Competition in on-device computing is intensifying as the artificial intelligence (AI) market shifts toward agents that execute commands. Until now, AI models have been developed through training and inference at large data centers, and these models had to retrieve information stored in external cloud spaces to answer user questions and commands. But because personal-use agents are designed to run 24 hours a day on ordinary PCs and smartphones, boosting the computing capabilities of computers and smart devices has become critical.
According to U.S. media reports on Sunday, Microsoft's developer conference "Build," which opens Monday in San Francisco, is expected to unveil a collaboration on the first Windows AI laptop using Nvidia central processing units (CPUs) as the main processor.
Nvidia CEO Jensen Huang also unveiled the company's first AI PC chip, the "N1 X," at his keynote speech at "GTC Taipei" in Taiwan the previous day, announcing a partnership with Microsoft to enter the AI laptop market. The N1 X is a chip for Windows PCs based on the Arm (a U.K. semiconductor design firm) architecture (design method). It is a chip capable of running Nvidia software on Microsoft's operating system and will be installed in Nvidia's "RTX Spark" laptop series. The PCs Nvidia previewed are laptops designed for AI agents and will be manufactured by makers including Dell, Lenovo, and HP.
Until now, the most commonly used PC processors have been CPUs from Intel and AMD built on the "x86" (architecture). Nvidia chips were used in tablets running "Windows RT," a lightweight operating system based on Windows 8, in 2012, but failed to make a significant impact. Analysts say, however, that the latest collaboration between Nvidia and Microsoft could have considerable ripple effects amid surging demand for AI agents.
Observers say the mainstreaming of agents is imminent, in which ordinary users — not just developers — can run agents on personal PCs or smartphones without needing to connect to the cloud via the internet. Demand for desktops and laptops suitable for running agents could surge. The strategies of Microsoft, which has been struggling in the PC market against Apple's MacBook series running Apple's own operating system, and Nvidia, which is seeking to expand its business beyond data center graphics processing units (GPUs) into CPUs, have aligned with each other.
OpenClaw, created by Austrian-born Peter Steinberger, drew enormous popularity after being released on the developer-sharing website GitHub in January this year. OpenClaw was released as an open-source (open) platform, allowing ordinary people to easily run agents. Inspired by OpenClaw's lobster logo, "lobster raising" has spread like a fad in China. The era has arrived in which non-developers can run agents on PCs 24 hours a day to perform tasks.
Although Microsoft CEO Satya Nadella disparaged OpenClaw in February, comparing the security-vulnerable program to a "virus," the agent's power is so significant that Microsoft established a related development team as OpenClaw's popularity soared. Omar Shahine, Microsoft Corporate Vice President (CVP), held a media briefing at the Mountain View, California, campus on April 27 and stressed that the "Lobster Project" would become an important business, saying, "Through open-source projects that leverage infinite AI capabilities and unlimited tokens, AI can do more than 10 times the work of humans." The Lobster Project is a business that incorporates OpenClaw-like functionality into M365, Microsoft's enterprise subscription service.
Apple is also expected to accelerate its on-device competition to prepare for the agent era. Billions of Apple devices, including iPhones, iPads, and MacBooks, could become powerful weapons for running agents. In particular, on-device computing is considered well-suited for incorporating AI while preserving Apple's management philosophy that emphasizes security and privacy protection.
Apple has struggled in the AI race, even losing the No. 2 market capitalization spot to Google, which has poured astronomical sums into AI infrastructure and is leading the competition. While Samsung Electronics (005930.KS) accelerated its on-device strategy by installing Bixby, Gemini, and Perplexity on smartphones, Apple failed to deliver results with Apple Intelligence and ultimately chose to integrate with Google's AI Gemini. Against this backdrop, on-device computing could serve as a card to turn the tide.
Apple already felt the power of AI agents through surprise earnings last quarter. Mac, a non-core business, was expected to post revenue in the low $8 billion range, but disclosed revenue came in at $8.4 billion. Increased consumption of the $599 entry-level MacBook (MacBook Neo) was one cause, but OpenClaw's popularity also had a significant impact. Mac mini consumption surged after the Apple desktop received praise as optimal for running OpenClaw. Apple announced that the Mac mini was the best-selling desktop in China. On the conference call, Apple CEO Tim Cook explained that one cause of the Mac sales increase was a surge in demand for Mac mini and Mac Studio devices for running local AI models such as OpenClaw, adding that it was an outcome Apple had not anticipated.
Against this backdrop, observers expect Apple to unveil a new on-device strategy at the Worldwide Developers Conference (WWDC), which opens on the 8th. While Big Tech firms are pouring hundreds of billions of dollars into AI data centers, Apple may focus on expanding AI functions running on various smart devices. The fact that incoming CEO John Ternus is a hardware expert lends weight to the view that Apple will move to strengthen on-device computing. The Information, citing Apple insiders, reported that "Apple is likely to demonstrate at WWDC the advantages it can offer when running AI models locally on devices, drawing on its experience designing custom chips for the iPhone, Apple Watch, and Mac."
