Nvidia's RTX Spark Laptops Go On Sale From Six PC Makers
Windows laptops built around a single Nvidia chip went on sale worldwide from ASUS, Dell, HP, Lenovo, Microsoft and MSI, starting at $2,599 and designed to run large AI models on the laptop itself rather than in a distant data center.
Nvidia's RTX Spark laptops went on sale worldwide from ASUS, Dell, HP, Lenovo, Microsoft and MSI, through the makers' websites and major retailers. Prices start at $2,599. Models include the Dell XPS 16, the HP OmniBook Ultra, the Lenovo Yoga Pro 9n and Microsoft's Surface Laptop Ultra.
Each laptop runs on a single Nvidia "superchip" that combines a Blackwell graphics processor and a Grace main processor sharing up to 128 gigabytes of memory. Nvidia says that is enough to run large AI models, the kind usually reached over the internet, directly on the laptop, so a user's files and prompts can stay on the device.
Nvidia also says the machines deliver up to one petaflop of AI computing, a quadrillion simple calculations a second, and can play demanding games at more than 100 frames a second.
"RTX Spark brings the power of NVIDIA AI, RTX graphics and CUDA together with Microsoft Windows in one PC," said Jensen Huang, Nvidia's founder and chief executive. Adobe has released test versions of Photoshop and Premiere built for the chip.
Santa Clara, Calif. — Nvidia's RTX Spark laptops, Windows PCs built around a single Nvidia chip, went on sale worldwide from six PC makers: ASUS, Dell, HP, Lenovo, Microsoft and MSI, with more systems to follow.
The machines are sold through the makers' own websites and major retailers, starting at $2,599, with prices and configurations varying by country. The first lineup includes the ASUS ProArt P14 and P16, the Dell XPS 16, the HP OmniBook Ultra and OmniBook X 14, the Lenovo Yoga Pro 9n, Microsoft's Surface Laptop Ultra, and MSI's Prestige N16 and Prestige N16 Flip AI+.
At the center of each one is what Nvidia calls a superchip: a Blackwell graphics processor, the part that draws images and does the heavy math for artificial intelligence, joined on one package with a Grace central processor, the part that runs the operating system and everyday programs. Taiwan's MediaTek helped design the central processor. The two halves share one pool of up to 128 gigabytes of memory, so a large AI model and a big video or 3D project can sit in memory at the same time.
That shared memory is the point of the design. Most people reach powerful AI models such as chatbots over the internet, where the computing happens in a company's data center. Nvidia is pitching RTX Spark as a way to run those models on the laptop instead, so code, documents and prompts stay on the user's machine. According to Nvidia, the chip delivers up to one petaflop of AI performance, meaning a quadrillion low-precision calculations a second, can edit 8K video and render 3D scenes as large as 90 gigabytes even when unplugged, and plays major games at 1440p resolution at more than 100 frames a second. Games supported at launch include Cyberpunk 2077 and NBA 2K27.
"RTX Spark brings the power of NVIDIA AI, RTX graphics and CUDA together with Microsoft Windows in one PC for creating, gaming and running personal AI agents," said Jensen Huang, Nvidia's founder and chief executive. CUDA is Nvidia's software platform for programming its chips, and AI agents are programs that carry out multistep tasks, such as writing and testing code, on a user's behalf. Yusuf Mehdi, an executive vice president at Microsoft, said Windows is being built so that AI "can run locally when it makes sense and reach the cloud when needed."
Software makers released versions on day one. Adobe put out beta, or test, versions of Photoshop and Premiere with native support for the chip, with Substance 3D to follow. Nvidia and Microsoft prepared developer tools, including the first build of the PyTorch AI framework with Nvidia's CUDA for Windows running on Arm-based chips, the low-power chip design used in phones and now in these laptops. Nvidia also released a 74-billion-parameter AI model, Nemotron 3.5 Bolt, sized to run on supported RTX Spark machines. Parameters are the internal settings a model learns during training, and their number is a rough measure of its size.