Skip to content
GPU Finder

AMD MI355X pricing and cloud availability

No single MI355X has fresh confirmed self-service on-demand stock. Check the listings for separately priced spot offers. Unknown or stale stock is not evidence of unavailable capacity. The on-demand catalog has no listings at this size.

How this stock answer is checked

Evaluated . Stock checks must be no more than 6 hours old; limited stock counts as in stock. Unknown, stale, out-of-stock and sales-only offers do not qualify. Confirm capacity with the provider before renting.

No in-stock listing selected. Historical availability can still be inspected below.

Stock and history methodology. Inspect listings and stock filters.

Compare AMD MI355X cloud rentals by GPU count, region and reported stock. Listings are still early, so check ROCm support, whole-node cost and provider requirements before renting.

Stock checks run regularly where supported; collection delays can occur and the cheapest in-stock offer can change between checks. Set a price alert to get an email when the MI355X price drops.

In-stock providers (all sizes):0
Catalog providers (all sizes):3
On-demand whole-node range:No matching offers
Instances:0

Page headline prices reflect the default offers: fresh reported stock, including limited stock, with self-service access.

MI355X rental prices by provider

Offer visibility

Choose whether to include out-of-stock or unconfirmed offers. Limited stock is included. Reported availability is not a capacity guarantee.

  • No offers match these visibility filters. Include unknown, out-of-stock or sales-only offers above to broaden the results.

Get notified when MI355X drops in price or reliability changes

One email per change, max once a day. We send a confirmation link first; one-click unsubscribe in every email.

Renting an AMD MI355X: what to check first

The AMD Instinct MI355X has 288 GB of HBM3E memory per GPU. Before choosing a rental, check your software stack, the number of GPUs you need and the exact configuration offered in your region.

Single GPU or a whole node?
Check the GPU count in the offer. A whole-node quote pays for every GPU in that node. The per-GPU equivalent helps compare rates, but it does not mean the provider rents a fraction of the server. Several GPUs also need software and a topology that can share the workload.
Verify ROCm support before a longer commitment
Check that your framework, ROCm release, container and required extensions support MI355X. A model that runs on another accelerator is not automatically validated on this one. Record the exact environment when testing so a working setup is reproducible.
Estimate the rental budget
Multiply the whole-node hourly rate by the hours you expect to keep the instance running, then add storage, transfer and minimum billing commitments. A cheaper hourly rate does not establish a cheaper completed job.
Size memory and confirm stock
288 GB does not guarantee that a model fits: weights are only part of memory use, so include runtime overhead, cache and intermediate state for your precision, context and concurrency. A listing does not reserve capacity, and unknown stock means capacity has not been established, not that the provider is sold out.

Hardware reference: AMD Instinct MI355X specifications. This does not verify any rental host's configuration.

Oracle documents an eight-MI355X bare-metal shape. That establishes a supported configuration, not current launch capacity for your account. Oracle MI355X configuration guide.

Budget example: a node priced at $40 per hour costs $240 for six billed hours before extras. This is arithmetic, not a current MI355X quote. Measure runtime on your workload before estimating a longer job.

MI355X rental questions

Why might there be no single-GPU MI355X quote?
Providers may offer larger node sizes, spot-only capacity or listings without fresh stock. Inspect GPU-count groups and stock labels separately. An empty single-GPU on-demand group does not establish that no MI355X capacity exists.
Does 288 GB guarantee my model will fit?
No. Weights are only part of memory use. Include runtime overhead, cache and intermediate state, with explicit assumptions for precision, context length and concurrency.
Is MI355X cheaper than B300?
Compare current offers for the same GPU count and rental terms after checking software compatibility. The cheaper hourly rate does not establish the cheaper completed job; measure your workload on the actual configurations.
Does a listing reserve capacity?
No. Confirm stock, account access, quotas and rental terms with the provider before starting. Unknown stock means capacity has not been established, not that the provider is sold out.
Compare B300 and MI355X rentals →

Compare listings and visit a provider →Estimate compute and egress →

MI355X (1× GPU) availability history — last 7 days

No provider held 1× MI355X ≥80% fully available across all 7 days. Polled hourly · 1 provider tracked · 7-day window.

≥80%50–80%<50%No data

Percentages show tracked time reported fully available. Limited stock can still be rented, but does not count as fully available here. Missing data is not 0%; this history is not machine uptime.

View:
Provider
13
14
15
16
17
18
19
Week
Digital Ocean71%

Building stock history

Stock tracking is supported for MI355X, but there are not enough retained observations yet to publish 30-day stock-history percentages. Percentages require at least 48 tracked listing-hours. Check each listing's observation time; missing stock data means unknown availability.

MI355X vs MI300X pricing watch

MI355X cloud inventory is still emerging, so this page tracks MI355X GPU providers and live rates as they appear. For the nearest AMD Instinct price baseline today, compare against current MI300X cloud pricing.

How does the MI355X compare on price? The cheapest cloud GPU prices per hour rank every model we track by its lowest per-GPU-hour rate and the cheapest listing in stock. For the most-compared training GPU, see H100 cloud pricing across every provider.

Providers listing this GPU

Related GPUs

Compare head-to-head