Microsoft's New Surface Computers Get Better at AI Tasks

On May 14, 2025, Microsoft announced two new Surface computers: the Surface Pro 12 and Surface Laptop 13. Both use a processor called Snapdragon X Plus, which includes special hardware designed to handle artificial intelligence tasks quickly and efficiently. These devices are part of Microsoft's push to bring AI capabilities directly into the computers people use every day.

The computers started selling in Hong Kong on July 15, 2025. In the UK, the Surface Pro 12 with the base configuration costs £800 (down from £999).

Why This Matters: A Shift to ARM-Based Chips

The Snapdragon processor represents a significant change in how computers work. Think of a processor like the engine of a car. For decades, Microsoft computers used one type of engine (called x86). Now they are adding a different type (called ARM), which was originally designed for phones and tablets.

The Snapdragon X Plus has a special chip inside called an NPU, or "neural processing unit." This chip is designed specifically for AI tasks. It can perform what's measured as 45 TOPS — a speed measurement for AI work. Microsoft requires Copilot+ computers to have at least 40 TOPS, which these new devices meet.

What does this actually do for you? Your new Surface can now handle certain AI jobs on the computer itself, without sending your information to the internet. It can translate languages, transcribe voice conversations, summarize documents, and suggest content — all while keeping your data on your device.

Two Types of Processors, Not One

Microsoft is not abandoning the older x86 processors from Intel. Instead, the company is offering both types. Earlier in 2025, Microsoft released Surface Laptop models with Intel processors. Customers can choose whichever works best for them.

Why two options? Many older programs and specialized software only work with x86 processors. A accountant using industry-specific software, an engineer using technical tools, or a company using decades-old business applications might need an x86 computer to work properly. ARM is newer, and while it is improving, it cannot yet run everything that older computers can.

This is similar to how Microsoft handled previous big changes — when computers went from 32-bit to 64-bit many years ago, the company offered both options until one became the clear standard. That gives people time to switch without forcing them immediately.

What This Approach Actually Lets Companies Do

For large organizations buying many computers, this choice matters. A sales team that travels constantly and cares most about battery life can get ARM computers with the Snapdragon chip. An engineering team that needs specific software can stay with Intel. Both groups get the same AI features, so training is simpler.

The company seems confident that ARM processors have matured enough for mainstream work. Battery life is better, programs run better on ARM now than they did a few years ago, and more software is being written for ARM from the start. This has made switching less risky.

Microsoft's real strategy here is clear: by making AI processing a requirement for Copilot+ computers, the company is pushing both computer makers and customers toward devices with this new hardware. In a market where most laptops look and feel similar, having built-in AI capabilities that actually work is one way to stand out.

The Broader Picture

Microsoft's continued investment in ARM-based Surface devices shows the company believes this shift is real and lasting, not a temporary experiment. The gradual rollout — starting in Hong Kong and moving to other regions — is Microsoft's standard approach with new hardware. It lets the company test how people actually use these devices before offering them everywhere.

For people shopping for a new computer, what matters most is whether you need the latest AI features, whether you have specialized software that requires older processor types, and how much battery life you value. These new Surfaces give you more choices to match your actual needs rather than accepting one-size-fits-all hardware.