Intel debuts Core Ultra processors at AI Everywhere event

Intel (INTC) on Thursday took the wraps off of its new Core Ultra processors for ultrathin laptops during its AI Everywhere event in New York.

The chips, which are available in new laptops starting today, pack a number of big firsts for Intel, including a chiplet design and neural processing unit. All of this comes as CEO Pat Gelsinger seeks to reinvigorate the semiconductor giant.

The Core Ultra processors will be available in three different configurations: the Core Ultra 5, Core Ultra 7, and Core Ultra 9. The Core Ultra 5 and Core Ultra 7 are available in new PCs now, while the Core Ultra 9 will land in the coming months.

Chiplets, or as Intel calls them, tiles, are a means of breaking up a single chip into different components, which can easily be swapped in and out depending on needs. The Core Ultra line has a compute tile, graphics tile, I/O tile, and SoC tile.

All of this is meant to make Intel’s Core Ultra chips more powerful and power efficient as it works to claw back market share from the likes of AMD (AMD) — and keep Qualcomm (QCOM) and Apple (AAPL) off its back.

Chiplets aren’t a new concept, though. AMD is already using them in its own chips, but it’s a big move for Intel and could provide a boost to the company’s sales as it seeks to prove it can innovate as fast as its rivals. Intel saw revenue of $63.1 billion in fiscal 2022 versus $79 billion in 2021.

The tiles themselves have a number of interesting capabilities. The compute tiles pack the chip’s performance cores, which focus on power-hungry tasks, and efficiency cores, which take over when you’re running less demanding programs. The SoC tile, meanwhile, gets Intel’s new Low Power Island, which can run tasks that require even less power.

The SoC tile also comes with Intel’s NPU, which fits into the event’s AI Everywhere theme. The NPU is meant to handle some onboard AI tasks, But don’t expect to be running something like ChatGPT directly on your laptop. That kind of functionality is still incredibly process intensive and has to be performed in the cloud.