Nvidia Unveils RTX Spark Superchip to Reinvent the Windows AI Laptop

Nvidia CEO Jensen Huang used his Computex 2026 keynote in Taipei to unveil the RTX Spark Superchip — a single-package processor that combines a Blackwell GPU, an Arm-based Grace CPU, and 128GB of unified memory inside a thin Windows laptop. It's Nvidia's most direct push yet into the client PC market, and it puts the company on a collision course with Intel, AMD, Apple, and Qualcomm.

Devices powered by RTX Spark are scheduled to ship this fall from Dell, HP, ASUS, Lenovo, MSI, and Microsoft. Huang said more than 30 RTX Spark laptops and roughly 10 desktop systems will lead the launch.

What's inside the chip

At full configuration, the RTX Spark Superchip includes up to 20 Arm CPU cores, a Blackwell GPU with 6,144 CUDA cores, 128GB of LPDDR5X unified memory, and up to 300 GB/s of memory bandwidth. The Blackwell GPU is roughly equivalent in performance to a desktop RTX 5070, according to Nvidia.

The CPU is a new Arm-based design Nvidia is calling Grace for client. It was co-engineered with MediaTek and is the first time Nvidia has put a custom Arm CPU into a consumer PC product. The chip also carries the full CUDA software stack — meaning developers can run the same AI workloads that today require a desktop tower or cloud GPU.

Why this matters: until now, "AI PC" mostly meant a laptop with a modest neural processing unit that could run small models. RTX Spark moves that bar to running models with tens of billions of parameters locally, without a cloud round trip.

A "new era" for Windows on Arm

Nvidia and Microsoft co-engineered the platform. Huang spent much of the keynote framing RTX Spark less as a chip and more as a new Windows form factor. He claimed the platform will "run every Windows app ever made," addressing the compatibility problem that has plagued previous Windows-on-Arm attempts from Qualcomm and others.

The pitch goes further than compatibility. Nvidia and Microsoft are positioning RTX Spark machines as the first true agentic AI PCs — laptops where autonomous agents can read your screen, control applications, and complete multi-step workflows locally. Microsoft is expected to elaborate on this vision at Build 2026 on June 2–3 in San Francisco.

The Microsoft Surface Laptop Ultra, announced alongside the chip, will be the flagship device. Dell, HP, ASUS, Lenovo, and MSI will ship competing designs in the same window.

Industry impact

For Intel and AMD, RTX Spark is the most serious external threat to their x86 laptop business in a decade. Both companies have spent the past two years marketing AI PC chips with integrated NPUs, but neither has a discrete GPU in the same league as Blackwell. Apple's M-series chips are competitive on efficiency but locked to macOS. Qualcomm's Snapdragon X line has Windows-on-Arm reach but no comparable AI silicon.

For developers, RTX Spark is a meaningful shift. The full CUDA stack on a laptop means an engineer can prototype, fine-tune, and run agentic workloads on the same machine, then deploy unchanged to cloud GPU servers. That's the workflow Apple has provided for Mac developers shipping to iPhone; Nvidia and Microsoft are trying to replicate it for the entire Windows ecosystem.

For enterprise IT, the appeal is more pragmatic. Running AI workloads locally on a corporate laptop means sensitive data does not need to leave the machine, simplifying compliance with regulations like the EU AI Act and HIPAA. Several IT decision-makers cited by CNBC said they expect to begin pilots with RTX Spark hardware in the fourth quarter.

What's next

Three things to watch over the coming months.

First, real-world performance. Nvidia's claims about local model performance need third-party benchmarking. Reviews from outlets like Tom's Hardware and Notebookcheck typically land within days of shipping hardware, and they will determine whether RTX Spark machines actually deliver desktop-class AI in a laptop chassis.

Second, pricing. Nvidia and its OEM partners have not disclosed prices. The 128GB unified memory configuration alone is expensive — analysts at Tom's Hardware estimated the bill of materials at $1,500 to $2,000, which would put fully configured RTX Spark laptops well above $3,000.

Third, the response from Apple, Intel, and AMD. Apple's WWDC 2026 keynote is on June 9. Intel's Panther Lake successor and AMD's next-generation Strix Halo refresh are both expected this fall. The competitive response will reshape the entire PC market.

Bottom line

RTX Spark is the first time Nvidia has put a desktop-class AI experience inside a Windows laptop, and the first credible attempt at making Windows-on-Arm a serious developer platform. If the launch ships on time and the software stack holds up, it will redefine what "AI PC" means — and force every other chip vendor to react. If you are buying a workstation laptop this fall, wait for the reviews.