# Graphics card NVIDIA Tesla H200 141GB PCIe HBM3e

**URL:** https://aio.lv/en/product--nvidia-900-21010-0040-000--11279161
**Data updated:** 2026-04-14 14:17 (EEST)
**Category:** Graphics Cards
**Manufacturer:** NVIDIA
**Manufacturer code:** 900-21010-0040-000
**EAN:** 8592978607388
**Warranty:** 2 Years
**Availability:** Available for order
**Price:** 37 151.99 €
**Prices and availability valid until:** 2026-07-22 23:59
**Price without VAT (21%):** 30 704.12 EUR
**Can be purchased by legal entities:** Yes (invoice)
**Buy:** https://aio.lv/en/product--nvidia-900-21010-0040-000--11279161

## Delivery options

| Delivery method | Delivery time | Cost |
|---|---|---|
| Office pickup (2b. Kengaraga str., Riga, Latvia, LV-1063) | 09.07.2026 | free |
| Pickup points | 14.07.2026 | 1.89 EUR |
| Courier delivery to door | 14.07.2026 | 4.99 EUR |

## Summary description

NVIDIA Tesla H200 is a data-center GPU in PCIe form factor with 141GB of high-bandwidth HBM3e memory. Designed for AI training, inference and high-performance computing workloads, it delivers large memory capacity and fast memory bandwidth for handling massive models and datasets. Suited for servers and workstations requiring accelerated compute performance.

## Long description

NVIDIA Tesla H200 141GB PCIe HBM3e is a data‑center GPU in a PCIe add‑in card form factor engineered for AI training, inference and high‑performance computing. With 141 GB of HBM3e high‑bandwidth memory and architecture tuned for large models and datasets, this card provides consolidated memory capacity and sustained bandwidth for demanding server and workstation workloads. The product is intended for deployment in rack servers, appliance nodes and workstation platforms that require accelerated tensor and matrix computation.
Advantages
The Tesla H200 delivers a high memory capacity combined with HBM3e memory bandwidth to reduce off‑chip transfers and improve throughput for large‑scale models. Its PCIe form factor simplifies integration into common server and workstation platforms without requiring specialized interconnects. The design targets sustained performance for mixed‑precision AI training and inference, as well as parallel scientific and data analytics tasks.
Suitable for
Training large neural networks that require very large GPU memory pools
Batch and real‑time inference workloads in data‑center environments
High‑performance computing tasks such as simulation, linear algebra and molecular modeling
Data‑analytics pipelines that process large in‑memory datasets
Server and workstation setups where a PCIe accelerator is preferred over proprietary interconnect cards
When it may not be suitable
This GPU is optimized for data‑center and professional compute workloads; it is not intended as a consumer or desktop gaming card. Systems with strict power, thermal or slot‑space constraints may require evaluation of compatibility before deployment. For small‑scale or single‑GPU desktop tasks, a lower‑capacity or consumer‑oriented GPU may be more cost‑effective.
Key specifications
Form factor: PCIe add‑in card
Memory type: HBM3e high‑bandwidth memory
Memory capacity: 141 GB
Target applications: AI training, inference, high‑performance computing
Designed cooling: passive or server‑chassis cooling (requires adequate airflow)
How to use
Install the card into a compatible PCIe slot in a server or workstation that meets the required power delivery and cooling specifications. Ensure the host system BIOS and OS support the GPU and that appropriate drivers and software stacks (CUDA, cuDNN, frameworks) are installed and configured. For multi‑GPU setups, configure interconnect and software to manage memory and workload distribution. Monitor temperatures and power draw during initial testing to confirm stable operation under expected workloads.
What to consider before purchase
System compatibility: available PCIe slots, chassis space and airflow
Power budget: PSU capacity and power connectors to support sustained load
Driver and software support for intended frameworks and libraries
Cooling strategy: chassis or rack cooling sufficient for passive or dense deployments
Deployment scale: single‑card versus multi‑card topology and host infrastructure
Usage recommendations
For best results, pair this GPU with server hardware designed for high airflow and reliable power delivery, install the latest vendor drivers and AI framework updates, and validate workloads using representative datasets to size memory and compute needs accurately.

## FAQ

### How much does the NVIDIA Tesla H200 141GB PCIe graphics card cost, and is it available for purchase?

The price is 37,151.99 € (valid at the time of publication and already includes all taxes). The item is available for order — 4 pcs. are in stock, and you can place your purchase right now with delivery across Latvia and Europe or choose pickup at the Riga collection point. We offer favorable purchase terms, including the option to split the payment into convenient monthly installments. Availability and price are updated in real time, so we recommend placing your order in advance.

### If I order this product today, when will I receive my order?

If you place your order today, delivery will be as fast and convenient as possible. You will be able to pick up your order on 09.07.2026 at the Riga pickup point or choose delivery across Latvia and Europe. We carefully inspect every purchase and pack it securely for safe transportation.

### What payment options are available?

AiO.lv offers several payment methods: Bank transfer, Bank cards on-line, Apple Pay, Google Pay, Leasing for legal entities. All transactions are 100% secure and encrypted.

### Can this product be purchased for a company (with VAT)?

Yes. AiO.lv works with legal entities in Latvia and across Europe. The product price already includes VAT. For certain product categories, as well as for companies from other EU countries with a valid VAT number, a 0% VAT rate may apply. We issue the invoice immediately after the order is placed, support electronic BIS invoices, and automatically generate all necessary accounting documents. You can complete the purchase for a company online — simply select the company profile and choose bank transfer as the payment method during checkout.

### What is the warranty period?

All NVIDIA products come with an official 2 Years warranty. It covers all manufacturing defects and malfunctions that may occur during use. The warranty is valid at all authorized NVIDIA service centers and through AiO.lv

### Can the product be returned after purchase?

Yes, you can return the product within 14 days of purchase if you bought it as a consumer (not a legal entity). Returns are possible without giving any reason — the main condition is that the product retains its original appearance and full set of accessories. We will make the process as simple and convenient as possible: just contact our customer support team, and we will help you arrange the return.

### Where should I contact for warranty service?

For warranty-related questions you can contact the AiO.lv (SIA VSPRO) customer support team. More information about warranty service: https://aio.lv/en/warranty

## Image

https://aio.lv/img/cache/product/11279161/68655245_large.webp