Preview
Microsoft opened preorders Oct. 7 for Surface Laptop Ultra, with availability beginning Oct. 16, giving enterprise technology teams a near-term opportunity to evaluate its new local AI hardware. The launch arrives before several related Windows AI experiences, making software readiness a separate procurement question.
Two devices, different deployment dates
Microsoft’s Surface announcement lists the laptop starting at $2,599 and the Surface RTX Spark Dev Box at $5,999. The desktop device is available for preorder exclusively through Microsoft’s U.S. store and is scheduled to begin shipping in November.
Both use NVIDIA’s RTX Spark platform. The laptop offers up to 128 gigabytes of unified memory, which is shared between the processor and graphics hardware. Microsoft cautions that the amount available to the graphics processor is less than total system memory and depends on the configuration and workload.
The Dev Box comes with development tools including Visual Studio Code, Git, Python and Node. Those details establish the intended technical audience; they do not establish how either machine will perform in a particular company’s development environment.
The software calendar is not the hardware calendar
In its separate Windows announcement, Microsoft said Microsoft Execution Containers became generally available on Windows 11. The technology lets organizations define the files and networks an agent may access and enforce those policies while it runs.
Other pieces are still ahead. GitHub HydraFusion’s local-model routing is scheduled for experimental preview later in October in the Copilot app, Copilot CLI and Visual Studio Code. Microsoft expects Copilot features using hybrid intelligence to start reaching Copilot+ PCs in the coming months, with availability varying by device, market and silicon platform.
The practical distinction is important: an Oct. 16 hardware purchase should not be treated as confirmation that every announced agent experience will be available that day. Experimental software also needs its own acceptance criteria.
What an enterprise pilot should answer
For buyers, the useful evaluation is narrower than a general promise of more AI capability. A pilot can compare an existing development task with the same task on the new hardware, recording completion quality, elapsed time and any cloud services still required. This is an evaluation framework, not a finding that local execution will be cheaper or faster.
IT teams can separately test whether an agent’s permitted files and network destinations match the assigned task, and whether reviewers can identify and reproduce its actions. Compute capacity and operational control answer different questions.
The launch therefore offers a confirmed equipment timetable, not a completed enterprise deployment case. The next milestone for technology leaders is evidence from their own workloads, alongside confirmation of which software capabilities have actually reached the devices they intend to manage.