Nvidia’s AI Blueprint Stretches from Seoul’s Data Centres to Your Desktop PC

Nvidia partners with Naver, LG for sovereign AI in Asia, teams with Microsoft on RTX Spark for desktop AI agents, and readies Vera Rubin chips.

The debate over whether Nvidia is a cyclical chip supplier or a permanent infrastructure layer just got a fresh test. While the market mulls that question, the company is locking in deals on two distinct fronts — sovereign AI builds in Asia and a push to embed its software into everyday Windows machines. The result is a narrative that no longer hinges solely on hyperscaler spending.

Seoul as a Sovereign AI Hub

Nvidia’s chief executive Jensen Huang was in the South Korean capital this week to cement partnerships that go far beyond chip sales. The centrepiece is an alliance with internet giant Naver to build a series of “AI factories”. The first, named Gak Sejong, will come online in the first half of 2027 with an initial capacity of 55 megawatts, with Naver already planning an expansion to 200 megawatts.

The logic underpinning these projects is “Sovereign AI” — the idea that nations and local champions must run their own AI infrastructure independently of foreign providers. Huang also met LG chairman Koo Kwang-mo to discuss robotics and autonomous driving, with LG building its own AI factory for cloud services on Nvidia’s technology. Additional technology partnerships cover SK Telecom and SK Hynix, tying Nvidia into the semiconductor supply chain as well as the data centre build-out.

The Desktop Dimension: RTX Spark and Windows Agents

At the same time, Nvidia and Microsoft are positioning the RTX Spark platform as a new class of Windows PC designed for personal AI agents. Nvidia’s software stack is embedded directly into the Windows environment, and major PC makers are already on board. Microsoft frames the project as part of a broader collaboration across gaming, AI and cloud, while Nvidia sees RTX Spark as a way to move local AI agents away from a dependence on remote infrastructure.

The company also announced the DGX Station for Windows, a desktop system for developing and running agents. The strategic intent is unmistakable: Nvidia wants its platform to be as entrenched on the desk as it is in the data centre. This second layer of demand — from creative workstations, enterprise desktops and personal machines — could broaden the addressable narrative well beyond the hyperscaler budgets that currently dominate the stock’s fortunes.

Vera Rubin: The Hardware Engine

Technologically, these initiatives rest on the forthcoming Vera Rubin platform, the successor to the Blackwell architecture. Vera Rubin combines Rubin graphics processors with an 88-core CPU manufactured by TSMC on a 3-nanometre process. Production is already running at full tilt, with first shipments to large cloud providers such as AWS and Google scheduled for the second half of 2026.

To avoid the supply bottlenecks of previous generations, Nvidia now taps a network of more than 350 factories globally. A new wireless server?rack design slashes the assembly time for AI clusters by roughly 95 percent, accelerating the conversion of its massive order backlog into live data centre capacity.

Stock Consolidation: A Pullback, Not a Break

Nvidia shares have cooled from the euphoria of earlier this year, but the decline looks more like a recalibration than a reversal. The stock last traded at €177.52, which is 12.34 percent below the 52?week high of €202.50 reached on 14 May 2026. Yet it still sits 44 percent above the 52?week low of €122.90 from June 2025 — a reminder of the long?term premium investors continue to assign.

The moving averages tell a similar story. At €177.52, the stock is just above its 50?day moving average of €175.49 and comfortably above the 200?day line of €161.71. The relative strength index stands at 45.1, suggesting a market that is questioning the slope of the growth curve without abandoning the thesis altogether. Annualised 30?day volatility of 43.48 percent means a rough week can happen quickly, but that is not the same as a broken story.

Year?to?date the shares have gained 43 percent, while the 12?month performance is 42.49 percent. The 30?day decline of roughly 5 percent reflects a more demanding environment, partly triggered by a sector?wide sell?off in US?listed chipmakers after weakness in Broadcom’s quarterly report. As the reference asset for the entire AI theme, Nvidia now absorbs nervous signals from across the supply chain rather than moving solely on its own announcements.

The Platform Question

The consensus analyst price target stands at €257.88, implying upside of more than 45 percent. That is not a promise — it is a measure of how much confidence still resides in the earnings power of the platform and how much execution the market expects.

A market capitalisation of roughly €4.3 trillion leaves little room for a merely good story. Nvidia must keep proving that AI demand is structural rather than a capital?expenditure surge concentrated on a handful of cloud clients. The RTX Spark and Windows?agent push directly addresses that challenge by extending the AI use case from model training and cloud inference to local workflows, software ecosystems and user behaviour.

If AI remains largely centralised, Nvidia can remain hugely important, but the stock stays hostage to hyperscaler budgets and chip cycles. If AI becomes a distributed computing layer on PCs, desktops and creative tools, the addressable narrative widens significantly. The recent dividend of $0.25 per share underscores that this is not an income story — it is a question of platform permanence. Whether personal agents translate into real workloads, and whether Nvidia remains the toll collector, will not be answered by a keynote. That answer will come in the quarters ahead, when announcements turn into revenue.