01 · Inference
Reasoning and agentic AI
Serve large reasoning models with long context windows and many concurrent users, at a lower cost per token.
NVIDIA HGX B300 supplier · Kuala Lumpur, Malaysia
Looking for an NVIDIA HGX B300 supplier in Malaysia? Hometec Marketing sources, configures and delivers NVIDIA HGX B300 servers for enterprises, research labs and AI factories. Tell us your workload and we will help you spec the right system.
Built for
Blackwell Ultra adds more FP4 compute and faster attention than the previous generation, which matters most where models think before they answer.
01 · Inference
Serve large reasoning models with long context windows and many concurrent users, at a lower cost per token.
02 · Training
Train and fine-tune large language models on a single node, then scale out to multi-rack clusters over 800 Gb/s networking.
03 · Compute
Accelerate simulation, scientific computing and data pipelines on the same platform your AI teams use.
About us
Your partner for enterprise AI infrastructure, from first enquiry to a running cluster.
Hometec Marketing supplies data center hardware to organisations that need dependable, high-performance computing. We focus on NVIDIA HGX systems because they set the standard for AI training and inference.
We work with leading server manufacturers to offer HGX B300 in air-cooled and liquid-cooled configurations, matched to your power budget, rack space and workload. Our team handles sourcing, configuration, logistics and after-sales support so your engineers can focus on models, not procurement.
Whether you need a single node for a research team or a multi-rack deployment, we give you a clear quotation, realistic lead times and one point of contact throughout.
We review your workloads, model sizes and facility limits.
We recommend CPU, memory, storage, networking and cooling options.
We manage sourcing, shipping and installation coordination.
We stay with you for warranty, spares and expansion.
Product
Eight NVIDIA Blackwell Ultra GPUs joined by fifth-generation NVLink and NVSwitch, with ConnectX-8 SuperNICs for scale-out. Compared with HGX B200 it delivers 1.5× more dense FP4 compute and 2× the attention performance.
| Form factor | 8× NVIDIA Blackwell Ultra SXM |
|---|---|
| FP4 Tensor Core | 144 PFLOPS sparse | 108 PFLOPS dense |
| FP8 / FP6 Tensor Core | 72 PFLOPS |
| GPU memory | 2.3 TB HBM3e (288 GB per GPU) |
| GPU interconnect | 5th-gen NVLink + NVSwitch, 1.8 TB/s per GPU |
| Networking | NVIDIA ConnectX-8 SuperNICs, up to 800 Gb/s |
| Host CPU | Dual Intel Xeon or AMD EPYC (by OEM) |
| System memory | Up to 8 TB DDR5 (by OEM) |
| Software | NVIDIA AI Enterprise, CUDA, Dynamo |
Configuration A
Drops into existing air-cooled data halls. The quickest route to deployment.
Configuration B
Twice the density per rack with direct-to-chip cooling, for facilities with liquid cooling in place.
Scale-out
Combine nodes into scalable units with NVIDIA Quantum-X800 InfiniBand or Spectrum-X Ethernet.
Performance figures are NVIDIA published results and projections and may change. CPU, system memory, storage and chassis details depend on the server manufacturer. Ask us for the exact specification of the system we quote.
FAQ
Answers to the questions buyers ask most about who supplies the HGX B300, where to buy it, and how and when it is delivered.
Hometec Marketing in Kuala Lumpur supplies NVIDIA HGX B300 servers to enterprises, research institutions and data centers in Malaysia. Email hometecmarketing@gmail.com or call +60 14-278 2768.
From Hometec Marketing, Lot A-0-01, Phoenix Business Park, Jalan 142/2, Megan Phoenix, KM10, 56000 Kuala Lumpur. Request a quotation through the contact form, by email or by phone.
An 8-GPU server platform built on NVIDIA Blackwell Ultra, with 2.3 TB of HBM3e memory, up to 144 PFLOPS of FP4 Tensor Core performance, 5th-generation NVLink at 1.8 TB/s per GPU and ConnectX-8 SuperNICs up to 800 Gb/s. It is designed for AI reasoning, model training and HPC. See the full specifications.
Send us your workload, number of nodes and preferred cooling (air-cooled 8U or liquid-cooled 4U). We recommend a configuration, send a quotation with lead time, then manage sourcing, delivery and after-sales support.
NVIDIA HGX B300 systems are shipping now. Delivery time depends on the configuration, quantity and current stock. We confirm the lead time in your quotation.
The price depends on the server manufacturer, CPU, memory, storage, networking and cooling you choose. We provide a detailed quotation for your exact configuration.
Contact us
Tell us about your project and our team will reply with pricing, lead time and a recommended configuration.
Hometec Marketing
Lot A-0-01, Phoenix Business Park,
Jalan 142/2, Megan Phoenix, KM10,
56000 Kuala Lumpur, Malaysia
Monday to Friday, 9:00 to 18:00