Nvidia's RTX Spark is the most ambitious Windows laptop chip in years: 20 CPU cores, 6,144 GPU CUDA cores, 128GB of unified LPDDR5X memory, and integrated graphics Nvidia compares to its RTX 5070 Laptop GPU. The first wave of machines ships in fall 2026 across Microsoft, Dell, Asus, Lenovo, HP, MSI, Acer, and Gigabyte. The pitch is straightforward — finally give Windows the kind of performance-per-watt jump Apple delivered with the M1 in 2020. The catch is the price tag.
Existing or comparable models in that announced lineup typically start at $2,000 to $2,500 and climb from there. The closest x86 analog, an AMD Strix Halo machine with 128GB of RAM, already runs $3,000 for the Asus ProArt PX13 GoPro Edition and $3,300 for the ROG Flow Z13. Nvidia's own DGX Spark desktop, built around the same GB10 chip family, costs roughly $4,700. A 15-inch Mini LED Surface Laptop Ultra with 128GB of unified memory is not going to undercut any of that.
Nvidia calls RTX Spark a 'superchip' and 'the most efficient PC chip ever built.' Microsoft is billing its Spark-equipped Surface Laptop Ultra as 'the most powerful thing we've ever made.' Adobe has signed on with optimized builds of Photoshop and Premiere. Notably absent from the launch: any published benchmarks or first-party performance numbers. Nvidia is asking the market to trust the spec sheet until fall.
Key facts
- 01RTX Spark packs 20 CPU cores, 6,144 GPU CUDA cores, and 128GB of unified LPDDR5X memory, with integrated graphics Nvidia compares to an RTX 5070 Laptop GPU.
- 02Fall lineup includes Microsoft's Surface Laptop Ultra, Dell XPS 16, Asus ProArt P14 and P16, Lenovo Yoga Pro 9n, HP OmniBook Ultra, MSI Prestige N16 Flip AI Plus, plus Acer and Gigabyte models.
- 03Comparable AMD Strix Halo machines with 128GB RAM already run $3,000–$3,300; Nvidia's own DGX Spark desktop with the GB10 chip costs about $4,700.
- 04Existing models in the announced lineup typically start at $2,000 to $2,500, with top configurations expected well above that.
- 05Nvidia CEO Jensen Huang called 'CPUs for agents' the company's new major growth driver during the two-hour keynote unveiling the chip.
Most of Nvidia CEO Jensen Huang's two-hour keynote was not about laptops. It was about agents. Huang framed 'CPUs for agents' as the company's new major growth driver, and RTX Spark is the consumer endpoint for that story — a Windows machine that can run sizable models locally without round-tripping to a data center. The 128GB unified memory pool is the giveaway. That capacity is overkill for browsing and Office, and it is exactly right for hosting a local agent that needs to keep a large model and its context resident.
Whether that pitch resonates with buyers writing the check is the open question. Apple's M1 moment worked because Apple started at the bottom of the stack. The original M1 shipped in the Mac mini, the MacBook Air, and the entry-level MacBook Pro. Mass-market pricing drove mass-market adoption, which gave developers a reason to ship native Arm builds. The M1 Pro and M1 Max did not arrive until nearly a year later.
Nvidia is inverting that order. RTX Spark is, in effect, an M1 Max or M1 Ultra-class debut with no Air-class machine underneath it. Nvidia has said lower-RAM Spark configurations will follow, but the launch SKUs are flagships, and the broader laptop market is already absorbing memory-pricing shocks that have pushed even 16GB and 32GB systems higher. The MacBook Neo landed at $599 and rearranged the consumer conversation. A $2,499 estimated entry price for a Spark laptop does not move the same volume of units.
The competitive picture for Windows on Arm gets more interesting either way. Buyers will have four viable chip lines this fall: Intel for balanced x86 compatibility, AMD for raw performance at some battery cost, Qualcomm for the longest battery life with weaker game support, and Nvidia for what should be the strongest graphics-and-AI combination on Arm. Microsoft and Nvidia have also pushed Riot Games to port its anti-cheat to Arm for Valorant and League of Legends, with similar work happening across Easy Anti-Cheat, BattlEye, and Denuvo. Gaming has been the soft spot for Arm on Windows, and that gap is finally closing.
The risk for Nvidia is that the technical win arrives in a year when consumers have less room to spend on a $2,500-plus laptop, and developers have less incentive to optimize for a chip that ships in low volumes at the top of the price stack. Apple solved the cold-start problem with cheap Macs. Nvidia is starting from the opposite end. Strong agent and creator demos will help, but the install base that makes a platform sticky is built at the entry level, not at the workstation tier.
RTX Spark is the most credible threat to Apple silicon's performance lead since the M1 shipped, and it is also the clearest sign yet that Nvidia views the PC as an AI endpoint rather than a general-computing device. The 128GB unified memory configuration only makes sense if Nvidia expects buyers to run real models on the laptop. That is a bet on where consumer AI ends up — local rather than cloud-only — and it is being placed at a price point that will test how badly the high end of the Windows market wants it.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
