Microsoft and Nvidia finally put a price on their AI laptop. The Surface Laptop Ultra, the first laptop built around Nvidia’s RTX Spark platform, opened for pre-orders on October 7 starting at $2,599, with shipping from October 16. Microsoft CEO Satya Nadella and Nvidia CEO Jensen Huang shared the stage at a San Francisco event this week to make the case that this is the machine that moves AI off the cloud and onto your desk.
It is a big swing. This is not another thin Copilot+ PC. It is a 15-inch workstation-class laptop with up to 128GB of unified memory, a full Nvidia Blackwell GPU, and native CUDA support on an Arm chip, aimed squarely at developers, creators, and anyone who wants to run serious AI models locally. And Microsoft is not being shy about the target: its launch materials take direct aim at the MacBook Pro.
What $2,599 actually gets you
The entry configuration pairs the reduced RTX Spark N1X part, an 18-core Nvidia Grace CPU with a 5,120-core Blackwell GPU, with 24GB of unified memory and a 512GB SSD. That is the $2,599 machine, and it is worth being clear about what it is: the headline 128GB memory figure belongs to the top configurations, not the base model.
The full-fat version steps up to a 20-core CPU, a 6,144-core Blackwell GPU, up to 128GB of LPDDR5X unified memory, and storage options up to 2TB. That configuration tops out at $5,899, which is serious workstation money, but it buys something no other Windows laptop currently offers: a single memory pool the CPU and GPU share, so a large AI model can use far more memory than the VRAM on a typical gaming laptop.
The spec sheet, condensed
| Base ($2,599) | Top config ($5,899) | |
|---|---|---|
| Chip | Nvidia RTX Spark N1X, 18-core CPU | Nvidia RTX Spark N1X, 20-core CPU |
| GPU | 5,120-core Blackwell | 6,144-core Blackwell |
| Memory | 24GB unified | Up to 128GB unified |
| Storage | 512GB SSD | Up to 2TB |
| AI compute | Up to 1 petaflop (FP4), runs 120B+ parameter models locally | |
| Display | 15-inch mini-LED touchscreen, 3270×2180, 120Hz, up to 2,000 nits peak HDR | |
| Size | Under 18mm thick, about 4.5 pounds | |
| Ports | 3x USB-C, USB-A, HDMI 2.1b, full-size SD reader, 3.5mm jack, new magnetic USB-C charging | |
| Wireless | Wi-Fi 7, Windows Hello face sign-in | |
| Battery | Up to 15 hours claimed (typical use) | |
A few details stand out. The magnetic USB-C charging port replaces the old Surface Connect, which long-time Surface owners will either love or mourn. The SSD is user-replaceable, a rarity in premium laptops. And Microsoft rebuilt parts of Windows 11 for the chip, including the task scheduler and a new power and thermal framework developed with Nvidia, to keep 20 cores fed without the machine running hot.
Nvidia’s MacBook Pro claims, with a grain of salt
The launch came with the obligatory benchmark chest-thumping. Based on Nvidia’s own preliminary testing, RTX Spark systems deliver up to 2.1x faster time-to-first-token, 4.3x faster AI image generation, and 6.2x faster AI video generation than a MacBook Pro with an M5 Pro chip.
Treat those numbers the way you would any vendor-supplied figures: as a directional claim, not a review. They are platform-level measurements, not independent tests of the Surface Laptop Ultra itself, and Apple will have its own numbers. The real test will be third-party benchmarks once retail units ship on October 16. Still, the gap Nvidia is claiming is large enough that even half of it would be meaningful for local model work.
How it undercuts HP
Microsoft is not the only company selling RTX Spark laptops, and the pricing looks deliberate. HP’s OmniBook X 14 starts at $2,999 and its 16-inch variant at $3,199, with higher configurations reaching $4,499 and about $4,999. The Surface Laptop Ultra starts $400 below HP’s entry point, which positions Microsoft’s machine as the value play in a category where “value” still means two and a half grand.
The catches: drivers, ARM, and $1,600 RAM jumps
Early reports urge some caution. Pre-release units reportedly shipped with immature drivers, and retail stability is unconfirmed until reviewers get final hardware. Memory upgrades are brutally priced: jumping from 64GB to 128GB adds about $1,600. And this is Windows on Arm, which still depends on emulation for older apps, so check your critical software before ordering. Microsoft demonstrated gaming with 60 fps in Gears of War: E-Day, but Call of Duty support for RTX Spark is only promised for 2027.
The machine is also not small. At 18mm thick and around 4.5 pounds, with peak system power between 90 and 140 watts under load, this is a portable workstation, not an ultrabook. The 15-hour battery claim is for typical use; hammer the GPU and that number will fall fast.
What this means: the AI PC grows up
The bigger story is the strategy. Microsoft is betting that a meaningful share of AI work, coding agents, local model experiments, complex business projects, can move off its expensive Azure data centers and onto machines where the customer pays for the hardware. Nvidia gets a crack at the Windows PC market, one of the last strongholds of Intel and AMD. And both companies get to tell a story Apple has been telling for years: the best AI computer is the one already on your desk.
That story cuts both ways. Apple is tightening macOS around AI agents at the same time Microsoft is opening Windows up to them, and the two companies are converging on the same idea from opposite directions. If you are weighing cloud subscriptions against local hardware, our October AI model price list and the new ChatGPT Pro tiers are the math to do first: $2,599 buys a lot of tokens before a local RTX Spark machine pays for itself.
