# Netweb Launches Make in India AI Supercomputers with NVIDIA

URL: https://technosports.co.in/netweb-make-in-india-ai-supercomputers/  
Published: 2026-02-20  
Updated: 2026-02-20  
Author: Hardik Dhamija

Indian computing OEM **Netweb Technologies** (NSE: NETWEB, BSE: 543945) has launched two landmark AI supercomputing systems at the India AI Impact Summit 2026 in New Delhi — both designed and manufactured in India under the Make in India mission, powered by NVIDIA’s latest Grace Blackwell silicon.

![](https://technosports.co.in/wp-content/uploads/2026/02/Netweb-1-1-1024x614.jpg)

## Tyrone Camarero Spark vs GB200: Full Specs Compared

The two products target very different segments — one is a compact desktop for AI developers, the other a rack-scale supercomputer for enterprises running trillion-parameter models.

| Spec | Tyrone Camarero Spark | Tyrone Camarero GB200 |
| --- | --- | --- |
| Platform | NVIDIA DGX Spark / GB10 Grace Blackwell Superchip | NVIDIA GB200 NVL4 |
| AI Performance | 1 Petaflop (FP4) | Up to 10 trillion parameter LLM inference |
| Memory | 128GB unified memory | Multi-node liquid-cooled rack-scale |
| Form Factor | Desktop — 5.9 × 5.9 × 2 inches | Liquid-cooled NVIDIA MGX modular server |
| Power Draw | 240W | Rack with 40 Grace CPUs + 80 Blackwell GPUs |
| Max Inference | 200B parameter models | Trillion-parameter real-time LLM inference |
| Local Fine-tuning | Up to 70B parameter models | Full training + inference |
| Software Stack | NVIDIA Nemotron models, NIM microservices, CUDA-X | Sovereign cloud stack, BlueField-3 DPUs, end-to-end security |
| Availability | Now shipping — select Indian AI organisations | Showcased at India AI Impact Summit 2026 |
| Listed On | NSE: NETWEB / BSE: 543945 | — |

## Why the Spark’s Power Efficiency Is the Real Story

The efficiency leap is striking: the Tyrone Camarero Spark delivers 1 Petaflop (FP4) at just 240W in a desktop the size of a thick book. In 2016, achieving a comparable 170 Teraflops (FP16) with the same 128GB memory required 3,200W of power — over 13× more energy. That’s not just an engineering milestone; it makes petascale AI compute genuinely accessible for individual developers, research labs, and startups that can’t operate a data centre.

![Netweb](https://technosports.co.in/wp-content/uploads/2026/02/Netweb-2-1-1024x576.webp)

The GB200 system targets the opposite end — enterprise and sovereign AI workloads. A single liquid-cooled rack integrates 20 Tyrone GB200 MGX systems with 40 [NVIDIA](https://technosports.co.in/?s=NVIDIA) Grace CPUs, 80 NVIDIA Blackwell GPUs, NVIDIA BlueField-3 DPUs, and high-performance storage with end-to-end security. Netweb’s 15-year collaboration with NVIDIA underpins the architecture.

Both systems ship with preinstalled NVIDIA Nemotron open models and NIM microservices, enabling vision search agents, image generation, AI chatbots, and agentic workflows out of the box — entirely on-premises.

For full product details, visit [netwebindia.com](https://www.netwebindia.com/).

Check out our AI hardware and supercomputing coverage on TechnoSports and our India AI Impact Summit 2026 roundup.

## FAQs

### What is the Netweb Tyrone Camarero Spark?

It’s one of the world’s smallest AI supercomputers — a Make in India desktop delivering 1 Petaflop (FP4) and 128GB unified memory at just 240W, built on the NVIDIA DGX Spark / GB10 Grace Blackwell platform, now shipping in India.

### Can the Tyrone GB200 handle trillion-parameter AI models?

Yes — built on NVIDIA’s GB200 NVL4 platform with 80 Blackwell GPUs per rack, it supports real-time LLM inference and AI training on models scaling up to 10 trillion parameters.