Blog

Microsoft CEO Satya Nadella said in January that 2024 will mark the year that AI will become a “premium part of every computer.”

The company already offers its Copilot chatbot assistant in the Bing search engine and for a fee in the Office productivity software. Now PC users will be able to hear more about how AI will be built into Windows and what they can do with it on the new AI PCs.

Build comes days after Google I/O, where the search giant unveiled its most powerful AI model yet and showed how its Gemini AI will work on computers and phones. Ahead of Google’s OpenAI event, it announced its new GPT-4o model. Microsoft is OpenAI’s lead investor and its Copilot technology is based on OpenAI models.

For Microsoft, the challenge is twofold: maintaining a prominent position in AI and bolstering PC sales, which have stagnated for the past two years after an upgrade cycle during the pandemic.

In a recent Dell note to investors, Morgan Stanley analyst Eric Woodring wrote that he remains “optimistic about the recovery of the PC market” due to comments from customers and recent “upward revisions to notebook original design manufacturer (ODM) builds.” .

Technology industry researcher Gartner estimated that PC shipments rose 0.9% in the quarter after a multi-year decline. Demand for PCs was “a little better than expected,” Microsoft CFO Amy Hood said during the company’s quarterly earnings call last month.

New AI tools from Microsoft could offer yet another reason for enterprise and consumer customers to upgrade their aging PCs, whether they’re made by HP, Dell or Lenovo.

“Although Copilot for Windows does not directly drive monetization, we believe it should increase Windows usage, Windows stickiness, customers to more expensive, more powerful PCs (and therefore more revenue for Microsoft per device), and possibly revenue.” from search,” Bernstein analysts wrote in a note to investors on April 26, a day after Microsoft reported earnings.

Although Microsoft will provide the software to handle some of the AI ​​tasks sent to the Internet, its computers will be powered by chips from AMD, Intel and Qualcomm for offline AI tasks. This could include, for example, using your voice to ask Copilot to summarize an offline transcription.

The key hardware addition to the AI ​​PC is what is called a Neural Processing Unit. NPUs go beyond the capabilities of traditional central processing units (CPUs) and are designed specifically to handle artificial intelligence tasks. Traditionally, they have been used by companies such as Apple to enhance photos and videos or for speech recognition.

Microsoft has not yet said what the AI ​​computers will be capable of without an internet connection. But Google’s PIxel 8 Pro phone, which doesn’t have a full computer processor, can summarize and transcribe recordings, recommend replies to text messages and much more using its Gemini Nano AI.

PCs with Intel’s latest Lunar Lake chips with a dedicated NPU are expected to arrive in late 2024. Qualcomm’s Snapdragon X Elite chip with an NPU will be available in the middle of this year, while AMD’s latest Ryzen Pro is expected sometime in the quarter.

Intel says the chips enable things like “real-time language translation, automated inference, and enhanced gaming environments.”

Apple has used NPUs for years and recently highlighted them in its new M4 chip for the iPad Pro. The M4 chip is expected to be released in the next round of Macs sometime this year.

Qualcomm, unlike Intel and AMD, offers chips powered by an Arm-based architecture. One of Microsoft’s sessions will talk about “The Next Generation of Windows on Arm,” which will likely cover how Windows runs on Qualcomm chips and how that differs from Intel and AMD’s versions of Windows.

Intel still controls 78% of the PC chip market, followed by AMD with 13%, according to the latest figures from Canalys.

In the past, Qualcomm has advertised Snapdragon Arm-based PCs, touting their longer battery life, slimmer design and other benefits like cellular connections. But earlier versions of Qualcomm’s chips were limited in what they offered consumers. In 2018, for example, the company’s Snapdragon 835 chip couldn’t run most Windows apps.

Microsoft has since improved Windows to handle traditional Arm applications, but questions remain. The company even has an FAQ page dedicated to computers running ARM hardware.

Microsoft will also host sessions such as “AI Everywhere,” covering how to “accelerate generative AI models” on devices that run in the cloud.

An “Azure AI Studio” session will look at how developers can build their own Copilot chatbots, which may be similar to what Google and OpenAI are doing with Gemini and ChatGPT. For example, imagine a company that creates a chatbot that can help employees choose health benefits.

WATCHING: Investing in the future of AI