NVIDIA DGX Spark 64GB Launches at $4,999 Amid Memory Shortage

Key points

• The 64GB DGX Spark will be available from Acer, ASUS, Dell, Gigabyte, HP, and MSI starting October 23, 2026.
• The device costs $4,999, which is $1,000 more than the 128GB model's initial launch price of $3,999.
• It features the NVIDIA GB10 Grace Blackwell Superchip, 64GB of LPDDR5x unified memory, and 273 GB/s memory bandwidth.
• The 64GB model supports up to 100-billion-parameter AI models, compared to 200 billion for the 128GB version.
• Two 64GB units can be clustered via NVIDIA Sync Cluster Assistant to pool 128GB of memory and achieve 1.7x the throughput of a single 128GB unit.
• The system includes the full NVIDIA AI software stack, including PyTorch, Ollama, and vLLM, and supports agentic AI workloads.

Background

The launch of the 64GB DGX Spark occurs amid a severe global memory shortage, with Micron warning that supply and demand may not balance until 2028. This follows the 128GB DGX Spark's launch in late 2025, which saw price increases from $3,999 to $4,699 due to supply constraints. ASUS had previously announced its Ascent GX10 64GB configuration on October 5, 2026, as part of its 'Ubiquitous AI' strategy to make local AI development more accessible. The 128GB model is currently out of stock in many markets, with prices rising to $6,950 or more.

How outlets are covering it

NVIDIA frames the 64GB DGX Spark as an 'accessible entry point' for developers and researchers, emphasizing its ability to run capable open-source models locally without cloud dependency. ASUS highlights the device's role in its 'Ubiquitous AI' strategy, targeting the 'early-majority' audience of developers and AI-curious teams. PCMag criticizes the pricing, noting that the 64GB model costs $1,000 more than the 128GB model's original launch price, despite having half the memory. Tom's Hardware frames the launch as a 'lifeline' for local AI fans amid the 'RAMpocalypse,' while NVIDIA's blog emphasizes the seamless scaling capabilities via clustering. All sources agree on the technical specifications and release date, but differ in their assessment of the value proposition given the memory shortage.

Why it matters

The 64GB DGX Spark addresses the growing demand for local AI infrastructure as agentic AI workloads increase token consumption and cloud inference costs. By offering a lower-memory configuration at a higher price, NVIDIA is attempting to balance supply constraints with market demand, while also providing a clear upgrade path for users who need more memory. This move reflects the broader trend of AI companies and developers seeking to keep data and models on-device for privacy and cost reasons, even as memory prices continue to climb.

What to watch

The 64GB DGX Spark will be available for purchase on October 23, 2026, from major PC manufacturers. NVIDIA is also launching the Sync Model Launcher at the end of October, which will simplify the process of running local AI models on single or clustered DGX Spark systems. The ongoing memory shortage may lead to further price increases for both the 64GB and 128GB models, and Micron expects the imbalance to persist through 2028. Users may need to consider alternative platforms, such as AMD's Ryzen AI Halo, which starts at $3,999, or wait for supply to stabilize.