GlobalFoundries discussed its technology roadmap; its plan to invest $1.4B in 2021; its expectation it will achieve profitability; and its intention to hold an IPO, perhaps in 2022...
GlobalFoundries has a unified growth vision for the whole company as well as local strategies to develop its major manufacturing clusters in Asia, Europe, and the U.S. The industry’s third largest contract maker of semiconductors believes that it is on the right track with its specialty technology strategy, which it believes will eventually lead to net profitability. The company plans to continue expanding its production capacities and introduce process technologies that are relevant for high-volume, long-lifecycle devices. Europe has a special place in GlobalFoundries’s strategy, so it makes sense to take a look at the state of things here.
Emerging applications need special technologies
The world has significantly reinvented itself in the past 10 – 20 years, but in the next 10 we are going to face far more dramatic changes both for business and for society. Historically, major business and social transformations have always been enabled by technological advancements. When homo learnt how to control fire and produce stone tools millions of years ago, the society transformed dramatically. Fast forward to today, our technologies got drastically more sophisticated and the number of devices and tools we use in our daily life — for personal, business, and/or work matters — is countless. And since the world is always reinventing itself, these devices tend to get smarter, and the number of them we use keeps growing.
Appliances like self-driving cars, autonomous vehicles, AI-enabled industrial equipment, exascale supercomputers, edge-computing devices, wearable electronics, and many other emerging applications need brand-new kinds of processors and system-on-chips. These new kinds of PUs and SoCs are only starting to evolve and even today they are sometimes made using manufacturing technologies tailored either for very high performance, or for very low power consumption. Meanwhile, what these fabrication processes need to offer is a combination of features specific to a particular product. Specialization of fabrication processes is an ongoing trend, but tailoring a node for a particular application does not mean an all-new technology developed for scratch. At times, introduction of a new library or insertion of a new module and subsequent re-qualification usually opens doors to numerous new possibilities.
22FDX hits shipments milestone
One of the good examples of such ‘expandable’ technology platforms is GlobalFoundries’s 22FDX that uses fully depleted silicon on insulator (FD-SOI) wafers.
The platform was introduced back in mid-2015 as a low-cost alternative to 14nm/16nm FinFET-based nodes. Initially, GlobalFoundries offered four versions of its 22FDX for different markets: 22FD-ulp (ultra-low-power), 22FD-ull (ultra-low leakage), 22FD-uhp (ultra-high performance), and 22FD-rfa (integrated RF and analog). Over time, GlobalFoundries added capabilities like embedded MRAM to the platform to make it more versatile.
Developers of applications that require maximum performance no matter the costs generally ignored 22FDX (perhaps, because it arrived later than competing FinFET technologies, or maybe because not all of its features could be used early in its lifecycle), but designers of various high-volume devices who appreciate low idle power, low leakage, and design costs in line with those of 28nm chips adopted the technology rather broadly.
GlobalFoundries said last week said that it had shipped over 350 million chips produced using its 22FDX node. The company indicated it has also had realized $4.5 billion in design wins, a sum that spans from the past into the future. To put the number into context, GlobalFoundries earned about $6.2 billion in revenue back in 2017. The contract maker of semiconductors has probably earned a lot more selling FinFET-based chips, but its 22FDX is a long-lasting node that will serve the company for quite a while, which perfectly aligns with its strategy.
Ecosystem for 22FDX’s active body biasing is ready
One of the key advantages of 22FDX is its software-controlled adaptive body bias (ABB) capability that allows to precisely fine-tune the transistor threshold voltage of a circuit and find the right balance between the chip’s performance (forward body bias, FBB), power consumption, and leakage (reverse body bias, RBB). Side bonuses of ABB include area optimizations and improved reliability, which greatly affects costs. The ABB feature is particularly important for ultra-low power chips like those used in IoT and wearable/hearable devices.
Unfortunately, very few of the 22FDX designs have ever used adaptive body bias capability as tools and IP to implement it properly were scarce until mid-2019. GlobalFoundries has been talking about adaptive body biasing for quite a while and at last year’s GTC it saidthat the ecosystem to implement adaptive body bias (ABB) was finally ready.
“Body biasing is arguably the most differentiating capability of FDX, but it has not been widely adopted because the complete body-biasing ecosystem has not been there,” said Jamie Schaeffer, a senior director of Mobility and Wireless Infrastructure Market at GlobalFoundries. “Over the last couple of years, we have built the first body-bias ecosystem, including the libraries and memories with process, voltage, temperature, and body bias characterization. There are body bias generators, and reference designs from Cadence and Synopsys. We are ready, from design and planning all the way through characterization and test.”
This year GlobalFoundries disclosed names of two companies — GreenWaves Technologies and Perceive — who have successfully managed to implement 22FDX’s ABB capability into their designs: an application processor platform for hearables as well as an edge inference processor for IoT devices.
Further demonstrating development of GlobalFoundries’s 22FDX platform is the announcement of the 22FDX+ platform that targets a broad range of applications that belong to IoT, 5G, automotive, and satellite communications worlds.
The 22FDX+ platform is essentially an evolution of 22FDX that promises to give developers some additional controls to enable high performance or ultra-low power as well as means to implement specialty capabilities.
The very first process technology from the 22FDX+ family will be 22FDX RF+. This node features digital and RF enhancements to meet requirements of next-generation front-end-modules (FEM) for smartphones. The technology will be available for production starting Q1 2021 at Fab 1 in Dresden, Germany.
Where Is 12FDX?
The announcement of the 22FDX+ platform this September begs the question: where is the 12FDX platform introduced back in 2016 and meant to be available for tape outs in the first half of 2019?
When it was unveiled four years ago, 12FDX promised to deliver what GlobalFoundries called a full node scaling over 22FDX and bring performance of 10nm-class FinFET technologies at a lower cost and power. The technology was expected to bring a 15% performance increase over then-leading-edge 14LPP/CLN16 nodes (at the same clocks and complexity) and “as much as 50% lower power consumption.”
Back in 2016, the technology was acclaimed by analysts, scholars, and industry peers. NXP publicly praised the platform, a rare occurrence, but did not publicly commit to use it.
To some degree, the history of 22FDX shows that ecosystem development for FD-SOI fabrication processes takes a rather long time. Therefore, while certain 12FDX platform-based manufacturing technologies might be ready or nearly ready and even IP companies might have some offerings in the pipeline, the interest for the node is low. Therefore, GlobalFoundries does not want to make the technology formally available or even announce an estimated time of arrival at this point.
“We have R&D getting ready for 12FDX, but not until we have a segment that we can differentiate a solution where we come in and do something on that front,” said Thomas Caulfield, CEO of GlobalFoundries. “This is not high-speed digital where it’s the race to the next feature side, because the complexity is really in the device and doing a single function. We are an SoC type company. So, we bring big features together of many features together in one piece of silicon. And the device doesn’t matter as much as the feature size. Yeah. So the answer is yes, we have a 12 FDX roadmap. We will launch it when it has the right solution space for our customers to differentiate their products and ours.”
GlobalFoundries itself does not deny that it had to delay its 12FDX platform because there are no market segments it could address just yet. Those companies who need performance and transistor density of 10nm-class nodes for their designs can now use GlobalFoundries’s 12LP+, whereas for low-power/low-cost high-volume designs there is 22FDX. At present, 12FDX is largely overkill for the markets addressed by 22FDX and more complex devices for emerging applications that will need 12FDX yet have to be developed.
“I think it’s really important that we articulate our strategy,” said Caulfield. “We are not a slave to device sizes. We don’t say we have to do the next device size for the sake of doing the next device size. We think about market segments. We think about the features that are needed. Device types, feature-rich IPs are more important than the size of the device.”
Analyst Rob Enderle agrees with this point of view.
“Some folks still seem to think that if they build something folks will buy it,” said Enderle. “Until developers embrace a new technology, it isn’t viable. So, until there is some developer interest this is a nice science experiment, nothing more.”
Fab 1 to double capacity
Having served for decades, GlobalFoundries’s Fab 1 in Dresden, Germany, is still one of the company’s primary workhorses along with Fab 8 in Malta county, New York. Indeed, the fab is used to make chips for automotive, mobile, IoT, industrial, and many other high-growth applications.
At present, Fab 1 can process wafers using 22FDX, 28SLP, 40/45/55NV as well as BCDLite technologies. GlobalFoundries plans for a range of 400,000 to 500,000 wafer starts per year in 2020/2021, but it has expansion plans for the production facility. Demand for these technologies appears to be significant enough to afford a plan to double the capacity of the fab to 900,000 ~ 1 million wafer starts in the coming years.
“Fab 1 has a little over 50% of the floor space available for capacity growth,” said Caulfield. “So, as we have generated this demand, this has become one of the key growth engines for us in the company. This is what we will need, you know, PPP, public, private partnership in co-investment to accelerate the build out of that, of that critical technology for the EMEA region and the rest of the world.”
Smart expansion and public private partnerships
GlobalFoundries has invested billions of dollars in various fabs throughout its history and its production facilities still need to depreciate, which is a reason why the company have never reported net profitability. Meanwhile, GlobalFoundries cannot cease its investments as its competitors are always expanding. Typically, GlobalFoundries spends about $700 million a year on expansion, but it will double the amount to approximately $1.4 billion next year to upgrade its manufacturing facilities, which includes Fab 1 and Fab 8.
“In our operating model, a typical year for us would be to spend about $700 million a year worldwide in our footprint to enable capacity growth or to bring new innovations in the marketplace,” said Caulfield. “The demand is so strong for our products heading into 2021 and beyond that we are going to double that next year, including a fair amount of that in our Fab 1 Dresden facility.”
But at this time the company intends to invest only with support from local governments. Production of semiconductors is a very capital-intensive industry, so it is not surprising that companies like GlobalFoundries need support from authorities.
“The reason semiconductors have been concentrated in certain parts of the world is because governments early on have decided that for national security interests, they needed to have that capability and meet co-investments,” said Caulfield. “To be competitive regionally governments, including the U.S., is doing this now, European union has done it as well […]. Because you know, we’re running businesses, we can’t be in uncompetitive positions when other nations of the world are attracting us to come in and do our manufacturing there. I think the European Union has a great track record of doing this. I believe they’ll, they’ll do more and we’re going to be looking to partner with them or continue our partnership to do more in Dresden.”
On Track for IPO in the second half of 2022
GlobalFoundries is on track to go public sometimes in the second half of 2022. The company says that it wants to hit certain financial milestones before going public and thanks to rising demand for chips, it is fairly sure that it will be able to hit them by 2022.
“It was always financial milestone dependent,” said Caulfield. “We wanted to go public when we were going to be able to get the true value of this enterprise and growing our scale was a big part of that. All that’s happened in the last year is given the acceleration of our industry and the business momentum we have is put higher confidence in hitting those financial milestones, therefore higher confidence in our 2022 target date.”
GlobalFoundries does not make promises that it will be net profitable by the time it goes public as it still needs to depreciate its fabs. Meanwhile, the company is confident that in a number of years it will be profitable, though it does not disclose when this is going to happen.
After changing its strategy rather drastically in 2018, GlobalFoundries became a much more cautious company in general. The contract maker of semiconductors has clearly made a bet on FD-SOI technologies and its 22FDX platform seems to be successful. At the same time, GlobalFoundries does not rush its successor — 12FDX — to the market as it does not foresee strong demand for this technology just yet.
Overall demand for chips is strong and GF seems to have a rather competitive lineup of process technologies. Therefore, the company plans to double its CapEx spending from $700 million to around $1.4 billion next year to upgrade its production facilities, including Fab 1 in Germany and Fab 8 in the U.S.
Traditionally, GlobalFoundries has received various incentives from local governments when it built or expanded its fabs. The company expects Bavarian authorities to help it with its expansion in Dresden too, but it has not yet made any formal proposals to the local government.
Having invested billions of dollars in its fabs over the years, GlobalFoundries has never reported net profitability as it still has to depreciate its fabs. The contract maker of chips is confident that it will be profitable one day several years down the road, but it does not set any goals publicly.
GlobalFoundries is on track for its IPO in the second half 2022 after it hits certain financial milestones and completes its expansions.