Computex 2024: Nvidia Teases Faster Chips, Project Rubin Revealed, Microsoft Copilot Partnership Announced

Nvidia fired a salvo at the Computex conference, showcasing its commitment to maintaining leadership in the AI chip market. The company unveiled a multi-pronged strategy that includes faster chip release cycles, a next-generation AI platform codenamed Rubin, and a collaboration with Microsoft to bring powerful AI capabilities to a wider range of devices.

Annual Chip Releases

Previously content with a two-year release schedule for its AI chips, Nvidia is shifting gears and moving to annual updates. CEO Jensen Huang cited the ever-growing demand for processing power in the AI field, a trend dubbed "computation inflation" by Huang. This accelerated release cycle highlights the competitive landscape in AI chips, where Nvidia is vying to stay ahead of the curve.

Blackwell Gets Supercharged: The Ultra Upgrade

The recently announced Blackwell platform is still on track for a 2024 release. However, Nvidia teased an even more powerful variant – the Blackwell Ultra chip – arriving in 2025. While specifics are under wraps, Nvidia promises significant performance gains with the Blackwell Ultra, all while achieving a remarkable 98% cost reduction and a 97% drop in energy consumption.

Project Rubin: A Peek into the AI Future

Looking beyond the immediate horizon, Nvidia unveiled a glimpse of its next-generation AI platform – Rubin. Scheduled for a 2026 launch, Rubin promises to be a paradigm shift. The architecture will reportedly integrate new GPUs and introduce a central processing unit (CPU) codenamed Vera. A key highlight of Rubin is its utilization of HBM4, the next iteration of high-bandwidth memory. This component is crucial for AI accelerators but has faced supply constraints due to high demand. By incorporating HBM4, Nvidia is addressing a potential bottleneck and ensuring smooth operation for future AI applications.

Nvidia and Microsoft Join Forces: Copilot Takes Flight on GPUs

Beyond chip advancements, Nvidia announced a collaboration with Microsoft to extend the reach of Copilot, a powerful AI integration platform. This partnership will bring the Copilot runtime to GPUs later in 2024. The significance lies in unlocking powerful AI capabilities for Microsoft's Copilot+ laptops, irrespective of the underlying processor architecture – be it AMD, Intel, or even Arm.

Windows Copilot+ Laptops With RTX

Nvidia also confirmed its presence in upcoming Windows Copilot+ laptops. These devices will sport Nvidia's RTX 40-series GPUs working alongside AMD's Zen 5 (Ryzen AI) CPUs and Intel's Lunar Lake processors. While the immediate future holds Windows AI laptops with these combinations, Nvidia hinted at the possibility of future iterations featuring Arm CPUs paired with Nvidia GPUs. This tantalizing prospect suggests Nvidia is keeping its options open to deliver the most performant AI solutions possible.