Microsoft Taps Intel's 18A Process for Next-Gen AI Accelerator

Intel has secured Microsoft as a major foundry client for its 18A process, sources say

Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust. In a nutshell: Intel has been struggling to establish a competitive chip foundry business for years. Now, industry sources indicate the US corporation is set to manufacture the next few generations of AI accelerators to support Microsoft's massive data center expansion, highlighting Intel's renewed push to capture a larger share of the growing AI hardware market. According to a recent report from SemiAccurate, Microsoft has chosen Intel Foundry to manufacture its Maia 3 AI chip. The machine learning accelerator, code-named Griffin, is expected to play a major role in Redmond's plans to develop new generations of self-sustaining data centers. Maia 3 is an upgrade to Microsoft's earlier Maia projects, which the company continues to present as a potential alternative to third-party AI chips. In a recent interview, Chief Technology Officer Kevin Scott confirmed that Redmond is developing its next AI chips even as most of its current AI cloud workloads still rely on Nvidia's GPU technology. If the Maia 3 project goes as planned, rumors suggest that Microsoft could continue using Intel's foundry to manufacture the next few generations of the AI chip. Maia 3 production would rely on Intel's 18A process, which provides some interesting food for thought if previous reports are accurate. Chipzilla was reportedly facing significant setbacks with its process node only a few months ago. Manufacturing issues such as low yields and poor quality threatened to derail Intel's earlier expectations. Now, the company is introducing 18A as a process ready for customer projects and the first advanced sub-2-nanometer manufacturing node available in North America. The Maia 3 chip could use Intel's 18A or even its 18A-P node. The second-generation 18A process is designed to improve both performance and power efficiency, with specific optimizations for threshold voltage and leakage. Microsoft's next-generation Maia chips could employ even more advanced nodes such as 18A-PT or 14A. Intel specifically designed the 18A-PT node for customers in the artificial intelligence and high-performance computing sectors. Building on the performance and power efficiency gains achieved with 18A, the new process offers greater package scalability and integration capabilities to further enhance generative AI workloads and chatbot performance. The number of rumors and manufacturing deals related to Intel Foundry is starting to resemble a financial roller coaster, but Microsoft's alleged Maia 3 plans should be taken with a grain of salt. Earlier this month, industry insiders claimed that Intel could begin manufacturing x86 chips for AMD. Meanwhile, the company's plans for its upcoming PC client CPUs appear to be proceeding as expected.

Microsoft to reportedly use Intel Foundry and 18A for its next-gen Maia AI accelerator

TL;DR: Intel's advanced 18A semiconductor process, produced at its Arizona Fab 52, is set for use in upcoming Panther Lake and Clearwater Forest chips. Microsoft plans to utilize this technology for its Maia 3 AI accelerator, highlighting Intel Foundry's growing role in high-performance chip manufacturing alongside partners like NVIDIA and ARM. Intel's 18A process or node, created in-house at its US-based Fab 52 semiconductor facility or Intel Foundry in Arizona, is set to be used in the company's upcoming Panther Lake and Clearwater Forest chips. The production-ready technology and Foundry are open for business, and Intel is looking to partner with big tech to manufacture new chips and technologies using the 18A or 18A-P process. One of these new customers, according to a report over at SemiAccurate (via TechPowerUp), is Microsoft. The company is apparently planning to use Intel's 18A process for its next-gen Maia 3 accelerator for AI, codenamed "Griffin." And with that, depending on the yields and success of its latest AI hardware, Microsoft could also tap future Intel process technology like 18A-PT and 14A for future Maia accelerators. Microsoft expressing interest in leveraging Intel technology for its AI accelerators isn't new. Microsoft CEO Satya Nadella released a statement about Microsoft's plans to use 18A for an undisclosed chip when Intel announced its Foundry roadmap in early 2024. "We are in the midst of a very exciting platform shift that will fundamentally transform productivity for every individual organization and the entire industry," Nadella said at the time. "To achieve this vision, we need a reliable supply of the most advanced, high-performance, and high-quality semiconductors. That's why we are so excited to work with Intel Foundry, and why we have chosen a chip design that we plan to produce on Intel 18A process." For 18A to succeed and Intel Foundry to become a notable and viable option in the chip market, Intel needs customers like Microsoft to make it happen. Other notable customers and partners for Intel's next-gen chip technology include NVIDIA, Broadcom, and ARM. As for Microsoft's Maia 3 accelerator, details remain scarce. The report indicates that the company could opt for the more advanced 18A-P process with RibbonFET and PowerVia technologies, which enhance both performance and efficiency.