Microsoft has confirmed an October 7 Windows and Surface event in San Francisco, with CEO Satya Nadella and Nvidia CEO Jensen Huang both set to appear on stage, per reports from The Verge and Engadget. The session starts at 10 a.m. Pacific and centers on RTX Spark, Surface hardware and the next step for local AI on Windows PCs. It is Microsoft’s first big Windows stage moment since the Copilot+ PC push in May 2024, more than two years ago, and the shared guest list is unusual: chipmakers rarely share a consumer keynote with the Windows team.
RTX Spark, which Nvidia and Microsoft first teased in June as a 1-petaflop superchip with the full CUDA and RTX ecosystem, is the centerpiece. The chip targets Windows PCs that treat large local AI models as a first-class workload rather than a bolt-on feature. Earlier this month Nvidia said confidentual computing configurations of its GPUs now run inference inside Apple’s Private Cloud Compute, so the company is simultaneously courting hyperscalers and client device makers. A joint appearance of Nadella, Huang and Windows and Devices chief Pavan Davuluri reinforces that both companies see local inference as the next PC battleground.
Why the Copilot+ window matters
When Microsoft launched Copilot+ PCs in May 2024, the pitch was on-device AI with NPUs inside Snapdragon X chips. The market share of Arm-based Windows laptops has since grown, but the segment has not produced a breakout consumer hit, and reviews have been mixed on local model quality. That is the gap RTX Spark is aimed at bridging. The Copilot+ certification required 40 TOPS of NPU throughput for local model execution, a mark most Intel and AMD laptop chips have not reached, which kept many OEMs from shipping qualifying products and madeCopilot+ a Snapdragon-adjacent term.
RTX Spark removes that gate by putting a full GPU stack with CUDA acceleration in the box, so the local model story no longer depends on a specific NPU arithmetic benchmark. For developers, that means a portable local AI target inside the Windows ecosystem. For surface and OEM partners, that means high-end inference in a thin chassis rather than a bulky workstation, and the possibility of lighter devices with real local model capability.
Reports from Engadget and PCMag confirm that the second track next to the Arm/Qualcomm Copilot+ designs is the plan: Windows PCs that treat large local models as a first-class workload, with Surface as Microsoft’s reference design and Dell, Lenovo, ASUS and Samsung as the volume channel.
Surface refresh and existing lineup
Microsoft’s Surface line is due for a refresh. The current Surface Pro and Surface Laptop use Snapdragon X Plus and X Elite chips, and November’s smaller Surface Pro and Surface Laptop got faster Snapdragon silicon at higher prices, a move TechSpot’s review called strong on graphics but hard to defend on value. A next generation Surface machine running an Nvidia-built superchip would be the biggest hardware shift in the line since the original Surface Pro 8 moved the tablet hybrid to x86 in 2021. Whether the event shows only a reference design or a sellable product remains unclear.
The competitive calendar adds pressure. Google confirmed preorders for the first Googlebook hardware category starting at $1,299, with the Lenovo Googlebook 15 arriving at $1,099 with a 3K OLED panel. Qualcomm’s new Snapdragon 8 Elite Gen 6 reaches 5 GHz on a 2nm process and pushes agentic AI across phones, laptops and earbuds. AMD bought Fei-Fei Li’s World Labs for $8.2 billion in an all-stock deal, pushing into spatial AI. Each of these forces Microsoft to defend the Windows PC franchise at the exact moment when the local AI story needs a flagship.
Pricing for RTX Spark laptops has not been confirmed. If Microsoft follows Copilot+ launch pricing, expect three tiers with RTX Spark variants at a premium over the base model, and a similar pattern from OEM partners. At least one Surface-branded reference design is a near certainty given Surface’s history as Microsoft’s spec-setter for the Windows ecosystem.
What to expect from Microsoft on stage
Microsoft’s own framing in the invite, as reported across outlets, is Windows, RTX Spark, Surface and OEM PCs, so expect new Surface hardware, Windows 11 updates built around local model execution, and maybe a new Copilot+ variant using Nvidia silicon. The company reportedly confirmed the broad outline to outlets in advance. Details on exact chips, SKUs and dates are the usual stage territory. Expect benchmarks too, since Nvidia’s own slides gave 1.8 times faster agentic task completion for the Vera CPU and RTX designs have a track record of measuring well on LLM token throughput.
The October 7 date lands in a busy stretch for AI hardware announcements. Nvidia’s DGX Spark 64GB, a variant for developers who want more memory in the local AI desktop box, arrived on October 2. NVIDIA GTC Berlin runs October 20-22. Apple’s iPhone Duo rollout starts October 16 with preorders, and availability begins October 23. Google’s Googlebook category is already shipping. Add the Microsoft event and October becomes the densest hardware month of the year, which is exactly the environment where a well-executed local AI PC story gets traction, and where a weak one gets buried.
Local AI versus the cloud
The event also matters for the longer argument over where AI workloads should run. Microsoft’s cloud business is the biggest beneficiary of hosted inference, but a local-first PC pitch moves some of that spending back to one-time hardware purchases, and it blurs the line between client devices and the server room. Nvidia’s client-side push, whether on Windows or in Apple’s cloud, is a direct play for the same developer mindshare.
Whether Windows PCs with a petaflop of AI throughput actually shift purchasing decisions, or become a niche for developers, is an open question. The Copilot+ segment has not answered it yet. October 7 is when Microsoft gets its chance to try.