What happens when Microsoft stops treating the Surface as a thin-and-light office laptop and drops an NVIDIA RTX graphics chip inside it? You get the Surface Laptop Ultra, the most expensive and most ambitious Surface laptop yet. It starts at $2,599, pre-orders are open now, and Microsoft and its partners expect units to start shipping on October 16.
Unveiled at Microsoft’s October 7 Windows and Surface event in San Francisco, the 15-inch machine runs on NVIDIA’s new RTX Spark platform with up to 128GB of unified memory. So is it a gaming rig, an AI workstation, or a MacBook Pro rival? Let’s break it down.
| Key fact | Details |
|---|---|
| Starting price | $2,599.99 (24GB RAM, 512GB SSD) |
| Top configuration | $5,899.99 (128GB RAM, 1TB SSD) |
| Pre-orders | Open since October 7, 2026 |
| Expected shipping | October 16, 2026 |
| Chip platform | NVIDIA RTX Spark (Grace Arm CPU + Blackwell RTX GPU) |
| Display | 15-inch mini-LED touchscreen, 120Hz, up to 2,000 nits HDR |
| Charging | Magnetic breakaway USB-C port, 140W adapter in the box |
| Colors | Platinum and Nightfall |
| Sibling product | Surface RTX Spark Dev Box, $5,999 |
What the Surface Laptop Ultra actually is
Here’s the thing. This isn’t a refreshed Surface Laptop with a faster processor. It’s built around a single NVIDIA chip, which Microsoft calls the RTX Spark Superchip, that pairs an up-to-20-core NVIDIA Grace CPU with Blackwell RTX graphics and a shared pool of memory. According to Microsoft’s official Surface Laptop Ultra page, that memory is unified, meaning the CPU and GPU draw from the same pool rather than the GPU having its own separate video memory.
Why should you care? Unified memory is the same idea Apple has used in its M-series Macs for years, and it’s a big deal for AI work. A large language model has to fit in memory to run quickly, and a laptop GPU with 8GB or 12GB of dedicated VRAM simply runs out of room. With up to 128GB shared, Microsoft says the machine can handle local agent workflows with models of up to 120 billion parameters.
One detail matters a lot: the CPU side is Arm-based, not Intel or AMD. Windows Central notes it belongs to the same architecture family as Qualcomm’s Snapdragon X chips, so apps that aren’t built natively for Arm run through Microsoft’s Prism emulation layer.
The key takeaway here is simple: the Surface Laptop Ultra is a pro workstation first and a Surface second.
Surface Laptop Ultra price and every configuration
Microsoft’s store currently lists 11 pre-order configurations across two chip tiers. The entry tier uses an 18-core CPU with a 5,120-core GPU. The upper tier steps up to a 20-core CPU and a 6,144-core GPU, which is the only way to get 48GB, 64GB or 128GB of memory. Every model gets the same 15-inch mini-LED screen, ports and design.
| CPU / GPU cores | Memory | Storage | Windows | Price |
|---|---|---|---|---|
| 18 / 5,120 | 24GB | 512GB | 11 Home | $2,599.99 |
| 18 / 5,120 | 24GB | 1TB | 11 Home | $2,999.99 |
| 18 / 5,120 | 32GB | 1TB | 11 Pro | $3,299.99 |
| 20 / 6,144 | 32GB | 1TB | 11 Pro | $3,699.99 |
| 20 / 6,144 | 48GB | 1TB | 11 Pro | $3,999.99 |
| 20 / 6,144 | 64GB | 1TB | 11 Pro | $4,299.99 |
| 20 / 6,144 | 64GB | 2TB | 11 Pro | $4,699.99 |
| 20 / 6,144 | 128GB | 1TB | 11 Pro | $5,899.99 |
Notice the jump. Going from the base 24GB model to the cheapest 20-core model costs about $1,100, and that’s the gap readers argued about most in Windows Central’s roundup of community reactions. That said, plenty of people expected a $2,999 starting price, so $2,599 landed better than feared.
There’s also a sweetener. Microsoft is running a trade-in offer worth up to $1,000 on qualifying purchases from October 7 to November 23, 2026, through the Microsoft Store online in the US and Canada.
Prices this high aren’t happening in a vacuum, either. Windows Central’s Cale Hunt points partly to the ongoing RAM and storage shortage, the same pressure we covered in why your next iPad and Mac will be 20% more expensive.
Specs, display and that magnetic USB-C port
Beyond the chip, the hardware is a clear step up from any Surface before it. Microsoft calls the screen its brightest Surface display ever, hitting 2,000 nits peak HDR brightness in a 10% window, with 100% DCI-P3 color, Dolby Vision IQ and a 120Hz refresh rate. It’s a touchscreen, but pen input isn’t supported.
| Spec | Surface Laptop Ultra |
|---|---|
| Display | 15-inch mini-LED touch, 120Hz, 2,000 nits peak HDR, 100% DCI-P3 |
| CPU | NVIDIA Grace (Arm), 18 or 20 cores |
| GPU | NVIDIA Blackwell RTX, 5,120 or 6,144 cores |
| Memory | 24GB to 128GB LPDDR5x unified |
| Storage | 512GB to 2TB, removable SSD |
| Ports | 3x USB-C, 1x USB-A, HDMI, SD card reader, headphone jack |
| Weight and thickness | Under 4.5 lbs, under 18mm |
| Security | Secured-core PC, Windows Hello face and fingerprint |
The headline design feature is what Microsoft calls the world’s first built-in magnetic USB-C charging port, billed as Magnetic Connect. Trip over the cable and it snaps away instead of yanking your laptop off the desk. Unlike Apple’s MagSafe, which only carries power, Windows Central points out that this breakaway port works with any USB-C cable and handles both power and data.
Cooling got serious attention too. Microsoft says this is Surface’s largest thermal system ever, with up to 2.5 times the thermal capacity of the 15-inch Surface Laptop (8th Edition) and bigger fans that spin slower to keep noise down. According to Windows Report, Microsoft also claims the laptop keeps more than 99% of its performance on battery power, which matters if you edit video or run models away from a wall socket.
Bottom line: this is the first Surface built to run hard for long stretches, not just in short bursts.
Can the Surface Laptop Ultra handle gaming?
Short answer: yes, surprisingly well. Long answer: it isn’t the laptop you should buy if gaming is your main goal.
Microsoft’s page confirms support for ray tracing, NVIDIA DLSS and NVIDIA Reflex. In closed demos, Windows Central reported smooth gameplay in Indiana Jones and the Great Circle, Fortnite and Alan Wake 2, and Tom’s Guide saw a Gears of War: E-Day demo running at a smooth 60fps. If you’re excited about that one, here’s everything we know about Gears of War: E-Day.
Here’s the catch. Many of those games ran through Prism emulation because the chip is Arm-based. Emulation works better than it used to, but anti-cheat systems in competitive multiplayer games have historically been the weak spot on Windows on Arm, and you’ll pay $2,599 or more for a GPU that a dedicated x86 gaming laptop can match for less.
“The Surface Laptop Ultra should be treated as a pro-grade laptop that happens to have some more-than-decent gaming chops.” That’s how Cale Hunt of Windows Central put it, and it’s a fair summary.
So, gaming is a bonus here, not the reason to buy.
What RTX Spark means for local AI
Let’s back up a little. Most AI tools you use today, from ChatGPT to Copilot, run in a data center. Your prompt goes to the cloud, gets processed, and comes back. That works, but it costs money per token, needs a connection, and means your data leaves your machine.
RTX Spark flips that. Microsoft says the platform delivers up to 1 petaflop of AI compute (a theoretical FP4 figure from NVIDIA) and supports CUDA, the software layer most AI developers already build on. In practice, that means tools like ComfyUI and Blender, both demoed at the event, can run locally on the GPU, and developers can test large open models without renting cloud GPUs. At the same event, NVIDIA showed DeepSeek V5 Flash running locally in around 60GB of memory, which only fits on the higher-memory configurations.
Who benefits most? Developers building agents, companies with proprietary code or customer data that can’t leave the device, and creators doing 3D or video work. If you’re still deciding whether any of this applies to you, our explainer on why you’ll need an AI PC this year is a good starting point.
“Memory is the new horsepower. Buy the RAM for the models you’ll run next year, not the ones you run today.”
Need even more? The Surface RTX Spark Dev Box is a compact desktop with 128GB of unified memory and up to 1 petaflop of AI compute for $5,999. It’s aimed at developers who want a dedicated box on the desk rather than a laptop, and Microsoft says it’s sold in the US exclusively on Microsoft.com.
Surface Laptop Ultra vs MacBook Pro
It’s impossible to look at this laptop and not think of Apple. Microsoft even made the comparison itself, claiming 2.1 times faster token generation than the MacBook Pro, though no test details were shared. Here’s how it lines up against the current 16-inch MacBook Pro, based on Apple’s published specs.
| Feature | Surface Laptop Ultra | 16-inch MacBook Pro |
|---|---|---|
| Chip | NVIDIA RTX Spark (Arm) | Apple M5 Pro or M5 Max |
| Max unified memory | 128GB | 128GB (M5 Max, 40-core GPU) |
| Peak HDR brightness | 2,000 nits | 1,600 nits |
| Touchscreen | Yes | No |
| Weight | Under 4.5 lbs | 4.7 lbs |
| Battery claim | All-day (no hours published) | Up to 24 hours video streaming |
| CUDA support | Yes | No |
| Magnetic charging | USB-C breakaway (power + data) | MagSafe (power only) |
The choice mostly comes down to software. If your AI or 3D workflow depends on CUDA, the Surface has an edge no Mac can match. If you live in macOS apps and want proven, very long battery life, the MacBook Pro remains the safer bet. And if you’re weighing Apple’s next moves, our Apple October event 2026 coverage is worth a look before you commit.
The takeaway: the Surface finally gives Windows users a credible MacBook Pro alternative, but battery life is the big unknown until reviews land.
Who should buy the Surface Laptop Ultra (and who should skip it)
At these prices, buying on hype is a costly mistake. Here’s a quick way to decide.
1. Buy it if you run AI models or CUDA tools locally
This is the core audience. If you fine-tune models, build agents, or run Stable Diffusion-style workflows, the 64GB or 128GB configurations give you room a typical laptop GPU can’t. For most AI developers, the 20-core, 64GB model at $4,299.99 is the sensible middle ground.
2. Pick the 32GB mid-tier for creative pro work
Windows Central recommends the 20-core, 32GB model at $3,699.99 for advanced users. Video editors, 3D artists and photographers get the bigger GPU without paying for memory they may never fill.
3. Skip it if you mainly want to game
A dedicated gaming laptop gives you x86 compatibility, fewer anti-cheat headaches and often more GPU for the money. The Surface can play games, but you’d be paying a premium for AI hardware you won’t use.
4. Wait for reviews if battery life is a dealbreaker
Microsoft promises all-day battery life but hasn’t published hour figures. With shipping starting October 16, independent battery tests should follow quickly, and pre-orders aren’t charged until shipping anyway.
What to do next
If the Surface Laptop Ultra fits your work, the smart move is to pre-order the configuration you actually need now, since your card isn’t charged until it ships, and check whether you have an old laptop worth trading in before the November 23 deadline. Then keep an eye on the first full reviews around launch day for real battery numbers and app compatibility on Arm.
Fast forward a year, and machines like this could make running serious AI on your own laptop feel normal. For now, the Surface Laptop Ultra is the boldest Windows laptop Microsoft has ever made, and it’s well worth a close look if you’re a creator or developer ready to bring AI home.
Frequently asked questions
The base model with 24GB of memory and a 512GB SSD costs $2,599.99. The top configuration with 128GB of memory and a 1TB SSD costs $5,899.99, with several options in between.
Pre-orders opened on October 7, 2026, and shipping is expected to begin on October 16, 2026, from Microsoft and partners such as Best Buy.
It can run demanding games like Indiana Jones and the Great Circle with ray tracing and DLSS, but many titles run through emulation on its Arm chip, so a dedicated gaming laptop is a better buy for gamers.
Microsoft promises all-day battery life but hasn’t published hour figures yet. It does claim the laptop keeps more than 99% of its performance on battery power.
It’s aimed at developers, AI specialists, engineers, scientists and creators who want to run large AI models and GPU-heavy apps locally instead of in the cloud.
Choose the Surface if you need CUDA, a touchscreen or Windows. Choose the MacBook Pro if you rely on macOS apps and want Apple’s proven battery life of up to 24 hours of video streaming.
