The NVIDIA RTX Spark laptops October 16 launch marks the day NVIDIA’s first end-to-end silicon platform for Windows AI PCs finally reached retail shelves. Six PC makers – ASUS, Dell, HP, Lenovo, Microsoft, and MSI – began shipping nine laptop models built around a single superchip that pairs an NVIDIA Blackwell RTX GPU and an NVIDIA Grace Arm CPU on one package, sharing a unified LPDDR5X memory pool of up to 128 GB. NVIDIA framed the rollout as the start of a category, not a product refresh, and priced the platform aggressively enough to draw attention from workstation vendors and notebook OEMs alike.
At the silicon level, RTX Spark is sold in two N1X tiers. The entry configuration pairs an 18-core Grace CPU with 5,120 Blackwell CUDA cores and a smaller LPDDR5X pool, while the top tier steps to a 20-core Grace CPU with the full 6,144 CUDA cores enabled and the maximum 128 GB of unified memory. Because the CPU and GPU draw from the same memory fabric, NVIDIA argues that large language models and diffusion pipelines can be staged without the round trips that plague discrete GPU laptops with narrow PCIe buses. The headline numbers – up to one petaflop of FP4 AI throughput and the ability to run models with up to 120 billion parameters entirely on-device with context windows reaching one million tokens – are aggressive even by workstation standards, and they explain why every launch partner built a flagship SKU around the chip.
Microsoft’s Surface Laptop Ultra is the most documented launch system and effectively serves as the reference design. Microsoft priced the entry model at $2,599.99 for the 18-core, 5,120-core, 24 GB, 512 GB configuration, and the flagship at $5,899.99 for the 20-core, 6,144-core, 128 GB, 1 TB build. Microsoft’s Pavan Davuluri was direct about the partnership, stating that the company “built Surface Laptop Ultra around NVIDIA RTX Spark” rather than the other way around – a notable admission for a vendor that typically drives its own silicon roadmap. ASUS pushed pricing even higher with the ProArt P16 RTX Spark, topping out at $6,999.99 for a 128 GB, 2 TB creator configuration aimed at the same buyers who previously spec’d mobile workstations with discrete RTX 6000 Ada cards.
Pricing across the six OEMs is split by roughly $200 at the entry configurations. NVIDIA’s official blog lists starting prices from $2,399, while CNET and Tech Insider’s retail tracking has OEM starting prices clustered at $2,599 across the lineup, a gap that reflects distributor markups, bundle configurations, and the fact that several partners chose to launch only at higher tiers. With nine SKUs available on day one and more in certification, buyers have a real spread of chassis, display, and storage options to choose from rather than a single reference design.
NVIDIA RTX Spark laptops October 16 launch: Berlin, Beijing, Seoul, San Francisco, Taipei, Istanbul
NVIDIA treated the launch as a global retail moment rather than a press release. The company scheduled six launch events worldwide, with Berlin’s Festival of Lights as the centerpiece. The Berlin TV Tower was lit in custom RTX Spark themes, and Jensen Huang appeared in person – a rare on-stage appearance that drew press coverage well beyond the hardware beat. Parallel events ran in Beijing, Istanbul, Seoul, San Francisco, and Taipei, each pairing retail availability with hands-on demos of on-device model inference. The geographic spread signals that NVIDIA expects RTX Spark to land as a holiday-season purchase, not a niche workstation order.
Why 120 Billion Parameters on a Laptop Matters
The 120-billion-parameter claim is the technical anchor of the launch. Running a model of that scale locally – meaning no cloud round trip, no subscription, and no data leaving the chassis – has been the white whale of on-device AI for three years. RTX Spark’s combination of one petaflop of FP4 throughput, 128 GB of unified memory, and a million-token context window is the first time a shipping laptop has credibly targeted that envelope. For PC enthusiasts, it means locally hosted coding assistants, image generators, and reasoning agents can run at the same scale that previously required a data-center accelerator. For OEM partners, it gives a marketing line that does not require asterisks. For AI-on-laptop buyers, it finally narrows the gap between what a thin-and-light can do and what a cloud endpoint can do.
Compact desktops are the obvious next move, and NVIDIA has already committed to a November 2026 ship date for small-form-factor RTX Spark systems aimed at developers and prosumers who do not need a battery. That cadence – laptops first, mini-PCs eight weeks after – mirrors the strategy NVIDIA used to seed RTX Studio into the creator market, and it gives software developers a holiday-season install base to target before desktop reviewers start benchmarking. The NVIDIA RTX Spark laptops October 16 launch will reshape how investors, regulators, and competitors weigh the next round of public-purpose commitments.
The NVIDIA RTX Spark laptops October 16 launch is best read as the start of a category rather than the end of a product cycle. With Blackwell-class AI throughput on every launch SKU, a unified memory pool that removes the old CPU-GPU bottleneck, six OEM partners, and retail footprints on four continents, NVIDIA has turned a superchip reference platform into a shopping event. Whether the 120-billion-parameter claim holds up under sustained workloads will be the story of Q1 2027, but on launch day the lineup is real, the prices are public, and the shelves are stocked.