GlobalFoundries CEO Tom Caulfield argues that GF has a big role to play in global manufacturing, as making chips becomes a political issue.
In the midst of global upheaval that has inspired the Biden Administration to earmark $50 billion to ensure that more advanced semiconductor manufacturing gets performed domestically, Tom Caulfield is scrapping for recognition for a chip company that is already here, that is already doing advanced IC manufacturing, but that hardly anyone outside the electronics industry has ever heard of because it specializes in making the types of ICs that few outside the electronics industry pay any attention to.
The company is GlobalFoundries. Caulfield is its CEO. If America’s goal is to be more self-reliant in the semiconductor business, Caulfield told EE Times, “they’re going to need a bevy of players to do that, and first in line is GF. We want to do our part, and we think it’s an important part.”
The electronics industry is being affected by two enormously disruptive events: the pandemic, which is probably a short-term disaster (fingers crossed), and the political antagonism between the US and China, which is looking to be a long-term issue. The combination has made the global supply of both commercial and military semiconductors look exceedingly tenuous, alarming just about everybody.
When it comes to public perception, processors have all the glamor. It started with “Intel Inside” and it’s still going on today as everyone from high performance computing specialists to online gamers to even some average smartphone users eagerly anticipate each new generation of CPUs and GPUs.
Many of the top CPU designers are household names (Apple, Intel, Qualcomm, etc.) and the leading manufacturers are also well-known (Intel, Samsung, TSMC). Meanwhile, unless you work for Analog Devices, On Semiconductor, Qorvo, Renesas, or any of the many successful semiconductor companies that aren’t involved in designing leading-edge processors, your mother has likely never heard of those companies. Unless your neighbors invest in UMC, or received a tip about SkyWater Technology’s upcoming IPO, they almost certainly don’t know either of those two even exists.
Public attention is usually not that important, but today it’s the crux of the matter. If the average person hasn’t heard of a company, then the odds are good that neither has the average Congressman, and Congressmen are about to divvy up $50 billion among companies in the semiconductor industry.
So, as the US government promotes its enticement of TSMC to build and operate a foundry within US borders, and as Intel loudly announces plans to set up a separate foundry business — and makes a case that a subsequent reward of government incentives would not be out of line — the CEO of the publicly obscure GlobalFoundries becomes available for a video call to talk about manufacturing semiconductors, Intel’s foundry announcement, and how GF fits into the big picture.
For most of the history of the semiconductor industry, the focus was on compute-centric applications: the personal computer, networking gear. Massive investment has been continuously made to get the technological return required: ever greater processing power.
Yet the compute-centric segment is only 30% of the entire semiconductor market (give or take a few percentage points) in terms of revenue, Caulfield points out. And despite that, most of the world’s investment in manufacturing capacity has been for this smaller part of the market. He called it “single-digit” manufacturing, a reference to the leading edge being at the 7nm node, going to 5nm, then 3nm, and perhaps even smaller.
For the past 10 years or so, starting with the rise of the smartphone, there’s been another, slowly growing trend — pervasive computing — “more things requiring intelligence,” Caulfield said. These devices don’t necessarily need more compute power; in fact, most don’t. They need high-voltage ICs, low-voltage ICs, embedded memories, RF and mixed-signal circuitry. Making these chips doesn’t require single-digit manufacturing capabilities; the advantages of going smaller have already been exhausted for most of these types of ICs.
These are the markets that GlobalFoundries caters to.
When the pandemic hit, Caulfield said its arrival accelerated the demand for pervasive computing beyond what anyone had expected. “It was a decade in a year…, a decade of technology adoption in one year in 2020 because of Covid,” Caulfield said. He allowed that’s a bit of an exaggeration, but he likes the phrase “a decade in a year” — and it’s accurate enough.
Caulfield said GF saw $1 billion more demand than it had been forecasting as recently as last in June, “and we’re roughly a $6 billion company.”
GF wasn’t prepared. Given that recent chip shortages have been global and cut across market segments, neither was anyone else.
The automotive industry was particularly hard-hit by those shortages. Caulfield said GF met with auto industry executives, who spelled out the consequences of ongoing shortages. If an automotive company shuts down a plant, then five or ten other companies in the ecosystem go down with it. That would be a lot of jobs lost.
GF said it took steps to help make up for some of those recent shortages. “We didn’t need a note from the White House to go do the right allocation of our capacity to keep auto makers going. We had to take hard actions.”
Of course, GF barely got noticed for it.
The upshot, however, is that this acceleration in demand for pervasive computing exposed an under-investment in manufacturing capacity for the other 70% of the IC market, Caulfield observed.
And amidst all this, Intel announces it’s going to establish a commercial foundry service. And of course that captures global attention.
“We were always going to get to the point where manufacturing was going to become fashionable again,” Caulfield said. “We live in a physical world, and someone’s got to make this stuff. Intel sits there as one of the few able to make. That they’re getting into foundry is a validation. They’ll be a force in that 30%, and we’ll be a force in that other 70%, to cover the space that TSMC plays in, because TSMC plays in 100% of the market.”
Interestingly, semiconductor technology is heading in a direction where the 70 percenters and the 30 percenters are going to be playing with each other more often. Functions that had been integrated in systems-on-chips (SoC) are heading toward being separated out again, and combined in other ways. There are different ways to perform (or just describe) this kind of disaggregation and reintegration, including chiplets, tiles, and systems in packages.
To this point, Caulfield referred to a recent presentation by the CEO of Intel. “Pat Gelsinger did a good job of explaining tiles — no longer trying to integrate a lot of features that were on traditional CPU chips and offloading to a nearby chip those that don’t need bleeding-edge manufacturing to create that function — I/O chips, maybe SERDES, things like that.”
And those areas are going to need investment too. It’s worth pointing out that the nature of those investments are going to be different. A manufacturer targeting the processor segment is going to spend perhaps 80% of its R&D budget on process development and equipment purchases — “on the on-wafer experience,” Caulfield said, and the balance on process development kits (PDK). The inverse is true for a company like GF.
“We spend the disproportionate amount on PDKs, on IP with partners like Synopsys, Cadence, Arm…, where they’re helping us build out our ecosystem. That’s where the innovation comes from,” Caulfield said.
Caulfield concluded, “I’m in a mission to help the industry to un-paint itself from the corner it’s in, where we’ve convinced the world and ourselves that the only innovation is Moore’s Law,” he continued. “I think that insults tens of thousands of engineers who work on RF, in embedded memory, high-voltage devices for electric vehicles.”
He held his smartphone up to the camera. “If you do secure pay transactions? We make the chip that does that. If you listen to the audio on your phone, we make the chips that do that. The touchscreen? We make the chips. The power management that makes your battery last longer? We make the chip. The complete front-end module? I defy you to find any phone, if it’s 5G-enabled, that doesn’t use a GF chip. All of that stuff takes innovation and it has nothing to do with taking a transistor from five nanometers to three nanometers.”
This article was originally published on EE Times.
Brian Santo is Editor-in-Chief of EE Times. He has been writing about technology for over 30 years, for a number of publications including Electronic News, IEEE Spectrum, and CED; this is his second stint with EE Times (the first was 1989-1997). A former holder of a Radio Telephone Third Class Operator license, he once worked as an engineer at WWWG-AM. He is based in Portland, OR.