What NVIDIA told Tokyo about RTX Spark, "from tool to partner" — reading the briefing through Japanese press coverage

In a room lined with RTX Spark machines, NVIDIA's Japan head stepped up to call RTX Spark "AI, a supercomputer, and gaming all at once," as Japan's PC Watch reported (in Japanese) ¹.

NVIDIA held the domestic launch event in Tokyo on Nov 2, according to PC Watch, and declared a full-scale push into the Japanese market together with its PC partners ². Japan's AI Watch added that the venue showed laptops that went on sale on Oct 16, along with compact desktops due in November ³.

AI Watch's headline for its briefing story reads, in translation, "NVIDIA explains the appeal of RTX Spark to Japan: with 1 petaflop of local AI, the PC goes from tool to partner" ⁴.

From tool to partner

According to AI Watch (in Japanese), NVIDIA's Japan head stressed that the PC is at a turning point: from a simple work tool to a "partner that runs AI agents," one that picks up on people's intentions and carries out tasks on its own ⁵. That is the picture the event put front and center, alongside the performance numbers.

Microsoft Japan's Surface lead also took the stage, and PC Watch says the talk looked back at the history of the first Surface, which carried NVIDIA's Tegra processor ⁶.

What NVIDIA's Jeff Yen said

Jeff Yen of NVIDIA did the explaining, as AI Watch and PC Watch tell it. These are NVIDIA's own claims, not numbers measured by the outlets.

• Why it was built: according to AI Watch, Yen said the biggest push behind RTX Spark was the explosive growth in demand for AI agents that run autonomously on the PC itself [[F483]]
• The gap to the cloud: AI Watch says Yen described the lag in intelligence and accuracy between the top cloud models and local models as shrinking from a gap of several months, as long as 8, down to about 2 months [[F484]]
• A cost estimate: per AI Watch, Yen showed an estimate for the same large batch of tokens: about 1,200 dollars through a cloud API, versus an electricity bill of roughly 20 dollars if open models such as Qwen 3.8 run locally [[F485]].
• On battery: PC Watch says Yen pitched that RTX Spark can deliver 98% of its AC-powered performance on battery [[F476]]
• A hybrid setup: also per PC Watch, local agents can be combined with frontier models in the cloud when needed [[F477]]
• Versus DGX Spark: AI Watch says NVIDIA stressed that, unlike the Linux-based DGX Spark, running Windows widens what the machine is practical for [[F486]]

The demos

The demo list was ambitious.

• Games: PC Watch reports that an esports title with anti-cheat, something Windows on Arm could not handle before, was shown running [[F478]]. AI Watch lists play demos of NBA 2K27 with DLSS 5 neural rendering and of Marvel Rivals with anti-cheat support [[F488]]
• Video generation: PC Watch says the event showed video generation in 4K/120fps/HDR [[F479]], and AI Watch adds a live demo of LTX 2.5 [[F488]]
• Video editing: in the hands-on area, a Hermes Agent assisted with editing inside DaVinci Resolve, and demos of scheduling by a local personal agent and of coding support were shown, according to AI Watch [[F487]]
• Daily tasks: PC Watch describes a demo where an agent pulls the day's tasks out of chat and email, and another where robot development is almost completed just by asking an agent [[F480]]
• A robot duck: AI Watch reports a demo in which Hermes Agent simulated the motion of the duck-shaped robot Microduck [[F489]]

Games, video, agents and a robot duck: the demos covered a lot of ground.

HP's president, and a headline

The next morning in Japan, TBS CROSS DIG with Bloomberg promoted an episode with Ketan Patel, president at HP. The headline of the post reads, in translation: "Your PC will soon be used by you plus 10 AI agents" ⁷.

The program's own blurb, written in Japanese, says RTX Spark lets very large AI models run on the PC, depending on the machine ⁸. That is TBS's wording, not an RTX Spark spec sheet. It also asks whether working by giving instructions from a phone while the PC executes will become the norm, and whether everyone will someday have a "small data center" at home ⁹.

If you are shopping

If these scenes made you want to shop, the product cards below show laptops from the partners named at the event. Before you buy, check the official support information for the tools you plan to run.

You can also browse RTX Spark laptops on Amazon.

Read next

Torn on memory size? Start with our comparison of memory sizes across outlets' tests. For how much to let a local agent touch, see how far to let your local AI agent go. For the battery angle, see running a big model with the charger unplugged. The companion ComfyUI piece is the ComfyUI article.