Microsoft Brings AI Processing Power to New ARM-Based Surface Devices

Microsoft announced the Surface Pro 12 and Surface Laptop 13 on May 14, 2025, expanding its Copilot+ PC range with Snapdragon X Plus processors and built-in AI capabilities. Both devices include 8-core Snapdragon X Plus chips paired with 45 TOPS NPUs — a measure of how much artificial intelligence work the processor can handle locally, on the device itself. These numbers place the devices squarely in Microsoft's bet that ARM architecture (a different design approach to chips than the Intel processors most of us grew up with) is ready for mainstream laptops.

The new devices became available in Hong Kong starting July 15, 2025, following Microsoft's usual pattern of testing hardware in Asian markets first. In the UK, the Surface Pro 12 with Snapdragon X Plus, WiFi, 16GB RAM, and 256GB storage is now priced at £800, down from an initial £999.

Why ARM? The Efficiency Argument

The Snapdragon X Plus represents Microsoft's ongoing shift toward ARM-based computing for high-end laptops. The 45 TOPS NPU meets Microsoft's Copilot+ PC requirements, which demand at least 40 TOPS of AI processing power for running AI tasks directly on your device rather than sending them to the cloud. This enables features like real-time translation, smarter voice processing, and running language models locally for productivity tasks — all without your data leaving the machine.

The 8-core Snapdragon X Plus sits somewhere in the middle: less raw power than the flagship Snapdragon X Elite chips, but designed to balance speed with battery life and heat management. For office work, this is a deliberate choice.

Both ARM and Intel, for Now

Microsoft is not ditching Intel. The company also offers Surface Laptop for Business models with Intel Core Ultra processors, which started shipping in February 2025 at roughly HK$11,788. You can still buy Copilot+ PCs with traditional Intel chips in both 13.8-inch and 15-inch sizes, depending on what software you need to run.

This matters because while ARM processors are brilliant at battery life and AI work, they still don't run every piece of enterprise software perfectly. Industry-specific applications, legacy code, and specialized tools often depend on Intel's x86 instruction set — the underlying language those chips understand. ARM translation layers (software that converts x86 instructions on the fly) have improved, but they are not flawless yet.

This pattern of supporting both old and new architectures is familiar from technology history. We saw it when Microsoft transitioned Windows from 32-bit to 64-bit computing — the company ran both in parallel until the market naturally settled on the newer standard. The ARM and Intel split here follows the same pragmatic logic: let customers adopt ARM where it makes sense, keep x86 available where it is essential, and let the transition happen at its own pace.

AI Processing Is No Longer Optional

The 45 TOPS figure across both new Surface devices signals a shift in how Microsoft thinks about baseline specs. AI processing capability has moved from a premium feature to something expected across the range. That computation enables transformer-based models — the technology behind modern AI — to run directly on your machine for everyday tasks: summarizing documents, transcribing meetings and identifying who spoke when, translating between languages, and suggesting content improvements all without talking to the cloud.

For organizations dealing with sensitive data — banks, hospitals, government agencies — this matters considerably. Data never leaves the device, which cuts regulatory risk and gives you control over where your information goes.

Pricing and Market Positioning

The £800 promotional price for the Surface Pro 12 places it below many premium Intel-based laptops while offering comparable screen quality, build materials, and the same AI acceleration. Microsoft's decision to launch first in Hong Kong, about two months after the May announcement, reflects how the company typically tests new hardware. Asian markets serve as a proving ground before rolling out to Western markets, especially for devices with newer chips or experimental designs.

The broader market context suggests Microsoft has concluded that ARM adoption is ready for the mainstream. Battery life has improved, software compatibility has gotten better through smarter emulation, and more applications are being written natively for ARM64 — the ARM instruction set used here. The friction of switching to ARM has dropped enough that productivity users can make the jump without major complications.

By building Copilot+ PC requirements around NPU-equipped devices, Microsoft is also nudging the wider laptop market toward AI acceleration. OEMs and customers have an incentive to prioritize devices with neural processing units, which helps differentiate Windows machines in an otherwise crowded laptop market where features have begun to look alike.

What This Means for Workplaces and IT Departments

IT decision-makers now have more granular options. A sales team that spends all day on video calls and presentations could run ARM-based machines for all-day battery and AI-enhanced presentation features. An engineering team needing specialized CAD software or legacy applications could stick with x86-based devices. Both work under the same management framework and get the same Copilot+ AI features, which simplifies training and policy rollouts.

Microsoft's support for both architectures suggests the company sees the ARM transition as gradual rather than disruptive. Organizations can move to ARM devices where they deliver clear wins — battery life, heat, cost — while keeping x86 machines where software compatibility demands it. This reduces the risk and uncertainty that usually haunts platform shifts.

The picture emerging here is of a market in transition, but not a rushed one. Microsoft is laying the groundwork for ARM to become normal in professional computing without forcing anyone off Intel until it genuinely makes sense. That pragmatism is worth noting.