Intel Announces Major Chip Architecture Breakthrough
Intel has unveiled what it describes as its most significant chip breakthrough in two decades. While specific details of the new architecture are emerging, the announcement signals a major effort by the company to reassert itself in the high-performance computing and AI markets. In a related move, Intel also struck a deal with a chip startup in which CEO Pat Gelsinger has a personal investment, indicating a strategy of using unconventional partnerships to drive innovation.
- The new architecture, featured in the "Panther Lake" Core Ultra series 3 processors, is built on Intel's 18A process node, equivalent to a 2-nanometer class node. This is the company's first major transistor architecture change in over a decade and incorporates RibbonFET gate-all-around transistors and PowerVia backside power delivery to improve performance and efficiency. - Intel's deal with a chip startup involves xLight, a company developing free-electron laser technology for EUV lithography, which is critical for advanced chip manufacturing. Former Intel CEO Pat Gelsinger is the executive chairman of xLight, and the startup is set to receive up to $150 million in U.S. government funding through the CHIPS Act. - The "Panther Lake" processors are projected to deliver over 50% faster CPU and graphics performance compared to the previous generation and include a new neural processing unit (NPU) for AI acceleration, offering up to 180 trillion operations per second (TOPS). - This architectural shift is a core component of Intel's IDM 2.0 strategy, a plan to regain manufacturing leadership from competitors like TSMC and Samsung. This involves investing heavily in new fabs in the U.S. and expanding Intel Foundry Services (IFS) to manufacture chips for external customers. - Intel's custom ASIC business is a growing focus, with annualized revenue exceeding $1 billion. The company aims to compete with Broadcom and Marvell by bundling custom chip design with its in-house manufacturing, and has secured a deal to produce custom 5G chips for Ericsson and AI-related chips for AWS on its 18A process. - The competitive landscape for AI compute is intense, with hyperscalers like AWS, Google, and Microsoft increasingly designing their own custom silicon. This "build vs. buy" decision for hyperscalers is driven by the need for workload optimization and cost reduction, though they often still partner with third-party providers to ensure capacity and speed to market. - Intel is also facing renewed competition in the PC market from AMD and Qualcomm, which is gaining traction with its ARM-based chips for Windows laptops. - The deal with the chip startup in which CEO Pat Gelsinger has a personal investment is with Fractile.ai, a British company focused on low-cost AI inference technology. This investment highlights a strategy of leveraging external innovation to compete with Nvidia in the inference market.
