Satya Nadella and Jensen Huang just bet the farm on your laptop doing AI’s heavy lifting instead of Microsoft’s cloud. The two CEOs took the stage in San Francisco on October 7 to unveil the Surface Laptop Ultra, a Windows machine built around Nvidia’s RTX Spark chip with up to 128GB of unified memory and 1 petaflop of local AI compute.

That’s enough horsepower, Nvidia says, to run 120-billion-parameter AI models entirely on-device, no cloud subscription in sight. Pre-orders opened October 7, with hardware shipping in mid-October pending final regulatory approvals, according to reports.

For Microsoft, the move is a wager that some of the work currently grinding through its costly Azure data centers can be shifted to high-powered Windows machines in businesses and homes — where, notably, the customers foot the hardware bills. For Nvidia, cracking the Windows PC market would let it chase one of the last big markets still dominated by longtime rivals Intel and AMD.

Microsoft’s invitation billed the event as “a conversation on how local AI will shape the next chapter of the PC.” It’s the company’s first major Windows-focused event in more than two years.

What RTX Spark actually is

RTX Spark isn’t a repackaged consumer GPU bolted onto an Arm chip. Nvidia shrank its Grace Blackwell data center superchip down to a 3nm consumer SoC — the same architecture running AI clusters in hyperscaler data centers, now fitting inside a laptop.

The chip pairs a 20-core Arm Grace CPU with a 6,144-core Blackwell GPU on a single die, connected by NVLink-C2C. Both share up to 128GB of LPDDR5X unified memory at roughly 300GB/s of bandwidth, so the GPU never has to copy model weights back and forth from separate VRAM. That shared memory pool is how a laptop can load a 120-billion-parameter model that would have needed a multi-GPU server rig two years ago.

Two configurations are available at launch, per Nvidia:

  • RTX Spark S2: 18-core CPU, 5,120-core GPU, 24–32GB unified memory — from about $2,999
  • RTX Spark S3: 20-core CPU, 6,144-core GPU, up to 128GB unified memory — up to roughly $7,000

The chips themselves were announced in June. Microsoft has spent the months since tuning Windows 11 specifically for the hardware, including a new Workload Profile Scheduling system that spreads tasks across the chip’s 20 CPU cores and a tuned Prism emulation layer for running x86 apps on Arm. Dell, Lenovo, and ASUS have already announced their own RTX Spark devices, though none have published benchmarks yet.

The awkward question: agents on your laptop

One of the thorniest challenges for Microsoft and Nvidia will be showing that AI agents — long-running AI systems that tackle complex problems on their own — can be safely contained on personal computers. Agents are exactly the kind of workload the Surface Laptop Ultra is built for, and they’re exactly the kind of software that makes security teams nervous.

The timing is pointed. Nvidia has been working to prevent a repeat of the hack of AI hub Hugging Face, and Apple is simultaneously tightening the process for granting AI agents full access to a Mac’s hard drive. Selling a machine whose whole pitch is “run powerful autonomous software locally” means Microsoft and Nvidia own the security story now, too.

Why this matters

This launch is the clearest signal yet of where the PC business is heading. For years, the pitch for AI features was “the cloud does the thinking, your device is a terminal.” The Surface Laptop Ultra flips it: your device does the thinking, and the cloud is optional.

Apple has been chasing the same idea with new Mac computers built around its own silicon. Intel and AMD, meanwhile, are watching the last market they truly dominate get squeezed from both sides. Microsoft’s edge here is distribution — Windows machines are already in businesses and homes everywhere, and developers working in CUDA, Nvidia’s entrenched parallel-computing platform, get to run it natively on a Surface for the first time.

If the reviews back up the numbers, the “AI PC” category finally gets a machine with actual teeth. If they don’t, it’s another expensive launch in a category that’s been long on promises and short on delivery.

Frequently asked questions

What is the Surface Laptop Ultra? It’s Microsoft’s new flagship laptop built with Nvidia, powered by the RTX Spark chip with up to 128GB of unified memory and 1 petaflop of local AI compute, designed to run 120-billion-parameter AI models without the cloud.

How much will the Surface Laptop Ultra cost? Microsoft hasn’t announced pricing yet. Nvidia’s RTX Spark platform comes in S2 (from about $2,999) and S3 (up to roughly $7,000) configurations.

When can you buy the Surface Laptop Ultra? Pre-orders for RTX Spark-based hardware opened October 7, 2026, with shipments expected in mid-October, pending final regulatory approvals.

Can the Surface Laptop Ultra run AI models locally? Yes. The top RTX Spark S3 configuration can run 120-billion-parameter models entirely on-device, with no cloud subscription needed, according to Nvidia.

Sources: Reuters, Notebookcheck, Tweaktown, byteiota, Windows Latest