First day cover ·
Ex-Apple Engineer Eyes Windows 11 Laptop Over NVIDIA CUDA
Ray Fernando, a 12-year Apple veteran, is testing the Surface Laptop Ultra for a week and says CUDA on a portable Windows machine is the draw.
When someone who spent more than a decade building Apple products starts seriously weighing a Windows laptop, it's worth paying attention. Ray Fernando says the pull is CUDA, and the idea that a machine light enough to carry can now run it. He's living with a Surface Laptop Ultra for a week before deciding whether to switch, and his early verdict is cautiously positive.
Why CUDA is the hook
In his X post, Fernando wrote that CUDA is "the thing most AI tools run on," and said the Surface Laptop Ultra puts that power "in a bag I can carry." That's the core of his case. CUDA is NVIDIA's platform for general computing on its GPUs, with libraries, debugging tools and support for frameworks such as PyTorch. Not every AI tool needs it, so "most" is Fernando's shorthand for how much AI development leans on NVIDIA's software stack. Microsoft is betting on the same dependency, promising CUDA-accelerated PyTorch on RTX Spark PCs.
Macs have been on the outside of that stack for a while. NVIDIA made CUDA 10.2 the last version supporting macOS back in 2019, and Apple's answer is its own Metal and MLX frameworks. What makes Fernando's position interesting is that he isn't a casual Mac user looking for an excuse. His Hugging Face profile lists him in the MLX community, so he knows Apple's side of the fence well. That's why a Windows laptop has to offer something specific to get his attention.
What he's testing so far
The unit he's reviewing is the top-end Nightfall configuration. In his unboxing video, he lists a 20-core CPU, a 6,144-core GPU, 128GB of unified memory, 1TB of storage and Windows 11 Pro. Microsoft sells that setup for $5,899.99. Fernando says it should handle AI models with more than 120 billion parameters. That's his expectation based on the hardware, and he hasn't published benchmark numbers in the material available so far.
He praised the keyboard, the large trackpad and the build quality. The detail that impressed him most was Magnetic Connect. He lifted the laptop by its charging cable, then pulled the cable free with a quick yank, a demo that shows how the magnetic connection behaves under load. On weight, he puts the Surface Laptop Ultra at 4.41 pounds, about 0.3 pounds lighter than another 16-inch laptop he owns with 128GB of RAM. Anyone who has carried a 16-inch workstation will notice that gap.
His verdict so far is "On paper it's a yes." He's still waiting on the software side, and the final call is still a few days out. His YouTube description says the video isn't sponsored and the opinions are his own. Taking him at his word, the open questions are everyday software behaviour and sustained performance, not hardware specs.
The OpenGL comparison
Former Windows president Steven Sinofsky, who ran Windows through Windows 7 and Windows 8, quoted Fernando's post and said it shows "the Spark strategy is playing out well." He then drew a parallel to a fight he watched from inside Microsoft:
"It's a replay of the OpenGL war SGI started. Nvidia embraced. Microsoft took on w/D3D. Quake rebuffed. Apple warmed up then split; went its own way w/Metal."
The history backs up the analogy. SGI introduced OpenGL in 1992 as a cross-platform graphics API, and Microsoft answered with Direct3D on Windows. id Software's Quake went with OpenGL, and Apple supported OpenGL for years before deprecating it in 2018 in favour of Metal. Brad Silverberg, who led Windows 95 at Microsoft, added his own memory of the period: "I remember meeting with Jensen in the early days to get him to support DirectX. He did and it worked out great for both him and msft," he wrote, noting that OpenGL had plenty of fans inside Microsoft as well.
My read is that the lesson isn't about graphics APIs specifically. Developers stick with the tools they already know, and hardware wins when it ships with those tools working on day one. Thirty years ago NVIDIA backed Microsoft's API. Today it's Windows that needs NVIDIA's CUDA to pull developers like Fernando across.
Bigger than one Surface
The Surface Laptop Ultra isn't the only RTX Spark machine in play. Microsoft named ASUS, Dell, HP, Lenovo, MSI and Surface, and NVIDIA's list also includes Acer and Gigabyte. A Windows user in the discussion pointed out that Microsoft earns Windows licence revenue on every one of these laptops, which is a fair way to see the commercial side of the push.
There's a catch, though. Every RTX Spark PC runs Windows on Arm, so app compatibility matters as much as CUDA does. Fernando hasn't shown that side yet. He also didn't mention Copilot once in his video, and Microsoft dropped the Copilot+ PC branding for this laptop. The pitch here seems to be about the hardware and the developer stack, not the Windows AI experience.
If you're on a Mac and doing AI work, the question isn't whether a Windows laptop can match Apple's build quality. It's whether CUDA's reach is worth the switching cost. Fernando's week with Windows 11 should show whether it outweighs years of macOS muscle memory, and his software impressions will matter more to buyers than the spec sheet.
