NVIDIA GTC 2025: AI Supercomputers Take Over—Where Are the Graphics?

NVIDIA just wrapped up GTC 2025, and if you were expecting groundbreaking graphics news, well… let’s just say the “Graphics” in Graphics Technology Conference is becoming more of a legacy name. Instead, this year’s focus was crystal clear: AI supercomputing. Whether you see this as an exciting leap forward or the first step toward a robot uprising is up to you—but one thing is certain: NVIDIA is going all-in on AI hardware.

One announcement that truly stood out was NVIDIA’s new lineup of AI personal supercomputers:

• DGX Spark – A compact AI workstation
• DGX Station – A beastly AI desktop supercomputer

These machines are built for running massive AI models and handling gigantic datasets. Let’s dive into what they bring to the table.

DGX Spark – A Mini AI Powerhouse

Think of the DGX Spark as the Mac Mini of AI development. Yes, NVIDIA is now making full-fledged computers—not just GPUs. And much like Apple, this system runs on an ARM-based CPU paired with their Blackwell GPU, forming a powerful GB100 Grace Blackwell SoC (System on Chip).

Key Specs

• 128GB Unified Memory
• 1,000+ TFLOPS of FP4 compute
• Supports AI models up to ~200B parameters (with accuracy trade-offs)
• Price: $4,000 (~?4 lakhs in India)

At first glance, this is a game-changer—a powerful AI workstation at a relatively affordable price. However, there are a few important limitations:

• No Windows support (yet). It runs a custom Linux-based DGX OS—which is great for developers but means you miss out on Windows software compatibility.
• ?Blazing-fast networking – With an 800GB/s ConnectX SuperNIC, NVIDIA claims you can stack multiple DGX Sparks together or even connect one to your laptop for extra AI horsepower.

Cheaper alternatives? Some OEMs like ASUS plan to offer $1,000 cheaper versions—but whether they’ll be available in India remains uncertain.

DGX Station – The AI Supercomputer for Your Desk

If the DGX Spark is an entry-level AI workstation, the DGX Station is an absolute monster. And before we even get into specs, just take a moment to acknowledge this:

784GB of unified memory. Yes, 784GB. Most PCs don’t even have that much storage, let alone RAM.

Key Specs

• GPU Memory: 288GB HBM3e (8TB/s bandwidth)
• CPU: 72-core Grace Neoverse V2 (paired with 496GB LPDDR5 RAM)
• Multi-instance GPU (MIG): Split the GPU into 7 isolated instances—basically turning your PC into a mini data center.
• PCIe Expansion: 3 PCIe slots for extra GPUs (because why not?).
• Price: $35,000 (~?3.5 crores in India, according to leaks).

This isn’t just a PC—it’s a full-scale AI research machine. NVIDIA is targeting AI R&D teams, universities, and enterprise AI workloads.

But let’s be real—it’s also perfect for YouTubers who want to flex with a “I Bought a ?3.5 Crore PC!!” video.

Final Thoughts – Should You Get One?

If you’re a serious AI developer, the DGX Spark could be a compelling option (provided you’re okay with Linux). For everyone else? The DGX Station is more of a scientific research tool than something you’d casually buy—unless, of course, you just happen to have ?3.5 crores lying around.

So, if you’re interested in these AI powerhouses, hit us up via our website or at our stores across India.

Stay tuned for more updates, and don’t forget to like and subscribe!