U.S. Chip Makers Embrace Collaboration


EETimes

May 21, 2021

As Congress debates funding levels for reviving U.S. chip manufacturing, commercial partnerships are forming around strategic technologies such as 5G wireless and AI chips.

In a bid to accelerate 5G wireless connectivity, GlobalFoundries is licensing gallium nitride technology from Raytheon Technologies. This collaboration is aimed at boosting not just 5G, but also subsequent 6G wireless infrastructure applications.

Meanwhile, IBM is touting its pivot to a “collaborative model” with foundry partners such as Intel and Samsung Electronics to fabricate its Power processors while developing energy-efficient AI chips.

GlobalFoundries said the wireless chip partnership announced Wednesday (May 19) includes joint development and commercialization of Raytheon’s GaN-on-silicon technology. The resulting devices would be aimed at boosting RF performance for 5G and 6G phones and wireless networks.

The foundry has yet to respond to our request for details on performance improvements.

The GaN-on-Si devices will be fabricated at GlobalFoundries’ Fab 9 facility in Essex Junction, Vt. As horse-trading advances in the Senate for funding U.S. semiconductor R&D, Sen. Patrick Leahy, D-Vt., chairman of the Senate Appropriations Committee, called the 5G chip partnership “good news for the nation’s semiconductor supply chain and competitiveness.

Under terms of the agreement, Raytheon will license its proprietary GaN-on-Si technology and “technical expertise” to GlobalFoundries, which will develop new semiconductor devices for next-generation wireless applications. The technology is touted as handling higher power and resulting heat levels beyond the capability of legacy wireless systems.

The deal with Raytheon reflects the U.S. foundry operator’s emphasis on key RF components required to keep pace in the race to deploy 5G and, eventually, 6G wireless networks. Those networks will transport everything from streaming video to big data collected by edge AI devices.

Company officials have stressed GlobalFoundries’ emphasis on “high-mix applications [with] modest volumes,” including those incorporating RF front-ends, microcontrollers and battery management. That pragmatic strategy varies markedly from cutting-edge process technologies offered by foundry leaders like Taiwan Semiconductor Manufacturing Co. (TSMC) and Samsung Electronics.

The partnership with Raytheon “will enable everything from AI-supported phones and driverless cars to the smart grid, as well as governments’ access to data and networks which are essential to national security,” GlobalFoundries CEO Tom Caufield said in announcing the agreement.

The ramping of advanced 5G chip production at a former IBM fab illustrates how the foundry business has evolved in the last decade. Chip designers such as IBM have sold off foundry facilities, instead seeking manufacturing partners as they move to smaller processor nodes.

“Given the extreme competition, market volatility and the scale of capital investments needed, many companies have pivoted from vertically integrated development to a model that leverages best-in-class strategic partnerships,” Zachary Lemnios, vice president of IBM research, noted in a blog post.

 “The trust and security of microelectronics created through partnerships requires in-depth understanding of a complex global supply chain that includes design services, EDA tools, intellectual property design elements, photomask design tools, foundry services, assembly, test, [and] packaging services and component packaging,” added Lemnios, a former U.S. assistant secretary of defense.

IBM currently licenses IC designs to GlobalFoundries and Samsung.

Besides AI chips, IBM’s hardware alliances also include a partnership with Intel aimed at fabricating the world’s first 2-nm chips. As it looks to rebuild its manufacturing prowess, Intel announced an alliance with IBM to development new fab processes. “Intel is joining a team of TSMC competitors to pool research and share the huge cost of developing ever-smaller transistors,” processor market analyst Linley Gwennap noted in a Forbes.com commentary.