NVIDIA RTX Spark Officially Launches at Computex 2026: 20-Core Arm CPU, Blackwell GPU, and 128GB Unified Memory Coming From Six OEMs This Fall
NVIDIA officially unveiled the RTX Spark Superchip at Computex 2026 — a 20-core Arm Grace CPU paired with a Blackwell GPU and up to 128GB of unified memory, launching in laptops from six OEMs in fall 2026.
Editor's Note ·
- Clarification:
- The article quotes Brett Ostrum, Microsoft's Corporate VP, as saying 'The work from creators, developers and AI builders has a common shape: massive scenes, long compile cycles, local models and datasets that no longer sit politely in the background. We built Surface Laptop Ultra to meet that work without flinching.' This quote was attributed to Axios, which bot-blocked our snapshot crawler (HTTP 403). The quote could not be independently verified verbatim from any readable source. The NVIDIA press release and other outlets contained different Ostrum quotes. The substance of Ostrum's statement — that Surface Laptop Ultra targets AI builders and creators — is accurate and confirmed by multiple sources.
Overview
NVIDIA officially unveiled the RTX Spark Superchip at CEO Jensen Huang’s Computex 2026 keynote on June 1, marking the company’s formal entry into the consumer laptop processor market after decades of absence from that segment. The chip — co-developed with MediaTek — pairs a 20-core Arm-based Grace CPU with a Blackwell GPU containing 6,144 CUDA cores and fifth-generation Tensor Cores, connected via NVIDIA’s NVLink-C2C interface, according to the official NVIDIA press release. The platform ships in laptops from six OEMs — ASUS, Dell, HP, Lenovo, Microsoft, and MSI — with Acer and GIGABYTE products following later, per NVIDIA’s announcement.
This follows The Machine Herald’s prior coverage from May 22, when the chip was still referred to by its development codename N1/N1X and no official product name or OEM device assignments had been confirmed.
What We Know
RTX Spark Superchip specifications. The chip features 20 CPU cores — a mixed configuration of 10 high-performance Arm Core X925 cores and 10 efficiency Cortex A-725 cores — alongside a Blackwell GPU with 6,144 CUDA cores and fifth-generation Tensor Cores with FP4 precision, according to PCWorld and the official NVIDIA press release. The integrated GPU delivers performance comparable to an RTX 5070 laptop GPU, Engadget reported. The platform claims 1 petaflop of AI compute, per Engadget, and supports AI language models of up to 120 billion parameters with a context length of up to 1 million tokens locally, according to the NVIDIA press release. NVIDIA also claims the platform can deliver 100+ fps at 1440p in gaming, enabled by DLSS 4.5 Ray Reconstruction with a second-generation transformer model, per the official announcement. The chip operates anywhere from single-digit watts up to 80W depending on workload, Engadget noted.
Memory and configuration. RTX Spark devices will be available with a unified memory pool ranging from 16GB to 128GB, shared between CPU and GPU, according to Engadget.
MediaTek co-development. The Grace CPU was developed in collaboration with MediaTek, while NVIDIA handles the Blackwell GPU architecture, the NVLink-C2C interconnect, and the platform software stack, per the NVIDIA press release.
OEM partners and device names. The initial fall 2026 RTX Spark laptop lineup includes the following devices, per the NVIDIA announcement:
- Microsoft: Surface Laptop Ultra
- Dell: XPS 16 Creator Edition
- HP: OmniBook models
- ASUS, Lenovo, and MSI (specific model names to be announced)
Acer and GIGABYTE will follow with devices at a later date. Laptops will span screen sizes from 14 to 16 inches and can be as thin as 14mm, according to Axios.
Microsoft Surface Laptop Ultra. Microsoft’s device is the flagship launch product. The Surface Laptop Ultra features a 15-inch PixelSense Ultra touchscreen with mini-LED backlighting — the first mini-LED display in any Surface laptop, as PCWorld reported — with peak HDR brightness of 2,000 nits, per Engadget. The laptop weighs less than 4.5 pounds and ships with dual fans, two Thunderbolt ports, HDMI, USB-A, USB-C, an SD card slot, and a headphone jack, according to PCWorld. It will be available in Platinum and Nightfall finishes, per PCWorld. Pricing has not been announced; the device is expected to ship in fall 2026, Engadget confirmed.
Software ecosystem. Adobe has committed to optimizing Photoshop and Premiere specifically for RTX Spark. Game compatibility on the Windows on Arm platform relies on a combination of native Arm builds, Prism emulation, and anti-cheat software support, Engadget noted.
What the Companies Said
Jensen Huang positioned the RTX Spark as a fundamental shift in how PCs are used. “The PC is being reinvented. For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — the PC does the work,” Huang said, according to the official NVIDIA press release.
Brett Ostrum, Microsoft’s Corporate VP, described the Surface Laptop Ultra’s target user in a statement reported by Axios: “The work from creators, developers and AI builders has a common shape: massive scenes, long compile cycles, local models and datasets that no longer sit politely in the background. We built Surface Laptop Ultra to meet that work without flinching.”
Andrew Hill, Microsoft’s Corporate VP of Surface, said the device is “the most powerful thing we’ve ever made,” according to Engadget.
What We Don’t Know
Neither NVIDIA nor its OEM partners have announced retail pricing for any RTX Spark laptop. Specific model names from ASUS, Lenovo, and MSI have not been disclosed. The exact manufacturing process node was not confirmed in official materials released at Computex. Real-world gaming performance will depend on the maturity of Windows on Arm software compatibility and anti-cheat support at the time devices ship. Detailed configuration options — memory tiers, storage capacities, and pricing — for all OEM models remain unannounced.
Analysis
The RTX Spark announcement is NVIDIA’s most significant product diversification since it exited the consumer mobile processor market. By integrating a Blackwell GPU — the same architecture it sells to hyperscalers in rack-scale AI systems — into a laptop SoC developed with MediaTek, NVIDIA is applying the same unified-memory architecture that Apple pioneered in its M-series chips while adding the RTX software ecosystem that no Arm laptop competitor can match: DLSS 4.5, Tensor Core AI acceleration, and an installed base of thousands of RTX-compatible games and creative applications.
The competitive positioning is direct. Qualcomm has dominated Windows on Arm with its Snapdragon X series, and Apple’s M-series remains the benchmark for Arm laptop performance-per-watt. NVIDIA enters with a GPU-first identity that differs from both: raw AI compute throughput and a gaming GPU lineage in a laptop SoC. Whether that differentiation translates to sustained software compatibility and real-world performance will not be assessable until independent benchmarks emerge when devices ship later this year.