RTX A6000
Workstation-class training and inference.
from $0.33/hr Market rate, 17 Nov 2026. Not our price.
GPU capacity desk
We help firms book in-country or data-residency GPU cloud when local delivery matters. You get a written quote, not a self-serve console. Current capacity we can quote is India-hosted, including H200 class capacity on request. We also take briefs for the UAE and Southeast Asia.
Market reference, 17 November 2026
These are dated official public list prices, not a live feed and not Shorecompute rates. They help a buyer size a budget. A written quote is the only commercial offer.
from $0.33/hr
Lowest official on-demand list in the display below, checked 17 November 2026. Market rates, not our prices. Not real-time.
Each bar is the lowest official public on-demand list price among the sources named in the market-reference table below. Checked 17 November 2026. Not a live feed. No availability counts. Not Shorecompute prices.
Workstation-class training and inference.
from $0.33/hr Market rate, 17 Nov 2026. Not our price.
Image, video and smaller model work.
from $0.34/hr Market rate, 17 Nov 2026. Not our price.
Inference and graphics-heavy jobs.
from $0.79/hr Market rate, 17 Nov 2026. Not our price.
General training and inference.
from $1.19/hr Market rate, 17 Nov 2026. Not our price.
Large-model training and inference.
from $1.99/hr Market rate, 17 Nov 2026. Not our price.
Long-context and larger models. H200 class capacity can be requested; no quantity is promised until a written quote.
from $3.59/hr Market rate, 17 Nov 2026. Not our price.
Frontier-scale training and inference.
from $5.98/hr Market rate, 17 Nov 2026. Not our price.
This is a requirements selector, not a live inventory check. Nothing is reserved until you receive and accept a written quote.
This is an invoice desk, not a marketplace with live stock. Capacity we can quote today is India-hosted. Briefs for the UAE and Southeast Asia are welcome and will be assessed against what can actually be arranged.
01
Region, GPU class, count, start date, term, residency needs, and intended use.
02
We check what provider programs can actually serve, without inventing inventory.
03
Or a referral, if we cannot contract it ourselves yet.
04
Xenofab invoices you only where we are the contracting party, and supports you on that paper.
A quote-desk version of a capacity lifecycle. There is no self-serve launch button here, and no lock-in beyond the term written in an accepted quote.
Compare
Use the dated market rates as a budget range. We then quote a specific configuration against what can actually be arranged.
Launch
After you accept a written quote, the provider delivers the GPUs. You do not get a live cluster console from this site.
Exit
The quote states the term. We will not claim an open-ended lock-in. Renewal, if any, is a new written agreement.
Repeat
More GPUs, a new region, or a longer term starts another brief and quote. Capacity is not standing inventory on our paper.
This is what agent and inference workloads often need, and what we can try to source. These are not Shorecompute-owned features, and they are not promised until a written quote says so.
Stable endpoints for agents and production inference, sourced from a provider that can host them in the agreed region.
Headroom for spikes, quoted as a range or a burst term, not as a live autoscaling pool we operate.
Programmatic access where the supplying provider offers it. Shorecompute does not run its own compute API.
Dedicated or isolated tenancy when the brief and the provider can support it. Isolation terms will be written into the quote.
Shorecompute does not publish its own rate card. We have no owned inventory and no signed reseller agreement yet. Request a quote and we will price a specific configuration against current provider or referred list terms.
Pay for time used, often billed per second or per minute. Common for short jobs and trials. Public lists move often.
A month or longer, often invoiced, often cheaper than on-demand, and frequently "contact sales." The natural fit for an invoice desk once a provider agreement exists.
Lower unit rates if the instance can be reclaimed. Only if your workload can stop or checkpoint.
Some lists sell 8-GPU machines as one unit. Node prices must not be read as per-GPU prices. A quote will state the unit clearly.
| GPU | Published on-demand examples |
|---|---|
| RTX A6000 48GB | Runpod Community Cloud $0.33/hr; Secure Cloud $0.69/hr. This is the citable official source for the "from $0.33/hr" figure on this page. |
| RTX 4090 24GB | Runpod Community Cloud $0.34/hr; Secure Cloud $0.89/hr. Not listed on the Lambda, CoreWeave or Leafcloud pricing pages checked on 17 November 2026. |
| L40S 48GB | Runpod Community Cloud $0.79/hr; Secure Cloud $1.09/hr. CoreWeave: 8-GPU node $18.00/hr on demand ($2.25 inference single GPU), spot node $7.88/hr. Yotta Shakti Cloud: 1x L40S VM ₹302/hr on demand. Not listed on the Lambda instances page checked on 17 November 2026. |
| A100 | Runpod Community Cloud: PCIe $1.19/hr, SXM $1.39/hr; Secure Cloud: PCIe $2.09/hr, SXM $2.19/hr. Lambda: 8x HGX A100 80GB $3.19/GPU-hr; A100 40GB $1.99. Leafcloud: A100 80GB €1.61/hr, Amsterdam, excl. VAT. CoreWeave: 8-GPU A100 node $21.60/hr on demand ($2.70 inference single GPU), spot node $9.65/hr. |
| H100 80GB | Runpod Community Cloud: PCIe $1.99/hr, NVL $2.59/hr, SXM $2.69/hr; Secure Cloud: PCIe $2.89/hr, NVL $3.19/hr, SXM $4.39/hr. Lambda: 8x HGX H100 $4.59/GPU-hr, 1x H100 SXM $4.89, 1x H100 PCIe $3.79. Leafcloud: €3.15/hr in Amsterdam, VM included, excl. VAT. Yotta Shakti Cloud (India): 1x H100 SXM VM ₹596/hr on demand; 8x HGX H100 bare metal ₹428 per GPU-hour. |
| H200 141GB | Runpod Community Cloud $3.59/hr; Secure Cloud $5.29/hr. CoreWeave: 8-GPU HGX H200 node $50.44/hr on demand ($6.31 inference single GPU), spot node $20.93/hr. Not listed on the Lambda instances page checked on 17 November 2026. |
| B200 180GB | Runpod Community Cloud $5.98/hr; Secure Cloud $7.99/hr. Lambda: 8x HGX B200 $7.69/GPU-hr, 1x $7.99. CoreWeave: 8-GPU node $68.80/hr on demand ($8.60 inference single GPU), spot node $34.11/hr. Yotta Shakti Cloud: 8x HGX B200 bare metal ₹473 per GPU-hour. |
Official pages used for the table:
Reserved and committed monthly terms are often unpublished or "contact sales" (Runpod enterprise, Leafcloud discounts for a month or longer). Yotta Shakti Cloud lists a dedicated 1x H100 SXM VM at ₹392,740 per month. The same official Runpod pricing page also lists lower-priced classes than $0.33/hr that are not shown in the chart above (for example RTX A5000 Community Cloud $0.16/hr). The "from $0.33/hr" headline is the lowest official list among the classes displayed. If we later sign a reseller agreement, Shorecompute quotes will be based on that provider's then-current terms, not on this table.
Shorecompute is a trading name of Xenofab LLC, a Wyoming limited liability company. We quote and, where a contract allows, invoice GPU capacity delivered through partner providers. Current capacity we can quote is India-hosted. Where a contract allows, Xenofab bills and supports the customer directly.
Shorecompute is a sourcing desk for GPU capacity in the UAE, India and Southeast Asia. It is not a self-serve GPU cloud, a GPU rental marketplace, or an inference API. Those are different products, offered by other companies. Names in the market-price table on this page are published list-price sources used as reference data only. They are not Shorecompute partners.
Registered office (registered office only, not a data center, warehouse or plant):
Xenofab LLC
Xenofab LLC also offers private AI for family offices at xenofab.pages.dev. That product is separate from this GPU desk.
Send a brief and we will reply from Xenofab LLC. The form opens an email to contact@xenofab.nylas.email with the subject line "GPU enquiry: [firm or name]".