Personal AI computers

The NVIDIA DGX Spark is a compact, desktop-class "personal AI supercomputer" built specifically for developers, data scientists, and researchers to prototype, fine-tune, and run advanced AI models locally.

Rather than relying constantly on cloud APIs or enterprise data centers, it brings data-center-grade Grace Blackwell performance right to your workstation in a quiet, low-power form factor.

NVIDIA DGX Spark desktop AI supercomputer, AI generated

NVIDIA DGX Spark desktop AI supercomputer. Source: Bloomberg / Bloomberg via Getty Images


Technical Specifications

Component

Specification

Silicon

NVIDIA GB10 Grace Blackwell Superchip

CPU

20-core Arm (10x Cortex-X925 + 10x Cortex-A725)

GPU Architecture

Blackwell (5th Gen Tensor Cores)

Compute Performance

Up to 1 PFLOP (FP4 Tensor performance)

System Memory

128 GB Unified LPDDR5X

Storage

4 TB NVMe SSD

Interconnect

NVIDIA ConnectX-7 (QSFP ports for local clustering)

OS Stack

NVIDIA DGX OS (Ubuntu Linux) + NVIDIA AI Enterprise


Core Capabilities & Strengths

  1. Large Model Local Inference: The 128 GB unified memory pool allows you to run local inference on state-of-the-art LLMs and multimodal models up to 200B parameters.
  2. Local Fine-Tuning: Efficiently fine-tune models up to 70B parameters on private datasets without sending sensitive IP or proprietary code to cloud providers.
  3. Always-On Autonomous Agents: Pre-integrated with tools like NemoClaw, vLLM, and TensorRT-LLM to run agentic AI frameworks and local developer assistants (e.g., Ollama, OpenClaw) securely in an isolated sandbox.
  4. Seamless Clustering: Multiple Spark units can be daisy-chained or networked directly via 200GbE ConnectX-7 links to pool memory and compute for heavier workflows.


While DGX Spark is NVIDIA's Linux-based desktop supercomputer for AI developers and researchers, NVIDIA RTX Spark is the consumer and creator variant co-developed with MediaTek and Microsoft.

Announced for premium Windows on Arm laptops and compact PCs, it integrates Grace Blackwell power into an energy-efficient form factor designed for daily productivity, local creative workflows, and high-end AI acceleration.

NVIDIA RTX Spark system-on-chip, AI generated

NVIDIA RTX Spark system-on-chip. Source: Bloomberg / Bloomberg via Getty Images


RTX Spark vs. DGX Spark

Feature

RTX Spark

DGX Spark

Form Factor

Windows laptops & thin mini-PCs

Desktop AI workstation / Dev Box

Operating System

Windows 11 (Windows on Arm)

NVIDIA DGX OS (Ubuntu Linux)

Power Envelope

18W – 80W (battery/laptop friendly)

Desktop power profile

Primary Use Cases

Creative suite apps, Copilot+ features, local AI, Arm gaming

Large LLM fine-tuning (up to 70B), local inference (up to 200B)


Technical Specifications

  1. Custom CPU Architecture: Features up to a 20-core Arm CPU (10x Cortex-X925 performance cores + 10x Cortex-A725 efficiency cores) co-designed with MediaTek.
  2. Blackwell RTX Graphics: Houses up to 6,144 CUDA cores with 5th-generation Tensor Cores delivering up to 1 PFLOP of FP4 AI compute performance.
  3. Unified Memory Architecture: Up to 128 GB of LPDDR5X unified RAM connected over NVLink-C2C, allowing the CPU and GPU to instantly share large assets without copying data across bus boundaries.
  4. Windows 11 Optimization: Built for Microsoft Copilot+ PC capabilities, complete with Prism x86 emulation and native anti-cheat support (Easy Anti-Cheat / BattlEye) for Windows gaming on Arm.


Which laptop manufacturers are releasing NVIDIA RTX Spark laptops and when will they be available?


NVIDIA and its hardware partners confirmed that the first RTX Spark laptops and mini-PCs will officially begin shipping in Fall 2026.

The rollout is divided across major OEM partners leading into the holiday season:

OEM Manufacturer Rollout

Launch Wave

Manufacturers

Expected Timing

Key Focus & Product Lines

Wave 1 (Lead Partners)

ASUS, MSI

Fall 2026 (Sept–Oct)

High-performance creator workstations (e.g., ASUS ProArt series) and thin-and-light gaming laptops.

Wave 2

Acer, GIGABYTE

Late Fall 2026 (Oct–Nov)

Portable developer laptops and high-end gaming/creator models.

Wave 3

HP, Microsoft

Late 2026 / Early 2027

Premium business workstations (e.g., HP Z-series) and flagship Windows on Arm hardware (e.g., Surface Laptop Ultra).


Key Launch Details & Pricing

  1. Pricing Expectations: Given the Grace Blackwell silicon, 1 PFLOP FP4 AI capability, and up to 128 GB unified memory configuration, initial retail prices for RTX Spark laptops are expected to start around $2,500 to $3,000+.
  2. Form Factors: Beyond standard 14-inch and 16-inch creator laptops, several OEMs (including HP and MSI) are also preparing ultra-compact Mini-PCs and desktop AI stations built on the same RTX Spark platform.
  3. Software Ecosystem: All shipping models will join Microsoft’s Copilot+ PC ecosystem and ship with optimized Windows on Arm scheduling, native Prism x86 translation, and full NVIDIA CUDA/TensorRT software support out of the box.



Source share.gemini.google/eu0yhTEwE9nF