In brief
- RTX Spark laptops combine an Arm-based processor, a Blackwell GPU, and unified memory.
- Hermes, Perplexity, and OpenClaw will be the first three AI agents supported by Windows at launch.
- Microsoft’s Prism emulator will be used for apps that don’t have a native Arm version.
- Copilot+ PC features, including Recall, won’t be available on RTX Spark devices at launch.
In a report published on 15 October 2026, The Verge covered Microsoft and NVIDIA’s preparations on the Windows side for RTX Spark laptops. Following a launch event in San Francisco last week, the work focuses on areas such as installing local AI agents, Arm app compatibility, and unified memory management. The RTX Spark platform combines an Arm-based processor, a Blackwell GPU, and unified memory.
How will local AI agents be installed?
The planned “Get Started” experience in Windows will help users find and install an agent. Agents are expected to be installed and updated through the Windows Store. The first options supported at RTX Spark’s launch will be Hermes, Perplexity, and OpenClaw; Microsoft is reportedly working to add other agents as well.
In The Verge’s test, getting Hermes ready to use went largely smoothly, while setting up OpenClaw’s local gateway caused problems and required the app to be uninstalled and reinstalled. The example suggests that although the aim is to simplify setup, the experience may vary by agent. According to the report, large local models can take longer to generate responses than cloud models, making long-running background tasks one possible use case.
Arm apps and unified memory
Microsoft’s Prism emulator will be used to run apps that don’t have a native Arm version. The report says Microsoft is optimizing Prism for the RTX Spark microarchitecture to adapt x64 apps to the Arm-based system. It also notes that compatibility improvements for Prism are ongoing.
On memory management, Microsoft’s work aims to balance power consumption, performance, and system responsiveness. A slider in NVIDIA’s Windows app will let users manually adjust the portion of unified memory allocated to the GPU. According to the report, this setting can help balance the needs of different apps and games, particularly on devices with 24 GB or 32 GB of memory. Microsoft’s Surface Laptop Ultra, meanwhile, can be configured with up to 128 GB of unified memory; the report does not describe this capacity as a feature common to all RTX Spark laptops.
Gaming and the Windows feature timeline
Microsoft and NVIDIA are also working on game compatibility and performance. In The Verge’s tests, many games ran through emulation, with results varying from one game to another. The examples in the report show some games reaching 60 fps at 1440p, with medium settings and DLSS set to performance; it also emphasizes that this is not on par with the performance of dedicated gaming laptops. Having the Xbox app prepare game shaders in advance while downloads are in progress is also among the planned Windows features.
Copilot+ PC features, including Recall, won’t be available on RTX Spark devices at launch. Microsoft reportedly expects to first make these experiences available to Windows Insiders later in the year. It’s therefore important to distinguish the platform’s day-one software features from those expected to be added later.
Impact on architecture, interior design, and visualization workflows
RTX Spark emphasizes running local AI agents and GPU-powered workloads on a laptop. However, the source contains no tests of compatibility or performance for architecture and visualization software. Offices should therefore verify before purchasing whether the apps they use have native Arm versions and how they perform through Prism.
The ability to change how much unified memory is allocated to the GPU allows users to adjust settings for different workloads, but the source provides no measurements of the results in professional 3D software. Teams considering local AI should also assess agent support and the installation process alongside how other apps sharing the same memory perform. Suitability for architectural workflows can be established through testing with the software and real projects they use.
Sources
1 sourceSource texts are not republished; short quotes are marked, everything else is our own summary and commentary.
What makes RTX Spark noteworthy for architecture and visualization teams is its aim to bring local AI agents to laptops alongside a GPU and unified memory. However, the source includes no compatibility or performance tests for architectural software, so there isn’t enough data to consider the platform a direct workstation alternative.
Offices in Turkey should assess their apps’ Arm support, how they run through emulation, and the memory configuration together when making a purchasing decision. They should also bear in mind that large local models may respond more slowly than cloud models. The source doesn’t include pricing or licensing details, so it’s important to clarify those separately.
Frequently asked questions
Which AI agents will be supported on NVIDIA RTX Spark laptops?
The first three agents supported at launch will be Hermes, Perplexity, and OpenClaw. They’re planned to be installed and updated through the Windows Store.
Will x64 apps run on RTX Spark laptops?
Apps without a native Arm version will be able to run through Microsoft’s Prism emulator. Compatibility depends on the app and the optimizations made for the platform.
Will Copilot+ PC features be available on RTX Spark at launch?
No. Copilot+ PC features, including Recall, won’t be ready at launch. Some experiences are expected to first become available to Windows Insiders later in the year.
