Nvidia says the RTX Spark chip “delivers amazing creating, AI development, and gaming” in slim laptops. Microsoft says the Surface Laptop Ultra was built for “world makers and creative pros who refuse to choose between power and portability.” Both of those statements are about a machine that starts at $2,599.99 and climbs to $3,699.99 — which means the one choice nobody is getting out of is the one involving a credit card.
That’s the tension at the center of this launch, and it’s worth sitting with before the preorder page does its job on you. Microsoft has positioned the Surface Laptop Ultra as a direct run at the MacBook Pro, with the AI era as the framing. The hardware story is Nvidia’s: RTX Spark silicon, two configurations, a 5120-core GPU paired with an 18-core CPU at the low end and a 6144-core GPU with a 20-core CPU at the top. Platinum or Nightfall. Preorders are open on Microsoft’s site.
And that is, more or less, what we actually know.
What Microsoft Told Us Versus What Reviewers Need
I review AI tools for a living, which mostly means watching companies describe performance instead of demonstrating it. This launch fits the pattern. The specs we have are core counts, which are the least useful numbers in the entire conversation. Core counts tell you how wide the chip is. They don’t tell you:
- How much unified or dedicated memory you get, which is the actual ceiling on what models you can run locally
- Memory bandwidth, which decides whether local inference is usable or just technically possible
- Sustained thermal behavior in a thin chassis under a long training or fine-tuning run
- Battery life when the GPU is doing anything at all
- How the Windows software stack handles Nvidia’s tooling day to day
For a machine pitched partly at AI developers, those five unknowns are the entire review. A 6144-core GPU that throttles after twelve minutes is a very expensive space heater. A 5120-core GPU with generous fast memory and a quiet fan curve might be the better buy. Right now nobody outside Microsoft and Nvidia can tell you which situation you’re in, and the preorder button doesn’t care.
The MacBook Pro Comparison Cuts Both Ways
Microsoft picked that fight deliberately, and it’s a reasonable fight to pick. Apple has spent several years owning the “powerful laptop that doesn’t sound like a leaf blower” category, and a lot of AI work has quietly migrated to Macs for exactly one reason: large pools of fast memory attached to a competent GPU, in a body you can carry.
The counter-argument for the Surface Laptop Ultra is obvious. Nvidia hardware means CUDA, and CUDA still means the shortest path from a GitHub repo to something running on your own machine. Anyone who has spent an afternoon coaxing a project into working on Apple silicon knows why that matters. If RTX Spark delivers real throughput in this form factor, Microsoft has a genuine argument that didn’t exist a year ago.
The problem is that “if” is carrying a $2,599.99 minimum. Apple’s pricing is also high, but Apple’s thermal and battery behavior in thin laptops is a known quantity backed by years of independent testing. Microsoft is asking buyers to accept a new chip, in a new chassis, on faith, at a price that leaves no room for disappointment.
Who Should Actually Preorder
Nobody, yet. Not because the machine is bad — I have no idea whether it’s bad — but because there is no version of this purchase that gets worse by waiting for benchmarks. Preordering a laptop is paying full price for the privilege of being an unpaid tester. The colorways will still be there in a month.
If you’re deciding between the two configurations, the $1,100 gap between them deserves more scrutiny than the core counts suggest. An 1100-dollar jump for roughly 20% more GPU cores and two more CPU cores is a bad trade on paper. It only becomes a good trade if the higher tier also brings more memory or better sustained performance, and Microsoft hasn’t given us enough to judge that. Wait for someone to run both side by side.
My Read
This is Microsoft doing the right strategic thing with incomplete public evidence. Putting Nvidia silicon in a premium portable is a sensible answer to where AI work is actually happening, and the Surface line was the obvious place to try it. The pitch is coherent in a way that a lot of AI hardware announcements are not.
But a coherent pitch and a working machine are different products, and only one of them is available for preorder. Treat the launch as a statement of intent. Treat the price as a reason to stay patient. When independent numbers land, I’ll tell you whether the Spark is worth the spend.
🕒 Published: