New RTX Spark laptops and desktops and a DGX Station for Windows aim to run large AI models locally instead of in the cloud
Microsoft and Nvidia are taking the accelerated computing they built for Azure and putting it inside Windows PCs. The goal is to run serious AI models and agents on the machine sitting on your desk, not in a data center somewhere.
The full reveal came on October 7, 2026, at an event in San Francisco. It followed earlier previews at events in May and June 2026, so this was less a surprise drop and more a season finale.
The headline hardware is a lineup of RTX Spark-based laptops and compact desktops. Those machines can deliver up to 1 petaflop of AI performance.
A petaflop is a quadrillion calculations per second. Put more plainly, that is the kind of number that used to belong on a spec sheet for server racks, not something you could fit in a backpack.
The RTX Spark devices can also handle models with more than 120 billion parameters. Parameters are roughly the internal settings a model learns during training, and more of them usually means a more capable model.
Pre-orders opened on October 7, 2026. Shipping is scheduled to start on October 16, which leaves a fairly short gap between the keynote and the doorstep.
On pricing, Microsoft’s Surface Laptop Ultra starts at $2,599, with higher configurations approaching $5,900. A Dev Box starts at $5,999. These are not impulse buys, which tells you exactly who the target audience is.
The DGX Station goes Windows
At the top of the range sits Nvidia’s DGX Station for Windows, also announced on October 7. It runs on the GB300 Grace Blackwell Ultra Desktop Superchip.
That chip pushes performance up to 20 petaflops. The station can run AI models with up to 1 trillion parameters.
For context, that is roughly 20 times the peak AI throughput of the RTX Spark machines. It also handles models more than eight times the size of the 120 billion parameter threshold the laptops and compact desktops are built for.
The software side matters just as much as the silicon. Windows 11 is getting new features aimed at AI agents, including Execution Containers designed to secure AI operations.
Think of Execution Containers as a sandbox for software that acts on your behalf. If an AI agent is going to click, browse and edit files for you, it helps to keep it in a fenced yard rather than letting it roam the whole house.
Windows 11 will also support hybrid operations that combine local and cloud capabilities. In practice, that means a workload can run on the PC when that makes sense and reach out to the cloud when it needs more muscle.
How we got here
This is not Microsoft’s first attempt at putting AI on the PC itself. Its Copilot+ PCs were an earlier push in the same direction.
The difference now is scale. Copilot+ focused on lighter on-device tasks, while the new hardware is aimed at much larger local AI workloads, the kind that previously almost always meant renting time in the cloud.
Microsoft is framing all of this around what it calls “agentic AI.” That refers to AI systems that do not just answer questions but take actions and complete multistep tasks. The company is positioning Windows as a leading platform for that category.
What this means for developers, enterprises and the AI market
The pitch to buyers rests on three things: privacy, latency and cost. Running a model locally means sensitive data does not have to leave the device, responses do not have to make a round trip to a server, and heavy users may avoid ongoing cloud bills.
That makes the core audience clear. Developers, creators and enterprises with steady AI workloads are the ones most likely to justify a $5,999 machine over a monthly cloud invoice.
For developers specifically, a desktop that can hold a model of more than 120 billion parameters changes the testing loop. Prototyping against a large model without waiting on remote infrastructure is a real workflow upgrade, assuming the software stack holds up in daily use.
That creates an interesting tension for Microsoft. Azure is one of its biggest businesses, and here it is helping customers keep some AI work off the cloud. The hybrid design looks like the hedge: keep the PC capable, but keep the cloud one call away.
Nvidia, meanwhile, gets another surface to sell into. Its accelerated computing already powers large cloud deployments, and now its chips are headed into laptops, compact desktops and a Windows-flavored DGX Station.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

9 hours ago
15








English (US) ·