One 748 GB memory pool
The Grace CPU's 496 GB of LPDDR5X and the GPU's 252 GB of HBM3e are linked as a single coherent pool, so large models and datasets are not squeezed into a graphics card's own memory.
MSI · NVIDIA DGX Station architecture
XpertStation WS300
A GB300 Grace Blackwell Ultra AI system in a tower.
One 748 GB memory pool and two 400 GbE ports, at your desk. Develop, fine-tune and run large AI models on your own data, in your own office, with a GST invoice and insured delivery across India from TechCure India.
- 748 GBcoherent memory
- 2 × 400 GbEQSFP112, ConnectX-8
- 72-coreNVIDIA Grace CPU
Price
₹1,55,99,979
In stock
What sets it apart
Data-center ideas, in one tower
A superchip, not a parts list
The NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip pairs a 72-core Grace CPU with a Blackwell Ultra GPU in one design built for AI.
Data-center networking
Two QSFP112 400G Ethernet ports from the integrated ConnectX-8 SuperNIC, plus a 10GBase-T port, for moving big datasets and linking systems.
Room to grow
Two M.2 PCIe 6.0 x4 slots are open for expansion, and three PCIe 5.0 x16 slots are available: one double-wide and two single-wide at x8 signalling.
What you can do with it
Built for the whole AI workflow
AI development
Train and fine-tune models on data that never leaves your premises.
Data science
Run GPU-accelerated analytics and pipelines on very large datasets.
AI inference
Serve large models locally for chat, analysis, search and content tools.
AI agents and physical AI
Develop and test autonomous agents and robotics-style workloads close to the team.
TechCure note Not sure it fits? Tell us the model size, team size and budget on WhatsApp and we will say honestly whether the WS300, a DGX Spark or a custom workstation is the right tool.
Specifications Official, from MSI
Everything in the chassis
| Platform | NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, NVIDIA DGX Station architecture |
|---|---|
| CPU | 1 × NVIDIA Grace, 72-core Arm Neoverse V2 |
| GPU | 1 × NVIDIA Blackwell Ultra GPU |
| Memory | Up to 748 GB coherent memory CPU memory: up to 496 GB LPDDR5X, 396 GB/s GPU memory: up to 252 GB HBM3e, 7.1 TB/s |
| Internal storage | 2 × M.2 2280 PCIe 5.0 x4 NVMe, populated with boot drives, connected to the CPU 2 × M.2 2280 PCIe 6.0 x4 NVMe, open for expansion, connected to the ConnectX-8 |
| Expansion | 1 × PCIe 5.0 x16 full-height, full-length, double-wide slot 2 × PCIe 5.0 x16 full-height, full-length, single-wide slots with x8 signals 1 × PCIe 2.0 x1 M.2 2230 for Wi-Fi / Bluetooth |
| Networking | 2 × QSFP112 400G Ethernet (NVIDIA ConnectX-8) 1 × 10GBase-T Ethernet (Marvell AQC113) |
| Management | 1 × 1000Base-T dedicated management port; ASPEED AST2600 with AMI MegaRAC firmware, IPMI 2.0 and DMTF Redfish API; eMMC for local BMC storage |
| Security | Chassis intrusion detection; on-board TPM 2.0 |
| Dimensions (W × H × D) | 245.7 × 529.0 × 570.4 mm (9.67 × 20.83 × 22.46 in) |
| Software | Ubuntu 24.04 LTS with NVIDIA AI developer tools, as listed on MSI's product page |
Source: MSI XpertStation WS300 datasheet, checked 10 Oct 2026. MSI notes that specifications are subject to change. Drives and accessories supplied with your unit can differ, so confirm the exact configuration with us before ordering.
Choosing TechCure guide
WS300 or DGX Spark?
MSI XpertStation WS300
For labs and teams that outgrow a desktop box
- GB300 Grace Blackwell Ultra
- Up to 748 GB coherent memory
- 2 × 400 GbE QSFP112
- Full tower with PCIe expansion and open M.2 slots
NVIDIA DGX Spark systems
For one developer or a small team
- GB10 Grace Blackwell
- 128 GB unified memory
- Up to 1 petaFLOP (FP4, with sparsity), from NVIDIA
- Small desktop box. See DGX Spark systems
Need Windows, gaming, video editing, CAD or trading instead? That is a job for a custom PC, not an AI workstation.
Why buy from TechCure India
Sold, delivered and supported by us
GST invoice
Full tax invoice with input credit for companies, colleges and research labs.
Insured delivery
Shipped insured to every serviceable pin code in India.
Setup guidance
We help you get to your first run: remotely anywhere in India, and on-site in Surat and Gujarat.
One point of contact
Warranty questions and service go through us. Warranty terms are on the product page; ask us before you order.
Questions
XpertStation WS300, answered
What is the MSI XpertStation WS300?
The XpertStation WS300 is MSI's tower workstation built on NVIDIA DGX Station architecture. It is powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip and is meant for AI development, data science, AI inference and AI agents.
How much memory does it have?
Up to 748 GB of coherent memory: 496 GB of LPDDR5X attached to the Grace CPU (396 GB/s) and 252 GB of HBM3e on the Blackwell Ultra GPU (7.1 TB/s), shared as one pool.
What networking does it have?
Two QSFP112 400G Ethernet ports from the integrated NVIDIA ConnectX-8 SuperNIC, one 10GBase-T port, and a separate 1000Base-T port for server management.
Can I add more storage?
Yes. According to MSI, two M.2 2280 PCIe 6.0 x4 slots are open for expansion, while the two PCIe 5.0 x4 slots hold the boot drives. Ask us which drives come with the unit you are buying.
How is it different from DGX Spark?
DGX Spark is a small, single-developer system with 128 GB of unified memory. The WS300 is a full tower on the larger GB300 platform with up to 748 GB of coherent memory and two 400 GbE ports, for labs and teams that outgrow Spark.
Can it run Windows or games?
It is built for AI and data-science workloads, not as a gaming or general-purpose Windows PC. MSI lists Ubuntu 24.04 LTS with NVIDIA AI developer tools. For gaming, editing or trading, a custom PC is the better choice.
Do I get a GST invoice, and do you deliver across India?
Yes. We issue a GST tax invoice with input credit for businesses, and ship insured to every serviceable pin code in India. The current price is shown at the top of this page and on the product page.
Tell us what you want to build.
Share your model size, team size and budget. We reply with the right system and a GST quotation.
XpertStation WS300 on NVIDIA DGX Station architecture. TechCure India is an independent reseller and is not affiliated with MSI or NVIDIA. MSI, NVIDIA, Grace Blackwell, NVLink and ConnectX are trademarks of their respective owners. Product photos are from the store listing; specifications are MSI's. Prices and stock are live from our store.