Who gets what, why it’s so important, and what questions remain unanswered.
GlobalFoundries’ proposed acquisition of IBM Microelectronics is the kind of deal that will have business schools talking for many years to come—a gargantuan combination of expertise and technology, built on the back of high-profile business successes and failures, long-running legal struggles and global politics—with far-reaching implications for all parts of the semiconductor supply chain.
The deal is so enormous in its reach, potential, as well as inherent risks, that even GlobalFoundries will need months to fully comprehend all of the pieces it is acquiring, where the possible synergies might be, and how it will impact everything from architectures to IP to EDA tools to packaging and manufacturing. But while this combination still must win regulatory approvals from a number of different government agencies around the globe, which no one is expecting to happen until at least next year, the planning has been underway for months. And alongside of all of this, there are plenty of questions.
“It’s going to take awhile to digest,” said one industry source at a chipmaker that works with all the major foundries. “The big concern for us is whether they eventually become an ASIC supplier instead of a pure-play foundry and turn into a quasi-IBM. We’re still waiting to hear a clarification on that. GlobalFoundries was a spinout of AMD fabs, so it was never a merchant pure-play foundry. Chartered was pure-play and they were very good at manufacturing. But when you put all the pieces together, it could be a tremendous money maker.”
In the right hands—a nimble market player with strong ecosystem relationships—there is an opportunity to generate more money that IBM was ever able to do with a merchant IC business. For 46 years, IBM was handicapped by the now infamous 1956 consent decree, which the company signed with the U.S. Department of Justice after being found guilty of anticompetitive business practices for bundling services, software and hardware in any configuration that would undercut competitors. While IBM continued to sell all three for its highest-margin business following the decree, it pursued lower-margin, higher-volume markets such as PCs and Unix-based servers in fits and starts, eventually selling its PC unit to Lenovo. The company largely ignored the mobile device market in favor of the infrastructure for what it termed “pervasive computing.”
“This looks like a win for both companies,” said Charles Janac, president and CEO of Arteris. “IBM has been exiting their component businesses for years in order to focus on systems and consulting, so this deal allows them to get out of a capital-intensive and unprofitable microelectronics business. In terms of GlobalFoundries, this allows them to leap into one of the leading positions in the foundry business as well as acquiring a proprietary ASIC design business and being the continued sole supplier to IBM/s mainframe/server business. They get two competitive fabs in Burlington (analog, mixed/signal and specialty processes) and Fishkill (digital) and lots of top notch technical resources. The $1.5B payment doesn’t hurt either.”
IBM will pay GlobalFoundries $1.5 billion to acquire its semiconductor business—handing over a huge sum to take the business off its hands and even more to liquidate redundancies internally. In addition, IBM will continue with its $3 billion advanced semiconductor research, to which GlobalFoundries will have access. The dollars changing hands in this deal are somewhat deceptive, though. It requires an investment by GlobalFoundries over the next decade—the time during which GlobalFoundries will continue to be the sole source for IBM’s processors down to 10nm.
That’s what has been reported so far, but there is much more to this deal. To begin with, IBM has some of the deepest semiconductor design, manufacturing and packaging expertise on the planet—broader and deeper in some areas that Intel—and it plans to continue reaping benefits from that technology for at least the next decade. It has expertise in stacked die (both 2.5D and 3D ICs), as well as silicon photonics, microfluidics; silicon on insulator (SOI); silicon germanium (SiGe); and RF SOI.
The SiGe and RF SOI are manufactured at its fab in Essex Junction, Vermont, which has been space-constrained for some time. RF SOI is an important technology for analog switches, while SiGe is important for power amplifiers, but IBM’s market footprint didn’t justify expanding the Vermont fab—particularly with the recent uncertainty about electricity rates in Vermont caused by the closing of the state’s nuclear power plant a decade ahead of schedule. GlobalFoundries, meanwhile, has the capability to produce both of those technologies in Singapore, a facility it acquired with the purchase of Chartered Semiconductor.
“We didn’t have SiGe—IBM brought that—but we do have a footprint in Singapore for RF SOI,” said Paul Colestock, senior director of segment marketing at GlobalFoundries. “But it’s not just the technology they’re bringing. They also have a very large set of customers for their own technologies, which is a big boost. It’s nice entry point into a business where we have not had a lot of progress—or at least not to the extent that they have.”
IBM also brings a patent portfolio that contains 10,000 patents in the United States alone, and more internationally. The question is what GlobalFoundries can do differently with this technology than IBM, which now views its semiconductor business as something of an albatross.
“The key is how GlobalFoundries’ implements all of this and what they do with it,” said Joanne Itow, managing director for manufacturing at Semico Research “IBM has great technology, but they haven’t always had success with it in terms of applications. Their technology is so complex that sometimes it’s hard to get to the market quickly.”
IBM was one of the original integrated device manufacturers. Unlike other high-tech companies, it draws from roots that date back to the early 1900s when companies controlled their entire supply chain. GlobalFoundries, in contrast, has managed an ecosystem of companies, from IDMs to fabless companies to IP and EDA companies. And while that minimizes the overlap and provides enormous upside potential, merging these disparate cultures isn’t so simple.
“The good news is that GlobalFoundries is picking up a strong technology partner,” said Mike Gianfagna, vice president of marketing at eSilicon. “TSMC has been killing it on process and technology, but as a customer we like to have strong options. We used to compete against IBM in the ASIC market. They’re very good at what they do. So from eSilicon’s point of view, this is a great move. Now there’s another strong competitor in the foundry business.”
Open-Silicon, meanwhile, believes the IBM acquisition will create new business possibilities with GlobalFoundries in the SiGe market, where IBM is currently the market leader in technology.
“IBM has interesting niche processing technologies that will open up new opportunities in the second- and third-tier markets,” said Dave Krishna, senior vice president of sales and marketing at Open-Silicon. “We’ve had SiGe opportunities in the past for packaging and test where the customers used IBM foundries. This may allow a GDSII handoff for things like high-speed ADC (analog-to-digital converter) and DAC (digital-to-analog converter) technology.”
Krishna noted that niche IP could open other opportunities in the overall ASIC market, as well.
GlobalFoundries’ Colestock affirmed the company’s commitment to working with both eSilicon and Open-Silicon. But the combined IBM/GlobalFoundries also will develop turnkey chips in the high-end ASIC business or something more collaborative. “We will not sell ASSPs, but we will address the wired communications, wireless communications infrastructure and storage markets.”
Included in that are enterprise and service providers switches and routers, high-performance SoCs with embedded memories, chips for mobile base stations, network controllers and gateways, and storage networking infrastructure switches, directors and adapters.
Colestock noted that FPGAs and graphics processors were the bread and butter of the foundries until 28nm, when they adopted a “mobile first” strategy. That leaves many other opportunities around the edges, and that’s one of the benefits opened up by the IBM acquisition.
EDA and IP opportunities
While none of the Big Three EDA companies would comment on the pending merger, numerous insiders at those companies said they believed there is a potential for a significant uptick in tools sales. Like Intel, IBM is a major user of EDA tools. But unlike Intel, IBM relies heavily on internally developed tools. The company invented many of the processes it used, and it created tools to automate some of the design tasks used in the processes.
A sale could throw open the market even within IBM for commercially developed tools and IP—standard, custom and configurable—to help lower costs and improve time to market. How that will play out is unknown, even within IBM. But the fact that there are unknowns is a good thing for third-party toolmakers.
At this point, no one outside of IBM has intimate knowledge of what IP will be included in the deal. Most of IBM’s IP, however, is likely to fall into areas that are outside of the standard, custom, and configurable IP developed by Synopsys, Cadence, Mentor Graphics, as well as ARM and Imagination Technologies. But all of it is a big boost to GlobalFoundries, which was considered well behind TSMC in IP, leaving GlobalFoundries in a race to catch up in that area over the past few years.
One area where IBM has really been pushing the envelope is at 14nm, where it will roll out its first finFET based on partially depleted SOI. At 14nm, GlobalFoundries has been focusing on bulk CMOS, along with Samsung and Intel. TSMC is using bulk at 16nm.
“With the IBM chip assets, Global Foundries should be able to compete for some of the largest foundry deals out there,” says Arteris’ Janac. “They become truly global. The challenge here is to make this much larger business work operationally. They have to make IBM Microelectronics profitable and they have to rationalize a complex portfolio of facilities, organizations and processes. Can they serve the largest ASIC customers cost effectively and reliably at this new level of business? It’s very promising but quite a challenge.”
What isn’t so clear is how various government customers will react to the deal. IBM was a known entity with a global footprint, but its manufacturing operations were in New York and Vermont. That put it on the safe buy list for many government agencies, both within the United States and among the U.S. government’s political allies.
GlobalFoundries is headquartered in California, but it is owned by the Emirate of Abu Dhabi through its Advanced Technology Investment Co. AMD had owned a portion of the company, but sold its final stake in GlobalFoundries to ATIC in 2012.
While that is not likely to disturb any commercial contracts—GlobalFoundries is a well-known foundry with deep technological prowess—governments around the globe have restrictions on where they can buy and distribute certain components for security reasons. The current deal is still being scutinized by the United States and other governments. Aside from defense contractors and smaller specialty foundries, the only other large U.S. chip manufacturer with its own fabs is Intel—a direct rival to GlobalFoundries in the foundry space.