$2.858	receipt
Hot Aisle	MI300X × 1	192 GB	Not published	$2.990	$2.990	receipt
Hot Aisle	MI300X × 8	192 GB	Not published	$27.120	$3.390	receipt
Crusoe	MI300X × 1	192 GB	US / Iceland	$3.450	$3.450	receipt
Core42	MI300X (from) × 1	192 GB	UAE	$3.500	$3.500	receipt
Spheron	MI300X × 1	192 GB	Not published	$3.590	$3.590	receipt
Oracle OCI	MI300X × 1	192 GB	Not published (global list)	$6.000	$6.000	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US East 2	$48.000	$6.000	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US West 3	$48.000	$6.000	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US East 2	$48.000	$6.000	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US West 3	$48.000	$6.000	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US North Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	EU North	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	CA Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US West Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	EU North	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	CA Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US North Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US West Central	$57.600	$7.200	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US Central	$59.041	$7.380	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US Central	$59.041	$7.380	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	UK South	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	UK South	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	FR Central	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	FR Central	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	KR South	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	KR South	$60.000	$7.500	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	DE West Central	$62.400	$7.800	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	DE West Central	$62.400	$7.800	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	US West	$62.853	$7.857	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	US West	$62.853	$7.857	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	AU Southeast	$67.200	$8.400	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	AU Southeast	$67.200	$8.400	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	SE Central	$67.200	$8.400	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	SE Central	$67.200	$8.400	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	AE North	$68.640	$8.580	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	AE North	$68.640	$8.580	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	PL Central	$73.920	$9.240	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	PL Central	$73.920	$9.240	receipt
Microsoft Azure	MI300X [ND96isr_MI300X_v5] × 8	192 GB	BR South	$96.000	$12.000	receipt
Microsoft Azure	MI300X [ND96is_MI300X_v5] × 8	192 GB	BR South	$96.000	$12.000	receipt

THE MI300X MARKET, BY THE NUMBERS

PROVIDERS

13

CONFIGURATIONS

48

MEDIAN $/GPU/HR

$7.20

SPREAD

24.0x

13 providers publish 48 MI300X configurations on the panel today. Published per-GPU hourly rates run from $0.500 (RunPod) to $12.00 (Microsoft Azure), a 24x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $7.20/GPU/hr. 44 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.

CHEAPEST MI300X BY GEOGRAPHY

italySeeweb$2.678/hr indiaCyfuture$2.858/hr united statesCrusoe$3.450/hr icelandCrusoe$3.450/hr uaeCore42$3.500/hr canadaMicrosoft Azure$7.200/hr united kingdomMicrosoft Azure$7.500/hr franceMicrosoft Azure$7.500/hr

MI300X PRICE TREND, DAILY PANEL SNAPSHOTS

Across every MI300X row on the panel, the daily floor held flat at $0.500/GPU/hr over 11 days of tracking. The median continuously-tracked listing held flat at $3.45. Snapshots run daily; the full history is public in the repo.

MI300X PROVIDER NOTES

RunPod · 1 config · 1 region · 1 in stock

Sets the panel floor for this model.

from $0.500/hr

SpotGPUs · 1 config · 1 in stock

1 of 1 configs reported in stock at last check.

from $1.450/hr

Vultr · 2 configs · 2 in stock

2 of 2 configs reported in stock at last check.

from $1.750/hr

DigitalOcean · 1 config · 1 in stock

1 of 1 configs reported in stock at last check.

from $2.590/hr

Seeweb · 1 config · 1 region · 1 in stock

1 of 1 configs reported in stock at last check.

from $2.678/hr

Machine0 · 1 config · 1 in stock

1 of 1 configs reported in stock at last check.

from $2.849/hr

Cyfuture · 1 config · 1 region · 1 in stock

1 of 1 configs reported in stock at last check.

from $2.858/hr

Hot Aisle · 2 configs · 2 in stock

2 of 2 configs reported in stock at last check.

from $2.990/hr

BUYING MI300X: WHAT THE PANEL SAYS

At the current floor, one MI300X running around the clock costs about $365 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $0.500/GPU/hr against $1.750/GPU/hr on an 8-GPU node (+250.0%). Per-GPU floors by config size: 1x $0.50 · 8x $1.75. Compare the alternatives below before committing - the same budget often buys more than one chip class.

MI300X VS THE ALTERNATIVES

MI300X vs H100 SXM

MI300X has 2.4x the memory and 1.6x the bandwidth; H100 has CUDA, NVLink and broader software support. For inference on mainstream stacks the AMD part often wins on cost; for training at scale, NVIDIA's fabric still leads.

MI300X vs H200 SXM

H200 narrows the memory gap (141 GB, 4.8 TB/s) with CUDA maturity. The decision is ROCm coverage plus the live price spread.

Compare live rates: MI300X vs A100 80GB SXM · MI300X vs H100 SXM · MI300X vs H200 SXM · MI300X vs MI355X · MI300X vs MI325X · MI300X vs MI350X · MI300X vs B200

MI300X PRICING FAQ

What is the cheapest MI300X cloud GPU right now?

The cheapest published MI300X rate on the panel is $0.500 per GPU-hour at RunPod (MI300X x 1, Global). Every price links to the provider's own published rate.

How much does a MI300X cost per hour?

Across 13 providers tracked today, published per-GPU hourly rates for MI300X run from $0.50 to $12.00, with a median of $7.20. Committed-use and reserved rates are excluded; everything shown is on-demand.

How many providers rent MI300X GPUs?

13 providers publish 48 distinct MI300X configurations on Compute Cafe. 44 of those rows were reported in stock by the provider's live endpoint at last observation.

Where is MI300X cheapest?

By published per-GPU hourly rate, the cheapest geography for MI300X today is italy at $2.68/hr (Seeweb). The spread across 14 tracked geographies runs to $12.00/hr in brazil.

Are MI300X prices going up or down?

Over the 11 days Compute Cafe has tracked the panel, the MI300X floor is holding flat: $0.500 to $0.500/GPU/hr (+0.0%). The median continuously-tracked listing moved from $3.45 to $3.45 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.

What does a MI300X cost per month?

At the current floor of $0.500/GPU/hr, a MI300X running 24/7 costs about $365 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.

Is a full 8-GPU MI300X node cheaper than a single card?

On today's panel: the single-GPU floor is $0.500/GPU/hr and the 8-GPU node floor is $1.750/GPU/hr - more expensive per GPU on a full node (+250.0%). Check the table for the exact configs behind each floor.

Is ROCm ready for production?

For inference on vLLM/PyTorch and standard training frameworks, yes - major providers run MI300X fleets in production. Custom CUDA kernels and niche libraries are the gaps; verify your specific stack.

Why is MI300X often cheaper than H100?

Supply and software moat: AMD prices for adoption, and buyers pay a premium for CUDA certainty. If your workload runs on ROCm, that discount is real savings, not a catch.