Microsoft’s Surface Laptop Ultra with NVIDIA’s RTX Spark chip and 128GB unified memory is already out of stock at $5,899.99 in the US — just days after launch. The configuration is listed as unavailable on Microsoft’s official product page, though lower-tier models remain purchasable. No restock date has been announced, and Microsoft has not confirmed whether the shortage reflects constrained supply, early demand spikes, or both.
The price applies to the US market only. No Indian pricing or availability details have been disclosed in the reporting. A UK or EU price is not mentioned. The device is not yet confirmed for India.
The Surface Laptop Ultra is the first laptop to ship with NVIDIA’s RTX Spark platform, a new class of mobile GPU built around massive unified memory pools and full CUDA support. Unlike traditional discrete GPUs, RTX Spark integrates up to 128GB of shared system memory, letting developers run large language models locally without cloud dependency.
NVIDIA claims it can handle 120B-parameter models with million-token context windows. A key draw for enterprises and researchers managing sensitive data. That capability explains why the top-spec model vanished first: it’s the only one that ships with the full 128GB UMA, 1TB SSD, and RTX Spark silicon. But raw memory size isn’t everything.
The Surface Laptop Ultra’s unified memory bandwidth sits at 273GB/s, less than half the 614GB/s offered by Apple’s M5 Max MacBook Pro at the same 128GB capacity. And while the RTX 5090 laptop GPU beats the M5 Max in prompt processing and token generation (up to +133% in LLM benchmarks), it’s capped at 24GB of GDDR7 VRAM.
Too little for the densest inference workloads. RTX Spark bridges that gap, but only where CUDA compatibility matters most. Which it does, deeply. Most AI frameworks, creative suites, and inference engines are built and optimized for CUDA first.
AMD’s Ryzen AI MAX 400 chips may offer up to 192GB of memory, but lack that software moat. For professionals, rewriting toolchains or hunting for CUDA alternatives costs more time than $5,899.99 saves. So yes: this is a niche machine. It’s not for general productivity or casual creators.
It’s for developers running local Llama 3.5-120B fine-tunes, ML engineers validating private-model pipelines, or studios rendering real-time neural graphics. All without sending data off-device. The Surface Laptop Ultra doesn’t compete with the MacBook Pro on bandwidth or battery life. It competes on developer inertia, CUDA lock-in, and the ability to keep 120B models entirely on your desk.
Microsoft hasn’t said when the $5,899.99 model will return. Wccftech has asked. We’ll update when they do.
CDFiled by The Computing Desk
Laptops, desktops, processors, graphics cards and the operating systems that run on them.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.