Loading page…
The Nvidia-powered laptop promises large-model capacity without cloud charges for local tasks. Its headline computing figure is theoretical, and demonstrated coding workflows still use the cloud.
Listen to this story
The Surface Laptop Ultra gives developers a $2,599 option for running some AI coding work on-device, with preorders open October 7, 2026, and shipping scheduled for October 16. Its local-model pitch comes with important limits: the 120-billion-plus-parameter claim and one-petaflop figure are manufacturer specifications, the latter based on theoretical FP4 performance with sparsity, and 128GB is shared system memory rather than a guaranteed GPU allocation. Microsoft’s demo also sent some requests to the cloud, so buyers can reduce cloud use for selected tasks, not eliminate it.
The RTX Spark Superchip combines a Blackwell GPU with up to 6,144 cores and a Grace CPU with up to 20 cores.
Microsoft says the laptop’s redesigned thermal architecture provides up to 2.5 times the thermal power capacity of current Surface Laptops.
The 15-inch touchscreen has a stated peak HDR brightness of 2,000 nits over a 10% window; it supports touch but not pen input.
Microsoft is offering developers a way to run large AI models on a portable computer rather than send every task to the cloud. The company opened Surface Laptop Ultra preorders on October 7, 2026, starting at $2,599. Shipping begins October 16, putting a price and delivery date on its Nvidia-powered local-AI laptop.
The laptop is built around Nvidia’s RTX Spark Superchip. Microsoft specifies a Blackwell RTX graphics processor with up to 6,144 cores and a Grace CPU with up to 20 cores. Configurations reach 128GB of unified memory: a shared pool used by both processors, rather than separate allocations for each.
Microsoft says the machine can run AI models exceeding 120 billion parameters locally and deliver up to one petaflop of AI performance. That computing figure is theoretical, based on FP4, a low-precision calculation format, with sparsity enabled. It is not a measured speed result for an everyday coding task.
Memory capacity also needs a distinction: 128GB is the maximum total system memory, not the amount guaranteed to be available to the graphics processor. Microsoft says the GPU’s addressable share is smaller and depends on the configuration and workload.
Local execution avoids cloud usage charges for those tasks; the laptop is not a cloud-free proposition. Microsoft says developers can use local models with GitHub Copilot to prototype agents, review code and test larger models. CNBC described the routing demonstration at the company’s San Francisco event.
Microsoft puts the system in a chassis less than 18mm thick and under 4.5 pounds. It says a new thermal architecture provides up to 2.5 times the thermal power capacity of its current Surface Laptops, a design intended to support sustained performance rather than only brief bursts.
The included charging cable attaches magnetically to the right-side USB-C port and releases easily. That port still handles video, data and other USB-C devices. Two additional USB-C ports sit alongside HDMI, USB-A, an SD card reader and a headphone jack. Microsoft also specifies support for up to three 4K external displays with compatible equipment.
Loading discussion...
Join the conversation
Explain what would make owning the hardware worthwhile for you.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.