We have previously discussed rumors in the business realm concerning Microsoft’s Copilot AI service transitioning to running directly on PCs rather than relying on cloud services. However, verifying these rumors had proven challenging until recently at Intel’s AI Summit in Taipei. During an on-the-record session with Tom’s Hardware, Intel executives confirmed that Copilot will indeed operate directly on PCs. Additionally, there has been talk among business associates about a 40 TOPS requirement for the latest AI Computers.
Microsoft has maintained a level of secrecy regarding its plans for AI computers, even allowing Intel to formally outline what constitutes an AI computer. The newly developed concept by Microsoft, Intel, and Inc. specifies that an AI Computer will incorporate a Neurological Control System (NPU), CPU, GPU, Microsoft’s Copilot, and a real Navigator directly integrated into the console.
While computers meeting these specifications are currently being shipped, this marks just the initial phase of the AI PC initiative. In response to inquiries about potential memory requirements for AI PCs, Intel disclosed Microsoft’s anticipated requirements.
Todd Lewellen, Vice President of Intel’s Client Computing Group, mentioned the progression towards the next generation of AI Desktops with a 40 TOPS NPU requirement. Intel is gearing up to introduce its next-gen product falling within this category.
As we advance towards the next generation, the capability to run more processes locally will increase, similar to the approach with Copilot where more components will operate locally on the client. While key functionalities will run on the NPU, not all components of Copilot will necessarily transition to local operation.
It is noteworthy that Intel’s Meteor Lake NPU offers up to 34 TOPS, while AMD’s Ryzen platform boasts an NPU with 39 TOPS, both slightly below Microsoft’s specified requirement. Qualcomm is set to launch its X Elite chips later this year, featuring 45 TOPS of performance.
Lewellen elaborated on Microsoft’s aim to enhance user experience with the new platforms. According to the manufacturer, Copilot should run on the NPU rather than the GPU to optimize battery life.
Looking ahead, Intel’s roadmap includes next-gen processors catering to various market segments for AI applications. The Lunar Lake processors slated for release later this year promise three times more AI performance on the GPU and NPU, surpassing the TOPS requirements.
In the interim, Intel is focused on augmenting the array of AI features embedded in its silicon. The company plans to introduce 300 new AI-enabled features on its Meteor Lake processors this year. Intel’s collaboration with OpenVino is expected to yield Intel-specific AI features as it transitions to the Lunar Lake generation.
The AI PC Accelerator Program by Intel aims to leverage developer enthusiasm and propel the adoption of AI software, aligning with the goal of selling 100 million AI PCs by 2025. Microsoft’s Copilot is set to run on NPUs from all vendors through DirectML, with higher TOPS translating to enhanced performance, marking the onset of a TOPS competition in both silicon development and marketing strategies in the coming years.
