Like most corporate communications, how companies talk about themselves should be taken with a couple of pounds of salt. The worst tend to be simplistically self-aggrandizing. Yet other examples can offer meaningful insights into an organization’s plans and self-perception.
That is particularly true following significant changes in leadership, such as the January announcement that Pat Gelsinger (then CEO of VMware) would replace Bob Swan as Intel’s CEO. On July 26 at the “Intel Accelerated” event, Gelsinger, Dr. Ann Kelleher, SVP and GM of Intel Technology Development and other company executives revealed a roadmap for new products that will be delivered through 2025 and beyond. What Pat Gelsinger and his team discussed is certainly worth considering, but how they talked about it is also intriguing.
The detailing of Intel’s new process and packaging innovations began with a bold disclaimer by Gelsinger. “The industry has long recognized that traditional nanometer-based process node naming stopped matching the actual gate-length metric in 1997.” In other words, the nanometer (or “nm”) related marketing terms and rhetoric that silicon manufacturers have depended on for 20+ years has reached the end of its useful life since it doesn’t meaningfully reflect or describe semiconductor performance.
To address that, Gelsinger said Intel will employ five new node names and associated innovations for Core and Xeon products launched through 2025 and beyond. Intel 7, Intel 4, and Intel 3 will leverage existing FinFet transistor optimizations and next gen extreme ultraviolet (EUV) production tools, resulting in a near 50% increase in performance per watt between now and 2H 2023.
Intel is also working on next gen EUV; Numerical High Aperture (High NA) EUV that will be used in future products. Those include Intel 20A which is planned to ramp in 2024 as the company enters what Gelsinger called the “angstrom era” of semiconductor production with RibbonFET, its first new transistor architecture since 2011.
RibbonFET is designed to deliver markedly faster switching speeds with the same drive current as multiple fins but in a smaller footprint. The new chips will also incorporate another Intel innovation; PowerVia, a backside power delivery technology that optimizes signal transmission by eliminating power routing on the front side of the wafer. Intel 18A (scheduled for 2025) will use the same technologies to deliver another significant performance boost.
Packaging and partner potatoes
Packaging is an increasingly important issue for Intel, especially considering the IDM 2.0 strategy that marks a significant effort by the company to sell Intel Foundry Services (IFS) offerings to “fabless” semiconductor customers. IFS is a new standalone foundry business led by Dr. Randhir Thakur, who will report directly to Gelsinger. At Intel Accelerated, Gelsinger noted that AWS has signed on to become IFS’s first customer.
On the packaging side, Intel will continue to leverage the EMIB (embedded multi-die interconnect bridge) 2.5D embedded bridge solutions it introduced in 2017 and Foveros, a wafer-level packaging capability that enables 3D wafer stacking. Intel also announced Foveros Omni, a highly flexible next gen technology that will be ready for volume manufacturing in 2023, and Foveros Direct (also available in 2023) which complements Foveros Omni by enabling an order of magnitude increase in interconnect density.
Along with its process and packaging innovations, Intel highlighted notable new future-focused collaborations, including strategic partnerships with ASML on the development of High NA FUV, and with Qualcomm which plans to use the new Intel 20A process technology.
Intel emphasized that its process and packaging breakthroughs were primarily developed at company facilities in Oregon and Arizona, highlighting the company’s position as the only leading-edge semiconductor player with U.S.-based research and development and manufacturing capabilities.
Intel also noted that its new and next-gen technologies were developed in close collaboration with partners in both the U.S. and Europe, which the company said are “key to bringing foundational innovations from the lab to high-volume manufacturing.”
So, what’s the takeaway from all this? Bombastic corporate rhetoric or evidence of substantial advances in Intel’s current efforts and future plans?
I would argue the latter. In concert with the March announcement (where, along with the IFS business, Gelsinger enumerated planned improvements for Intel’s products and fabs), this latest announcement qualifies as the second half of a “here’s where we’re going/here’s how we’ll get there” one-two punch.
Some might demure, noting that Intel’s plans require the on-time delivery of numerous, highly complex new and significantly updated technologies and production processes, any of which might slip or stumble along the way.
That’s a valid view, but two points undercut it. First, Pat Gelsinger is a significantly different leader than Intel has seen in a very long time. He is deeply technically astute and knows (as shown by his 11+ years at EMC and VMware) how to make a good company great. Gelsinger has a reputation as a fair but rigorous leader who demands honesty and transparency, and he has a history of choosing team members who display similar characteristics. Count on him and Intel to catch and address issues before they become significant problems.
Second, the new partnerships and the customer that Intel highlighted in its announcement suggest that the company has made believers out of some significant players. Qualcomm’s plans to use the new Intel 20A process technology is a notable vote of confidence. Given ASML’s leadership in global semiconductor manufacturing equipment and technologies, it’s hard to think of a better partner for Intel to collaborate with on High NA FUV. Signing up AWS as the first customer for IFS looks like both a significant win and a serious shot across the bows of third-party foundries, like TSMC and Samsung.
Looked at another way, if Intel’s March announcement focused on strategic vision, this roadmap to 2025 and beyond, discussion of new and in-development solutions and revealing of new partners and customers was all about practical delivery. Intel knows where it wants to go, and has the motive, means and leadership to get there. The next five years should be beneficial for Intel’s semiconductor customers and instructive for its competitors.