You’ll soon hear a lot about AI-powered “supercomputers” that don’t work for you

No, your new AI PC or Copilot+ PC is not good at running AI — at least, not with any significant amount of computing. True AI processing takes place behind the closed doors of the sprawling data centers currently popping up all over the United States. Nvidia, which has become a trillion-dollar company thanks to artificial intelligence, is now asking you to put an ounce of that cloud on your desk.

Nvidia It first announced the $4,000 DGX Spark An artificial intelligence-powered calculating machine, later called Project Numbers, during… Consumer Electronics Show 2025. If you don’t remember the details, I don’t blame you. While CEO Jensen Huang has been talking up the AI ​​and graphics capabilities of the company’s RTX 50-series GPUs, the Spark has been pushed further back on Earth as a home machine specifically designed for high-end AI workloads. The company went so far as to describe it as “a new class of computers.” Advertisement post —Despite the fact that it’s powered by a Blackwell chip, an architecture that appears across many of Nvidia’s other lines.

The DGX Spark is scheduled to begin shipping on Wednesday. Regular Nvidia partners like Acer, Asus, Lenovo, MSI, Dell, and Gigabyte are already gearing up to roll out their own versions of the device. You may not find these products when you wander the sparse halls of your nearest Best Buy store, but Nvidia said that it will ship them to Micro Center stores in the United States. Nvidia Make a big deal Sparks is distributed to major companies such as OpenAI and Microsoft, as well as Elon Musk in… Starbase is headquartered in Texas. Maybe the billionaire founder of xAI will plug it in and use it vibe code Centered around artificial intelligence Wikipedia competitor.– which he promised would help people “understand the universe.”

You won’t use the “AI supercomputer” for anything but AI

I’ve seen versions of Spark from both Nvidia and Acer IFA 2025. They both looked like shiny little PCs, but they didn’t actually run Windows. Spark runs a custom Linux distribution built on Ubuntu loaded with many of Nvidia’s AI tools for AI image models and LLMs, or large language models. The ARM-based 20-core CPU comes accompanied by a Grace Blackwell GPU. If all that matters is core count, the Spark will tie in with Nvidia’s GeForce RTX 5070, one of the lower-end GPUs.

On the face of it, these specifications don’t look like a supercomputer. But inside, you’ll find much greater performance and power consumption than a typical desktop computer. The 2.65-pound Spark box has 128GB of system memory and 4TB of storage. Its Blackwell chip promises AI computing performance of up to 1 petaflop, which is several times more than the 170 teraflops offered by Nvidia’s 2016 DGX-1 AI computing machine, though it’s worth noting that throughput, a measure of how fast a GPU can perform a given number of floating-point operations per second, is a rough measure. The Spark also runs at 240 watts, compared to the older model’s power draw of 3,200 watts.

For another loose example of AI compute power, DGX Spark promises to perform around 1,000 TOPS, or trillions of operations per second. Although that falls short of the RTX 5090, which has 3,352 TOPS, it outperforms any equivalent PC of the same size, and its memory puts it on the edge for developing and designing the next chatbot. For comparison, Qualcomm’s upcoming Snapdragon X2 Elite Extreme chip It’s supposed to be much better for AI than its predecessor, thanks to a redesigned NPU, but that can only claim 70 TOPS of AI performance. Your typical PC is still limited to running very low-end AI models or background AI tasks.

Oh, and it all comes with a retail price of around $4,000. Don’t worry: Nvidia doesn’t expect every Joe Schmoe to buy one of these for Running Windows 11 recall. The DGX Spark is designed for budding AI developers, students, or perhaps just curious AI enthusiasts who can afford to drop the equivalent of two $2,000 RTX 5090 GPUs To purchase a specialized computer. Its real mission is to get more developers to create AI apps that people actually want to use, or as Hwang puts it, “the next wave of discoveries.” Hopefully this will take the form of something beyond the promising chatbot interface Home decor repair or Treating the abstract concept of unity.