What happened
Microsoft revealed its new line of AI-focused PCs at an event in San Francisco. The hardware includes the Surface Laptop Ultra, starting at $2,600, and the Surface RTX Spark Dev Box, priced from $6,000. These machines, as reported by TechCrunch AI, are powered by Nvidia's RTX Spark chip and are designed to run AI models locally.
Dell also announced its own RTX Spark PC, the XPS 16 Creator Edition. Alongside the hardware, Microsoft is shipping a revamped Windows 11 with a new feature called "Execution Containers" for sandboxing AI agents.
How the room's reading it
The consensus among developers is that this is Microsoft's most aggressive play yet for the AI practitioner market — long dominated by Apple's MacBooks. The trade-in offer targeting MacBook Pro owners is a clear signal of intent. Infra teams on X are focused on the new "Execution Containers" in Windows 11, seeing it as a serious platform-level bet on local agent development and orchestration.
While the price point puts these devices out of reach for most consumers, the immediate sell-out of the highest-end model suggests a pent-up demand from professional builders. The move is seen less as a hardware launch and more as a strategic effort to make Windows the default operating system for building and running on-device AI.
Sailfish's take
We see the hardware as a sideshow. The real story for builders is Windows 11's new "Execution Containers" — it's the first serious OS-level primitive for running sandboxed AI agents locally. This finally makes building complex, stateful agents on Windows a practical reality, not just a cloud-tethered experiment. We've shipped enough cloud-based AI to know the performance and cost ceilings are real.
We'd start prototyping applications that need deep, private access to the local file system and peripherals. But we'd hold off on buying the new Surface. The important work right now is exploring what this new software layer unlocks on existing machines. The question isn't if the hardware is fast enough; it's whether a new app ecosystem can be built on local agents.