NVIDIA DGX Spark 64GB Launches at $4,999 as Memory Costs Bite

NVIDIA is launching a 64GB version of its DGX Spark AI supercomputer at $4,999, shipping October 23 through six manufacturer partners. The move comes as memory costs have forced the 128GB model’s price to $6,950, nearly double its original $3,999 tag. The compact system retains the Grace Blackwell Superchip and supports up to 100-billion-parameter AI models on device, letting developers run local AI without cloud costs.

Why NVIDIA cut the memory in half

Rising memory and component costs have made the original 128GB DGX Spark unaffordable for many developers. NVIDIA’s 128GB Founders Edition launched at $3,999 last year, jumped to $4,699 in February, and now sits at $6,950. By releasing a 64GB variant at the old $4,999 price point, NVIDIA keeps the platform accessible while maintaining the same Grace Blackwell Superchip, ConnectX-7 networking, and full AI software stack.

Clustering two 64GB units beats buying one 128GB alone

Two 64GB units clustered together deliver 1.7x the performance of a single system in NVIDIA’s own testing, according to the company’s Qwen 3.8 27B benchmark. Two 64GB boxes cost roughly $10,000 for 128GB of pooled memory, versus $6,950 for a single 128GB unit. The clustering uses NVIDIA’s Sync Cluster Assistant software, which requires no additional setup beyond a QSFP cable connection.

Local AI model support and target market

The 64GB DGX Spark runs models up to 100 billion parameters fully on device, with two clustered units supporting 200-billion-parameter models. NVIDIA ships the system with Agent Toolkit, CUDA-X libraries, Nemotron open models, and runtimes like Ollama, vLLM, and PyTorch pre-installed. The system targets developers, researchers, and AI enthusiasts who want to experiment with models and proprietary data without paying cloud computing fees per task.

What this means for NVIDIA investors

NVIDIA’s stock rose 1.3% to $233.95 on October 4, with Meyka grading the company a B+ and forecasting $224.84 by year-end. The DGX Spark price hikes reflect both rising memory costs and strong demand for local AI hardware. Analyst consensus remains bullish at 4.0 out of 5, with 26 buy ratings and 2 strong buys. However, Meyka’s valuation metrics flag caution: the PE ratio sits at 40.06 and the stock trades at 24.7x book value, suggesting limited margin of safety at current levels.

Final Thoughts

NVIDIA’s 64GB DGX Spark at $4,999 addresses cost-conscious developers while the 128GB model’s price surge reflects tight memory supply. With Meyka grading NVDA a B+ and analyst consensus bullish, the stock remains favored, but valuation multiples leave little room for error.

FAQs

Disclaimer:

The content shared by Meyka AI PTY LTD is solely for research and informational purposes.  Meyka is not a financial advisory service, and the information provided should not be considered investment or trading advice.