TLDR
- Microsoft rolled out the Surface Laptop Ultra, a new premium laptop built around Nvidia’s AI hardware.
- Starting price is $2,599, with preorders open now and shipping set for October 16.
- A new Microsoft tool called Execution Containers aims to keep AI agents from accessing data without permission.
- Meta’s Muse assistant and DeepSeek’s V4 model are both coming to the new Windows hardware.
- Memory chip shortages have driven up prices across the AI PC market, Microsoft included.
Microsoft stock ticked up 0.23% in after-hours trading Wednesday following the unveiling of its newest AI hardware push. The announcement centers on bringing AI processing directly onto personal computers rather than routing it through the cloud.
The centerpiece is the Surface Laptop Ultra, priced from $2,599 and open for preorder now. Microsoft says it will begin shipping on October 16.
Inside the machine is Nvidia’s Blackwell RTX Spark GPU. Nvidia CEO Jensen Huang compared its processing power to the company’s 2016 DGX-1 server, a machine that originally sold for $250 million.
Pavan Davuluri, Microsoft’s executive vice president of Windows and devices, introduced the laptop at a San Francisco event. He pointed out its magnetic USB-C charging port, a design similar to Apple’s MagSafe, alongside two additional standard USB-C ports.
The Ultra weighs 4.4 pounds and has a 15-inch screen. Specs top out at a 20-core processor, 128 gigabytes of memory, and a terabyte of storage, with that configuration priced at $5,899.
Gartner estimates currently place Microsoft outside the top six PC makers by unit shipments. AI features are central to how the company hopes to change that.
Alongside the hardware, Microsoft unveiled Execution Containers, a system built to stop AI agents from taking unauthorized actions or pulling data they shouldn’t touch.
Anthropic, OpenAI, and Nvidia have all signed on to use the new containment tools. According to Davuluri, IT teams will be able to define rules that Windows then applies automatically across company devices.
Local AI Gets a Boost
Several AI models that previously ran only in the cloud can now operate directly on the new devices. That includes Nvidia’s open-source Nemotron model and a version of DeepSeek’s V4 model from China.
Davuluri said DeepSeek’s V4 needs at least 60 gigabytes of memory to run locally. He claimed it beats OpenAI’s GPT-5 on certain coding and reasoning benchmarks.
Microsoft’s Copilot assistant will take a hybrid approach going forward. Copilot chief Jacob Andreou explained that harder tasks stay in the cloud, while simpler ones can run on-device to cut costs or protect user privacy.
Meta is bringing its Muse AI agent to Windows as a dedicated app. The open-source OpenClaw agent system is also expected to integrate with Microsoft’s new security framework.
Cost Remains a Sticking Point
Pricing is shaping up to be the biggest obstacle for AI-capable PCs industry wide. A global memory chip shortage is the main driver behind rising costs.
Nvidia’s DGX Spark desktop, for comparison, now costs $6,950 after a roughly 75% price increase since its debut. Apple’s closest MacBook Pro equivalent, with a 40-core graphics chip and similar storage, runs $6,700.
Microsoft’s base Surface Laptop Ultra costs more than Apple’s entry-level MacBook Pro. It does offer 24 gigabytes of memory versus 16 gigabytes on Apple’s starting configuration.
Moor Insights & Strategy analyst Anshel Sag pointed to a timing mismatch. He said the software needed for local AI has finally caught up, but the hardware required to run it has become harder to afford.
Microsoft has built its own PCs since 2012. CEO Satya Nadella said this week that the company views its role in computing as entering a new phase shaped by AI.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
