GF’s First Decade: Finding Their Center

By: Dave Lammers

The second of a three-part series looking back at GF’s first 10 years, and looking forward to the next decade and beyond.

GLOBALFOUNDRIES’ first decade was a ten-year period during which the foundry met a series of remarkable challenges, ranging from technology transitions that tested the mettle of its staff to decisions about how to differentiate GF from its competition.

How GF got from delivering just a few technology solutions for AMD in 2009 to where it is today – with a portfolio of RF-based solutions dominating the marketplace and a serious thrust underway in fully depleted SOI – is arguably the most interesting story in the semiconductor industry over the past decade.

The journey to build an at-scale global semiconductor manufacturer
Source: GF

I talked to Gregg Bartlett, now GF’s head of engineering and technology, who was recruited to join the brand-new company at the very beginning. Bartlett remembers 2009 as the year when he joined GLOBALFOUNDRIES and immediately was thrust into what was a demanding start for the new foundry, but it also was a time when management learned to assess how much risk it should take on technology and product transitions – a valuable lesson.

A Challenging Beginning

GF, spun out from AMD in early 2009 with financial backing from Abu Dhabi’s Mubadala investment fund, was immediately tasked with bringing up a new AMD product architecture, on a new 32nm high-k/metal gate process.  That combination proved to be very challenging to build in high volume production.

Bartlett recalled, AMD’s flagship new “APU” part –which pioneered the combination of a graphics processing engine and a CPU core on one die – required focused execution at GF’s flagship fab in Dresden, Germany. “It was quite a humbling time, a really difficult time – but we persevered,” Bartlett said.

Patrick Moorhead, now an industry analyst at Moor Insights & Strategy, was part of AMD’s management team a decade ago. “If you look at it from a historical perspective, Llano had one of the first large pieces of graphics IP that was integrated with the APU. At that time, no smart phone processor was delivering anything like it. But getting it going came in fits and starts, and there were super, super challenges in getting that part going.”

Bartlett said the main lesson GF management learned was to manage the amount of risk being taken. “Right out of the chute, GF was trying to be something that was going to be hard to do, introduce a revolutionary new silicon technology. We tried to change our chip architecture, while at the same time making technological innovations. I look at our company today, and can say we have learned a lot about working with customers to better manage technology innovation with customer design innovation.”

From Three to 300 Customers

Then in 2010 GF’s management, with backing from the leadership at the Mubadala fund, acquired Chartered Semiconductor. GF and the Dresden fab had produced just a few different parts for a small group of customers who needed the leading-edge technology offered in the fab, and Bartlett said GF’s management realized it needed to learn to deal with multiple customers with various process technologies.

“We asked ourselves then: How do you change a fab with three customers to one with 30 customers, and then 300 customers? How do you change your business processes? The answer was, of course: not easily. Chartered brought us skills and processes to deal with hundreds of customers all at once,” Bartlett said, praising the “customer orientation” that Chartered’s Singapore team brought into the GF culture.

GLOBALFOUNDRIES acquired the Singapore site from Chartered Semiconductor on January 13, 2010. Chartered Semiconductor originally opened the fab in 1995. It was one of the first 200mm fabs. Source: GF

“What Chartered had figured out was how to always have contingency plans for customers to ensure they hit their market introductions,” he said.

Bartlett gave credit to the leaders at the Mubadala investment fund, saying they were “willing to make the large investment in acquiring Chartered, knowing how transformative the acquisition was.” The acquisition gave GF several 200 and 300mm fabs in Singapore, and roughly 200 customers. The acquisition began a transformation of the new foundry, creating a process of dealing with many dozens of customers.

“This was an inorganic transformation that accelerated our ability to operate as a global foundry. Today, Singapore is a very successful part of our corporation and has proliferated processes and talent across our company,” Bartlett said.

A ‘Very Fortuitous” Cooperation

Building on lessons learned from Dresden’s work on the 32/28nm node, five years later the new Malta fab faced the challenge at the 14nm node, when FinFETs were first introduced.

This time, GF and AMD came through it smoothly, with Moorhead describing the happy endpoint: “When the first versions of (the AMD 14nm) Ryzen came out, as an industry analyst I had to ask myself, ‘What is not going to work?’ It is one thing to ship a few thousand samples, but shipping hundreds of thousands, and then millions, is a different thing. It was a super-impressive ramp, to say the least. And to do that while outperforming many of Intel’s products on cache cell size, on performance of the transistors, on power consumption and the amount of heat, was truly remarkable.”

GF’s groundbreaking ceremony to officially start construction of the Admin 2 office building as part of Fab 8 in Saratoga County, New York. Source: GF

Learning to Collaborate

To achieve that 14nm success, GF went through an honest self-appraisal. Bartlett, who by then was GF’s chief technology officer, recalls that at the time, members of the IBM technology alliance all had come up with different answers to the 14nm node. IBM was pursuing an SOI-based process that included its deep-trench embedded DRAM, while GF and Samsung went their separate ways toward bulk CMOS finFET 14nm processes.

While GF and Samsung were holding intermittent discussions about if and how to cooperate, they realized cooperation was a good way to manage supply risk for a very large customer in the mobile space. This brought GF and Samsung together to collaborate on deployment of a single technology between the two companies.

When it became apparent that there was a deal, that turned out to be very beneficial,” Bartlett said. By working with Samsung on 14nm, GF accelerated its transition to go into high-volume manufacturing. At the same time, GF was able to grow the technology development team in Malta that is now differentiating that platform for more applications and customers today.

A First Step Toward Differentiation with FD-SOI

Bartlett said that when GF was in the process of acquiring IBM Microelectronics, management concluded it needed a technology diversification strategy, something different than what its main competitor was doing at the 28nm node.

“FD-SOI was our first conscious choice to differentiate on a technology platform. Rather than go imitate the industry leader, we decided to drive a superior technology solution for market segments we believed needed FDSOI such as IoT and integrated RF,” Bartlett said.

GF’s 22FDX platform technology relies on planar transistors as well as fully-depleted SOI wafers
Source: GF

That ability to make tough decisions, to be flexible, was needed again last year, when GF’s new management team, led by Tom Caulfield, concluded that it needed to pivot away from 7nm to be more relevant to our customers. This pivot freed development resources, allowing more investment in the technology platforms that the majority of GF’s customer base was using, including 12nm FinFET, 22FDX®, RF, silicon photonics and other platforms.

Moorhead said he believes the 22FDX process has power advantages for edge processing that will play a major role as 5G RF and data processing converge. “The opportunity for GF is huge. 5G — and all of the associated circuitry that goes with it – will be fundamentally connected to the Internet of Things. That is going to define the next ten years. It gets us to hyper-connectivity,” Moorhead said.

I asked Bartlett if GF’s experiences over its first decade had created a battle-tested workforce, better able to meet future challenges.

“We’ve learned a lot from our first ten years of existence, understanding which decisions we got right, and which ones we got wrong and why. And when we got them wrong, to have gone through them together as one GF team means we are better able to deal with future challenges.” he said.

“Will we make mistakes going forward? Very likely, but not because we hadn’t considered the risks. When I look at our company today, we have learned a lot about how you support our customers, how you manage the risk you are taking, and most importantly about being relevant to the customers where their needs align with core capabilities. We know who we are, where our core competencies are and are excited for a future going on offense with those capabilities.”

In the final installment in this series, we will examine the future of GF and how its differentiated portfolio will enable its customers and change the industry that’s changing the world.

About Author

Dave Lammers

Dave Lammers

Dave Lammers is a contributing writer for Solid State Technology and a contributing blogger for GF’s Foundry Files. Dave started writing about the semiconductor industry while working at the Associated Press Tokyo bureau in the early 1980s, a time of rapid growth for the industry. He joined E.E. Times in 1985, covering Japan, Korea, and Taiwan for the next 14 years while based in Tokyo. In 1998 Dave, his wife Mieko, and their four children moved to Austin to set up a Texas bureau for E.E. Times. A graduate of the University of Notre Dame, Dave received a master’s in journalism at the University of Missouri School of Journalism.