NVIDIA unveiled RTX Spark at Computex 2026, an Arm-based superchip that packs a 20-core CPU, an RTX 5070-class GPU, and up to 128GB of unified memory into a single package. The chip delivers 1 petaflop of AI performance and can run large language models with up to 120 billion parameters entirely on-device.
The RTX Spark platform targets Windows on Arm laptops and desktops, directly challenging Qualcomm’s Snapdragon X Elite and Apple’s M-series chips in the AI PC space. NVIDIA priced the entry-level RTX Spark systems starting at ,799, with premium configurations reaching ,499.
What Makes RTX Spark Different
The unified memory architecture is the headline feature. Unlike traditional x86 systems where the CPU and GPU each have separate memory pools, RTX Spark uses a shared 128GB pool accessible by both the 20-core Arm CPU and the integrated RTX GPU. This eliminates the data copying bottleneck that plagues current AI workloads on laptops.
NVIDIA demonstrated RTX Spark running a 120-billion-parameter language model locally, something that typically requires a dedicated server with multiple GPUs. The company positioned this as a privacy advantage: your data never leaves the device.
RTX Spark Tiers and Pricing
NVIDIA announced three tiers for RTX Spark systems:
- RTX Spark 5070: 20-core Arm CPU, RTX 5070 GPU, 64GB unified RAM, starts at ,799
- RTX Spark 5070 Ti: 20-core Arm CPU, RTX 5070 Ti GPU, 96GB unified RAM, starts at ,999
- RTX Spark 5080: 20-core Arm CPU, RTX 5080 GPU, 128GB unified RAM, starts at ,499
ASUS, Lenovo, and HP announced the first RTX Spark laptops. The ASUS ProArt P16 leads the lineup, followed by Lenovo’s ThinkPad X1 Carbon with RTX Spark and HP’s EliteBook X G2a. All three are expected to ship in Q3 2026.
Impact on Intel and AMD
NVIDIA’s entry into the PC chip market rattled Intel and AMD stock prices. Intel dropped 4.2% and AMD fell 3.8% following the announcement. Both companies have been slow to integrate high-performance NPUs into their mobile processors, and NVIDIA’s unified memory approach gives it a technical edge for AI workloads.
Qualcomm, which currently dominates the Windows on Arm space with Snapdragon X Elite, faces the most direct competition. NVIDIA’s chip offers significantly more GPU performance and double the memory capacity of Qualcomm’s top-tier offering.
Software Compatibility
The biggest question surrounding any Arm-based PC chip is software compatibility. NVIDIA partnered with Microsoft to optimize Windows on Arm for RTX Spark, and the company claims that over 90% of the top 100 Windows applications now run natively on Arm. For apps that still rely on x86 emulation, NVIDIA built in hardware-accelerated translation layer that the company says delivers 85% of native performance.
For developers, NVIDIA released updated CUDA toolkit support for RTX Spark, meaning existing CUDA-based AI and ML code can be compiled for the Arm platform without major rewrites. PyTorch and TensorFlow both received Arm-native builds optimized for RTX Spark’s unified memory architecture.
FAQ
Can NVIDIA RTX Spark run ChatGPT locally?
RTX Spark can run open-source language models with up to 120 billion parameters locally on the device. Closed models like ChatGPT require an internet connection since they run on OpenAI’s servers, but you can run comparable open models like Llama 3.1 120B or Mistral Large entirely offline.
How does RTX Spark compare to Apple M4 Ultra?
Apple M4 Ultra offers up to 192GB of unified memory, which is more than RTX Spark’s maximum of 128GB. However, RTX Spark’s GPU delivers significantly higher raw compute throughput for AI inference tasks. The M4 Ultra has better power efficiency, while RTX Spark trades battery life for raw performance.
Will RTX Spark laptops have good battery life?
NVIDIA claims up to 18 hours of battery life for RTX Spark laptops in standard use. Running AI workloads locally will drain the battery faster, but NVIDIA’s power management system dynamically allocates power between the CPU and GPU based on workload.
Can I upgrade RTX Spark RAM later?
No. The unified memory in RTX Spark is soldered directly to the chip package and cannot be upgraded after purchase. You need to choose your memory configuration at the time of purchase.
When will RTX Spark laptops be available?
The first RTX Spark laptops from ASUS, Lenovo, and HP are expected to ship in Q3 2026, with pre-orders opening in July 2026.
