Globalfoundries is a leading pure-play foundry, headquartered in New York, but with a wafer fab in Dresden, Germany. That campus forms an anchor for the Silicon Saxony initiative and it could be a key part of Europe’s plans to re-invigorate its semiconductor fortunes in pursuit of strategic independence.
We started by asking Caulfield about the European politician’s proclaimed goal of doubling Europe’s chip manufacturing capacity to 20 percent of global output by 2030 when currently its share is said to be down at 6 or 7 percent by some analysts. Others have pointed out that Europe’s share of the more advanced chip manufacturing on 300mm-diameter wafers has been less than 1 percent of global output (see Europe’s 300mm IC manufacturing falls below 1% of world output). With the continent’s share on a downward trend surely 20 percent is an unrealistic goal?
“There are plenty of examples from history where it has been important to set ambitious goals. The most obvious is JFK when he said the US would put a man on the moon within a decade. The most important thing is it produces a change in mindset. You have to change what you are doing. It forces a different paradigm and then you get rid of the blockers,” Caulfield said
“If Europe gets to 17 or 18 percent by 2030 no one is going to say it has failed. They’ll say things are going in the right direction. It is ambitious and the world is not standing still. So it will be 20 percent of a much increased global manufacturing capacity,” Caulfield said. He added, that given that so much semiconductor manufacturing investment is going to be made why not make a significant part of it in Europe
In the last few years Globalfoundries has benefited from a funding boost at its wafer fab in Dresden by way of an Important Project of Common European Interest (IPCEI) on microelectronics. There is much discussion of a second one from which Globalfoundries would likely benefit.
Globalfoundries is investing in the current fab at Dresden – $1 billion over the next two years – but Caulfield has indicated that building another significant wafer module in Dresden would be possible, dependent on investment support (Globalfoundries moves towards next Dresden wafer fab). “Globalfoundries makes the lion’s share of any investment, but we need government and partners to make it make economic sense,” Caulfield said. “Without government partnership and the right investment, those ambitious goals don’t stand a chance.”
There is also the issue that sometimes with government initiatives the spending gets sliced thinly between scores of worthy recipients. Clearly for an advanced 300mm wafer fab module the support has to be of the appropriate scale.
Having discussed Globalfoundries’ potential role in boosting volume chip production in Europe and its tentative plan to build a wafer fab module to share the Dresden infrastructure, we asked Caulfield about Europe’s goal of developing or attracting a leading-edge manufacturer of chips at the 2nm node.
“Europe shouldn’t worry over issues of technology leadership for two reasons. One: You can’t do anything in semiconductors without lithography and Europe has ASML the leader in lithography. Nobody can do anything in semiconductors without giving capex to ASML, so Europe has great control of the semiconductor industry.”
Caulfield prefixed his second reason by pointing out that the industry has moved beyond the point where you moved from node to node to obtain the greatest performance per dollar in personal computer dominated compute-centric electronics. In the subsequent era of the smartphone there was an explosion in the diversity of semiconductor needs including RF circuits, touch-screen displays, sensors and so on. “Our industry went from computer-centric to pervasive deployment. The chip companies are not short of single-digit nanometer chips, it is the diverse types of chips they are short of. It is no longer Moore’s Law with two ‘O’s, it is More with one ‘O’,” said Caulfield.
Next: 2nm or not 2nm
Caulfield continued with his second reason for Europe to be cheerful: “In the context of Europe, it has FDSOI.” This is the fully-depleted silicon-on-insulator manufacturing process used by Globalfoundries in Dresden under the process name 22FDX. “FDX is a major European technical achievement. The wafers are made by a French company Soitec. Development was done by STMicroelectronics and Globalfoundries.”
But nonetheless European politicians talk of obtaining 2nm process manufacturing capability. Are they wrong?
Caulfield responded: “Shrinking transistors is a large and expensive challenge. Is it necessary? Well, the use cases diminish over time. At each node, less customers participate.”
But Caulfield acknowledged that pursuing access to the leading edge is what the US is doing.
“There are a lot of perspectives. With the pandemic and shortages politicians have spent more time trying to understand semiconductors. The US is taking a holistic approach. They need single-digit nanometer processes, but also memory and storage, and analog and mixed signal from the likes of ADI [Analog Devices Inc.] and Texas [Instruments].”
“For Europe to get to 20 percent it needs to cover its bases. Domestic chip manufacturing should cover the most important industries; that’s industrial and automotive in Europe. Focus on 28nm, 12/14nm, 40nm, the things you use today and then, maybe, also move to single-digit nanometer.”
But what about the idea that the single-digit nanometer process of today becomes the “workhorse” platform process of tomorrow, supporting diverse needs at greater transistor density and integration? If you are slow to the current leading-edge, will you be slow to the next broad-based technology?
Caulfield said: “It’s a false narrative. The node becomes the platform when you add blocks of intellectual property. We invested in 28nm and that is partly where 22FDX came from.”
“Will everything get swept down to single-digit nanometer? No. The benefit is just not there. That’s why there are fewer and fewer customers at finer geometries. Also they [single-digit nanometer technologies] do not lend themselves to RF, high power, low voltage, which are getting embedded in established process nodes. That’s what’s causing some of the stalling of Moore’s Law. Some of the most profitable companies in semiconductors today are manufacturing at 65 and 45nm and they have no intention of moving.”
Next: Nothing to it
Nonetheless the Europe Commission is seemingly trying to entice Intel and TSMC to bring advanced process manufacturing to Europe
“I am not saying Europe should not have it [leading-edge capability] but it’s only 30 percent of the market. A disproportionate amount of energy and effort have been put into that,” said Caulfield.
There are reports that Globalfoundries’ and Intel’s interests could become aligned by a merger (see Intel in discussions about buying Globalfoundries). Caulfield is absolute in his denial. “There’s nothing to the rumor. Our shareholder is very proud of this asset,” he said. After 12 years under the ownership of Mubadala, the sovereign wealth fund of Abu Dhabi, there is a plan for an initial public offering of shares in Globalfoundries. “There is a difference between owning 100 percent of an asset and owning a share in an asset. It is a question of balance,” Caulfield explained.
Caulfield sees the industry’s move to a chiplet-style assembly of components as a confirmation of the limited applicability of leading-edge silicon processing.
“Multichip modules [MCMs] have been around since I began in this industry. Any time you could integrate monolithically you would sweep up the functions in the MCM, because it’s cheaper. Now those economics don’t work,” he said. It started with the high-end processor makers Intel and AMD, he added. “The situation is really telling you single-digit nanometer is not leading to broad-based applications. You need to surround the [processor] with additional process technologies,” Caulfield said.
Next: Campus strength
Nonetheless, it seems the United States, China, Japan and Europe all want to catch up with Taiwan and South Korea with leading-edge capability. And yet Morris Chang, founder and past CEO and chairman of leading foundry TSMC, has warned against different regions of the world trying to achieve self-sufficiency (see TSMC founder warns ‘on-shoring’ chip manufacturing could backfire).
“You have to take a holistic approach. If you can make the wafers but they all have to go elsewhere to be packaged it can become a bottleneck. You have to consider the end-to-end capability,” said Caulfield. “But politicians are realizing semiconductors are the new oil. The whole economy depends upon it. It’s too strategic for society and for nations to leave it to be concentrated in one place.”
Caulfield said he agreed with Chang and TSMC on the policy of building wafer fabs in campus clusters. “We have this footprint in Dresden. The fastest way to increase capacity is to expand on an existing footprint. It can make an 18-month to two-year time difference starting in a new location.”
This is something to be born in mind when considering the potential arrival in Europe of TSMC, which would most likely need to create a new campus infrastructure. Intel also seems to be indicating it will go for a new German site rather than expand at its site in Dublin, Ireland (see Intel in talks with Bavaria over wafer fab location).
So will Globalfoundries be building a new wafer fab in Dresden?
“If and when IPCEI-2 comes to pass, absolutely,” said Caulfield. “We’re ready to do our part to boost the European economy.”
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.