The year of AI PCs

Microsoft CEO Satya Nadella said in January that 2024 would mark the year AI becomes “the first-class component of every PC.”

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 learn more about how AI will be integrated into Windows and what they can do with it on new AI PCs.

The 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. Before the Google event, OpenAI announced its new GPT-4o model. Microsoft is the main investor in OpenAI and its Copilot technology is based on OpenAI’s models.

For Microsoft, the challenge is twofold: maintaining a leading position in AI and strengthening PC sales, which have been in the doldrums for two years following an upgrade cycle during the pandemic.

In a recent note on Dell to investors, Morgan Stanley analyst Erik Woodring wrote that he remains “optimistic about the recovery of the PC market” due to customer feedback and recent “revisions to the rise in original laptop manufacturers (ODM) releases.

Technology industry researcher Gartner estimated that PC shipments rose 0.9% in the quarter after a multi-year slump. PC demand has been “slightly better than expected,” Microsoft Chief Financial Officer Amy Hood said during the company’s quarterly earnings conference call last month.

Microsoft’s new AI tools could offer another reason for businesses and individuals to upgrade their aging computers, whether they’re made by HP, Dell or Lenovo.

“While Copilot for Windows does not directly drive monetization, we expect it to increase Windows usage, Windows stickiness, customers to more powerful, more expensive PCs (and thus more revenue for Microsoft per year). device) and likely search revenue,” Bernstein analysts wrote in a note to investors on April 26, the day after Microsoft’s earnings release.

While Microsoft will provide the software needed to handle some of the AI ​​tasks sent over 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 a transcription without a connection.

The key hardware addition to an AI PC is something called a neural processing unit. NPUs go beyond the capabilities of traditional central processing units (CPUs) and are designed to specifically handle artificial intelligence tasks. Traditionally, they are used by companies like Apple to enhance photos and videos or for voice recognition.

Microsoft has not yet specified what AI PCs would be capable of without an Internet connection. But Google’s PIxel 8 Pro phone, which lacks a full computer processor, can summarize and transcribe recordings, recommend replies to text messages and much more thanks to its Gemini Nano AI.

Computers powered by Intel’s latest Lunar Lake chips with a dedicated NPU are expected to arrive in late 2024. Qualcomm’s Snapdragon during the quarter.

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

Apple has been using NPUs for years and recently highlighted them in its new M4 chip for the iPad Pro. The M4 chip is expected to launch on the next series of Macs this year.

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

Intel still controls 78% of the PC chip market, followed by AMD at 13%, according to recent data from Canalys.

In the past, Qualcomm has promoted Snapdragon Arm-based computers by touting their longer battery life, thinner design, and other benefits such as 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 apps on Arm, but questions remain. The company even has an FAQ page dedicated to computers running on ARM hardware.

Microsoft will also host sessions such as “AI Everywhere” explaining how to “accelerate generative AI models” on devices running in the cloud.

An “Azure AI Studio” session will look at how developers can create their own Copilot chatbots, which can be similar to what Google and OpenAI are doing with Gemini and ChatGPT. Imagine, for example, a company creating a chatbot that can help its employees select health benefits.

WATCH: Investing in the future of AI