NVIDIA GB10 SuperChip delayed into Q4

No sign of the DGX Spark.

Once again, we return to the delay of the GB10 SuperChip, perhaps NVIDIA’s most delayed product yet (aside from the NVIDIA Spot). This chip is meant to power a compact workstation known as DGX Spark, previously announced as Project Digits. It’s a small system running the NVIDIA DGX operating system, featuring one processor and one ConnectX chip inside a compact chassis.

This platform is likely the closest competitor to AMD’s Strix Halo, a single processor combining CPU cores and a powerful GPU, designed primarily for AI workloads. Each chip comes with 128 GB of LPDDR5X memory, clearly aimed at AI training tasks.

We’re revisiting this topic because DGX Spark was initially expected to launch in May (according to ComputerBase), and later in July. Yet after an entire summer, there’s been no sign of reviews or users showcasing the system. As we enter the fourth quarter, NVIDIA’s official page still displays the same “notify me” message:

Source: NVIDIA

Meanwhile, third-party vendors such as ASUS and HP are confirming that their systems are set to launch sometime this quarter, though none have shared specific release dates. ASUS even contacted us after our previous post to clarify that, despite what their webinar slides implied, there is still no confirmed schedule. It appears no one outside NVIDIA really knows when Spark will ship.

Scan UK, who first listed September 15th, then 30th, now doesn’t ‘even mention any date:

Source: Scan UK

NVIDIA was also rumored to introduce N1-series SoCs for consumer laptops and desktops, but no such announcement came at either CES or Computex. The company has never officially confirmed those plans.

Recently, NVIDIA’s CEO confirmed that the GB10 SuperChip carries the internal codename “N1,” supporting earlier rumors that the two share more than just similar specifications, they are essentially the same chip. Perhaps “N1” was always meant to remain internal, and NVIDIA never planned to release a consumer version at all.

Source: NVIDIA , ComputerBase