B300
NVIDIA · Blackwell Ultra
288 GBHBM3e
8 TB/sBandwidth
4.5 PFFP8 dense
$4,899 /GPU/mo$6,552 · −25%
AMD · CDNA 3 · OAM · 8-GPU UBB
Rent the AMD Instinct MI325X by the month.
The AMD Instinct MI325X carries 256 GB of HBM3E at 6 TB/s with 2.6 PF of dense FP8 on CDNA 3. Its strength is memory: a 405B-parameter model fits on four GPUs at FP8.
256 GBHBM3E per GPU
6 TB/sMemory bandwidth
2.6 PFFP8 dense
Infinity FabricInterconnect
Why the MI325X
What the MI325X is best at
- 256 GB HBM3E. 1.8× the memory of an H200 per GPU.
- 6 TB/s bandwidth. Fast enough for large-batch, memory-bound serving.
- 2.6 PF FP8 dense. More FP8 throughput than H100 or H200.
- Large-model inference (400B+)
- Long-context serving
- Fine-tuning with large batches
MI325X is currently sold out; the next allocation date is shown on this page.
Pricing
MI325X monthly price by server size
One flat price per GPU, the same from 1 to 8 GPUs. The market median is the public on-demand list price of specialist GPU clouds × 730 hours.
| Server | CoreGPU / month | Market median / month | You save / month | You save / year | ≈ per GPU-hour |
|---|---|---|---|---|---|
| 1× MI325X | $2,299 | $772 | $9,264 | $3.15 | |
| 2× MI325X | $4,598 | $1,544 | $18,528 | $3.15 | |
| 4× MI325X | $9,196 | $3,088 | $37,056 | $3.15 | |
| 8× MI325X | $18,392 | $6,176 | $74,112 | $3.15 |
Prices in USD, excluding taxes. Market median checked Nov 4, 2026 across specialist GPU clouds; hyperscalers excluded. All GPU prices
Compare
MI325X vs H200
MI325X offers 256 vs 141 GB and more FP8 throughput; H200 wins on CUDA-only tooling.
| MI325X | H200 | |
|---|---|---|
| Architecture | CDNA 3 | Hopper |
| GPU memory | 256 GB HBM3E | 141 GB HBM3e |
| Memory bandwidth | 6 TB/s | 4.8 TB/s |
| FP8 dense | 2.6 PF | 1.98 PF |
| FP4 dense | n/a | n/a |
| Interconnect | Infinity Fabric · 896 GB/s | NVLink 4 · 900 GB/s |
| Power (TDP) | 1,000 W | 700 W |
| Price per GPU / month | $2,299 | $2,699 |
| Price per GB of memory | $8.98 | $19.14 |
| Price per FP8 PFLOPS | $884 | $1,363 |
| Availability | Back Dec 2 | 34 available |
Specifications
AMD Instinct MI325X technical specs
Per-GPU figures from the vendor datasheet. Tensor throughput is dense (without sparsity).
| Model | AMD Instinct MI325X |
|---|---|
| Architecture | CDNA 3 |
| Form factor | OAM · 8-GPU UBB |
| GPU memory | 256 GB HBM3E |
| Memory bandwidth | 6 TB/s |
| FP8 Tensor (dense) | 2.6 PF |
| FP4 Tensor (dense) | Not supported |
| Interconnect | Infinity Fabric · 896 GB/s |
| Power (TDP) | 1,000 W |
| GPUs per server | 1 to 8 |
| Availability | Sold out · next allocation Dec 2 |
FAQ
MI325X rental: common questions
Another question? Our engineers answer within one business day.
Contact usHow much does it cost to rent an AMD Instinct MI325X per month?
$2,299 per GPU per month at CoreGPU, about $3.15 per GPU-hour, versus a market median of $3,071 ($4.21/hour × 730 h). A full 8-GPU server is $18,392 per month, taxes excluded.
Is the MI325X available right now?
Not right now. The current allocation is sold out and the next one opens on December 2, 2026. Join the waitlist and we email you before units go back on public sale.
How many MI325X GPUs can I rent?
From 1 to 8 GPUs per server, at the same price per GPU. Servers with 8 GPUs come with the full Infinity Fabric mesh. Need more? Rent several servers or ask for a private allocation.
MI325X or H200?
MI325X offers 256 vs 141 GB and more FP8 throughput; H200 wins on CUDA-only tooling. See the full MI325X vs H200 comparison.
What do I get with a MI325X server?
Dedicated GPUs (never shared or sliced), root SSH access, your choice of a clean OS or a ready image with ROCm, PyTorch and vLLM, local NVMe scratch, and month-to-month billing with no setup fee.
Alternatives
Other GPUs in stock now
Get the MI325X restock email first
The next allocation opens December 2. Waitlist members are emailed before public release.