Microsoft's Surface Laptop Ultra Ships With 128GB Memory and Mystery USB-C Port

Two RTX Spark devices, one design philosophy

Microsoft announced two new Surface devices arriving later this year, both powered by Nvidia's RTX Spark chips. The Surface Laptop Ultra is a 15-inch clamshell with a 2,000-nit mini-LED display (company-reported as the brightest Microsoft has shipped on a Surface device), a larger trackpad with haptic feedback, 128GB unified memory, and a thermal design budget of up to 80 watts. The Surface RTX Spark Dev Box is a desktop form factor with the same chip, 128GB memory, and a 100-watt thermal envelope, housed in an aluminum 3D-printed chassis with 1,000 air vents.

Microsoft prioritized performance, battery life, and display quality over weight or thinness. Andrew Hill, corporate vice president of Surface product, told The Verge: "When we went through the priority order of what we're going to design for, performance, performance, performance, battery life, battery life, battery life, display, display, display."

The Surface Laptop Ultra includes six ports (two USB-C left, one wider USB-C right, one USB-A, one HDMI, one SD Card reader). The wider USB-C port remains unexplained; Microsoft declined to discuss it during hands-on sessions but signaled additional details would arrive later. The device drops Microsoft's traditional Surface Connect magnetic charging port entirely.

Inside, the Surface Laptop Ultra shows internal redesign for repairability, with components clearly marked and arranged in a grid layout. The device uses two fans for cooling and maintained warm-to-touch (not hot) thermal output even under near-full load during testing.

Local AI compute and the cloud metering question

Microsoft is explicitly marketing both devices to developers and creators who want to run inference workloads on owned hardware without paying per-token to cloud providers. Hill noted: "You'll be able to do a lot of work locally on a thing you own, and if you want to let it rip, cool, you're not on a meter."

The company is notably downplaying Copilot Plus PC branding for these devices, despite both qualifying for Copilot Plus features. The shift signals a pragmatic focus on professionals evaluating cost-per-inference rather than consumers buying AI assistants as a feature bundle.

The thermal envelope difference (80W laptop vs. 100W dev box) will directly affect sustained inference performance. The dev box's higher power budget means better sustained Tensor core utilization, relevant for production AI workloads and DLSS-based gaming.

Pricing remains unannounced, but Microsoft's recent memory-cost pressure (from what the industry terms "RAMageddon") suggests both devices will occupy premium market segments. Without stated pricing or independent benchmarks, buyers cannot yet compare total cost of ownership against cloud alternatives.

What to track before purchasing

Verify actual power consumption and thermals under your specific inference workload before committing. A 100W dev box may sustain performance where an 80W laptop throttles on continuous batch inference. Request thermal and latency benchmarks from Microsoft for your use case rather than relying on general performance claims.

Confirm repairability scoring from iFixit when units ship (Microsoft achieved 8/10 on the Surface Laptop 7 in 2024). Compare the final pricing against used or refurbished datacenter GPUs for the same VRAM tier; the ownership cost gap may narrow faster than expected.

Test the trackpad haptics on the Laptop Ultra in person if you spend eight hours a day dragging and scaling UI elements. It's a real quality-of-life feature, not a spec sheet claim, but only matters if your workflow is actually UI-heavy.