NVIDIA Wants to Own the Whole PC Now

NVIDIA Is Done Being the Component You Bolt On

For three decades, NVIDIA's place in the PC was clearly defined: it made the graphics card. Intel or AMD ran the CPU. Microsoft ran the OS. OEMs bolted it all together and sold you a box. That arrangement made NVIDIA extraordinarily rich, and it made every other stakeholder comfortable.

RTX Spark is NVIDIA's formal resignation from that arrangement.

NVIDIA unveiled RTX Spark at Computex 2026 — a new superchip that the company says reinvents Windows PCs for the era of personal AI agents, offering "a new class of computer that moves from tool to teammate." The hardware specs are the first signal that this is something genuinely different. The chip fuses two chiplets into one SoC: a Blackwell GPU with advanced Tensor Cores, connected to a multi-core Grace CPU, with unified memory supporting substantial AI models in local memory — something conventional gaming laptops cannot match.

That architectural choice is the key differentiator. Instead of splitting system RAM and video memory, RTX Spark exposes unified memory as one pool shared by CPU and GPU — the same recipe NVIDIA used in its GB10 Grace Blackwell module, and what lets a thin laptop hold substantial language models entirely in memory.

MediaTek, described as a leader in Arm-based SoC designs, collaborated with NVIDIA on the custom CPU design, contributing to its power efficiency, performance, and connectivity. The CPU is an Arm-based design featuring a mix of high-performance and efficiency-focused cores.

What Jensen Huang Actually Said

The keynote framing was unambiguous. Huang emphasized that the PC is being reinvented, moving from a tool where you launch apps to one where you can ask the system to do work. He positioned RTX Spark as bringing NVIDIA's core technologies — CUDA, RTX, and its AI platform — into a single superchip.

In follow-up remarks, Huang emphasized the company's commitment to this shift, suggesting it as a significant platform move for the industry.

That commitment extends beyond a single generation. NVIDIA indicated that future generations of its platforms will include Spark variants, with planned roadmap successors, representing a multi-generational platform commitment rather than a one-off experiment.

The Apple Question NVIDIA Won't Quite Answer

The architecture — CPU, GPU, and unified memory on one die — is structurally similar to what Apple has been shipping in M-series chips since 2020. Every journalist at Computex noticed. When directly asked about competing with Apple Silicon, Huang deflected with a compliment: "Apple's ecosystem is excellent as you know, and they have a world-class silicon roadmap. But that's not our focus. Our focus is to reinvent the PC. Windows is of course improving, but the basic architecture of a PC has largely been the same now for about 40 years, and we want to reinvent it."

That's a diplomatic non-answer. RTX Spark is absolutely entering Apple Silicon territory — unified memory, efficiency-first design, on-device AI inference. The difference is NVIDIA is doing it in the Windows ecosystem, not building a closed garden. Which means the real competitive pressure isn't Apple-vs-NVIDIA; it's Intel and AMD, who have been selling x86 laptop chips into this market for decades and now face a credible challenger that has CUDA, DLSS, and the entire generative AI developer stack baked in at the silicon level.

For decades, the Windows PC stack had a stable shape: Intel or AMD supplied the CPU, Microsoft supplied Windows, NVIDIA or AMD supplied graphics when needed, and OEMs assembled the final machine. RTX Spark challenges that arrangement by putting CPU, GPU, unified memory, CUDA support, and AI acceleration into a single platform pitch.

The Honest Caveats

The specs are real; the experience isn't yet testable at scale. Windows on Arm still has compatibility baggage, battery life claims need independent testing, and OEM thermal designs will determine whether Spark feels like a breakthrough or another very expensive demo.

NVIDIA confirmed these chips will run across a range of power envelopes, with each OEM free to pick its own cooling approach. In a thin chassis, the chip will likely be power-limited, which in turn makes it thermally and frequency-limited. That's not a death sentence — it's physics — but it means the OEM execution will matter enormously. A thicker, better-cooled chassis will outrun an identically specced wafer-thin machine on sustained workloads.

There's also the gaming angle, which NVIDIA is careful not to over-promise. RTX Spark systems won't have discrete GPU capability, which affirms that high-end gaming isn't the focus — NVIDIA will likely target "good enough" gaming for the mainstream and push hard on agentic AI as the main selling point. Whether that's enough to sell laptops in a category where gamers often drive the premium tier remains to be seen.

The Fall Launch Line-Up

NVIDIA said multiple laptops and desktops using RTX Spark will launch this fall from partners including Dell, HP, Asus, Lenovo, and MSI, initially targeting creators, gamers, and AI developers. Microsoft Surface is also in that list — the Surface Laptop Ultra was described as symbolically important because it gave Microsoft something it has lacked for years: a Surface machine that feels like it is setting the agenda rather than following it.

The breadth of OEM support matters as much as the chip itself. Building a full product and partner ecosystem is a much larger challenge than simply building and shipping a chip — those OEM and software partners need to trust it's worth committing time and resources to NVIDIA's platform. The multi-gen roadmap announcement is partly about the chip, and partly about giving Dell and Lenovo and HP a reason not to treat this as a one-off SKU.

The strategic read here is simple: NVIDIA no longer wants to be the component you add to a PC. It wants to be the reason the PC exists. Whether that ambition survives contact with actual consumers — with their x86 software libraries, their anti-cheat requirements, their skepticism about Windows on Arm — is the question that won't get answered until fall shipments hit real hands.

---