There is no secret that AMD probably had the most versatile manufacturing engineers in the industry. Their innovative approach to manufacturing enabled AMD to capture 25% of world-wide CPU market [in the early 2000s] with just a single 200mm wafer Fab, while Intel had several 300mm-wafer Fabs in operation.
As a part of their Asset Smart - Asset Light strategy, which I exclusively uncovered on the previous publication http://www.tomshardware.com/news/amd-corporate-culture,5206.html , AMD spun off its manufacturing division and formed GlobalFoundries, a company whose sole business is to go head to head against TSMC, UMC, Chartered Semiconductor and all other large contract manufacturers.
But, here's a twist - unlike regular start-ups that struggle for cash, GlobalFoundries was founded with $4.5 billion in cash, with future backing going up to seven billion USD, to be spent on building a foundry in New York State [Malta] and continuously improving the Dresden site. In the light of recent announcements of acquiring various staff, we requested to sit down with GlobalFoundries and learn what's going on. Thanks to Jon Carvill, we managed to talk with Tom Sonderman, VP of Manufacturing Systems.
BSN*: Hi Tom, could you please introduce yourself to the readers of BSN* - what is your position in GlobalFoundries and what are your responsibilities?
Tom: I am serving as vice-president of Manufacturing Systems. In the real-world, this means I am in charge of manufacturing operations for GlobalFoundries, making sure our engineering teams are on the same page. I am also working with other divisions to make sure our partners have the right information at the right time.
BSN*: A lot was said about GlobalFoundries, but can you tell us what is the headcount of you guys, now that you are separated from AMD?
Tom: We are 2700-strong company.
BSN*: Out of those, what is the percentage between core-business [engineers] and non-core personnel [administration, marketing etc.]?
Tom: Roughly 10% in based in the USA- the rest is located in Dresden, Germany. Our US staff is spread between California and the New York state, the state of our future Fab 2 complex.
BSN*: Not many people know about your manufacturing team located at IBM's Fab in East Fishkill, NY… what is the headcount there, and what are these engineers working on?
Tom: Today, the IBM site is home to 70 engineers who are working on new process technologies hand-in-hand with IBM engineers. As you know, GlobalFoundries is a part of both manufacturing alliances headed by IBM and these engineers are our contribution to creation of new manufacturing processes and materials, such as the next-generation SOI [Silicon on Insulator http://en.wikipedia.org/wiki/Silicon_on_Insulator].
Given our plans for growth in NY State, we will put more and more manpower in NY. All of our present efforts were with SOI... but now it is likely that we will expand in some other areas, depending on our partnership with IBM. Our wholly-owned facility Albany Nanotech is dedicated to work with IBM on most advanced manufacturing processes. As we all know, nanotechnology starts at 11nm and we are investing so that we can be ready when the time for 11nm arrives.
BSN*: Obviously, any foundry player has to be compared to the industry leader. Just like in the world of CPUs, proven semiconductor manufacturing leader is Intel Corporation. How do you fare against the industry leader as GlobalFoundries?
Tom: We're killing the latency between Intel and us in 32nm. AMD was late with 65nm and introduced 45nm with a nine month delay. With 32nm we are reducing this on just a single quarter or three months. Also, please note that we are implementing this model for a fraction of the cost, thanks to the shared cost model inside the IBM alliance.
Our model is allowing us to bring the high-end mentality forward, instead of low-cost mentality.
We are bringing production back in the US because we need top doctors, top scientists, top everything. Our economy is built upon excellence in every field and this is very important right now because we're in a downturn and other companies are looking into our offering. We are doing the same thing in foundry space as AMD - bringing competition to the market leader, e.g. Intel.
BSN*: Can you update us on the Fab development? Where are you with your internal milestones when it comes to production of customer silicon?
Tom: As you probably know, we have Fab 1 facility in Dresden. The Fab1 Module 1 [ex-Fab36] is currently configured as a 100% 45nm High-Performance SOI Fab production. The move to 32nm SOI is in progress and we're moving closely aligned with AMD. The managing of F1M1 didn't change from the times while AMD was in charge. This facility manufactures most of AMD processors and we are moving forward in order to be ready for 32nm processors [sexa-core Sao Paolo and 12-core Interlagos, Ed.] in the second half of 2010.
The Birth of a super fab...
Fab1 Module 2, formerly known as Fab30 and Fab38 is bringing up the 32nm bulk-technology. We are currently installing the equipment in the F1M2 and adding more units than there were in the past. We are on track to be ready for initial customer designs in 4Q09 with products available in mid-2010. From 32nm bulk, we are going to introduce a 28nm half-node die-shrink, so that our customers can extract enhanced performance from their original 32nm bulk silicon. We will be as flexible as possible to our customers, enabling them to extract maximum performance from available process technology.
As you can read here, we believe that this Mega Fab, with its 50,000 WSPM [Wafer Starts per Month] has the potential for leading edge capacity. At GlobalFoundries, we analyzed what's right and what's wrong in the foundry business and we want to challenge the traditional models. Traditionally, a foundry would start with a low-end process and then move upwards. We decided to do things from opposite end of scale. The world and world's demands have changed.
There are no more "simple" designs. The processors of today feature hundreds of millions of transistors, and in case of multi-core processors and graphics cards in this and next decade, we will start to talk about billions of transistors. In order to be able to produce production-worthy yields on a billion transistor chip, you need to have the best engineers in the field.
With GlobalFoundries, we are bringing leading edge to the market, technologies that only IBM, AMD and our alliance partners were privy to. Our competition plans to bring only 1% of WSPM in 45/40nm, while we want to ramp-up aggressively in 32nm, 28nm and moving on forwards, bring 22nm process in both SOI and bulk silicon flavors to the foundry market.
AMD had to stay competitive with Intel and the company learned how to manage a leading-edge foundry with a fraction of resources. Even with our vast financial capability, we will not stop being nimble and most importantly fast.
BSN*: You mentioned AMD and its arch-rival, Intel. What is the relationship between AMD and GlobalFoundries, now that the deal passed and you operate as a separate entity?
Tom: AMD has an equal ownership with ATIC and they have equal controlling interest. AMD brings the volume, technology and human resources, while ATIC bring "patience-capital", they're the biggest customer. They're the alpha-shareholder in this company. This is a very capital-intensive business and we're really happy with the plans ATIC has in store.
The board is consisted out of four members from ATIC and four from AMD, thus we are on equal terms.
Our partners at ATIC walked in knowing what they are getting into… we are a $4.5 billion start-up with secured seven billion dollars in securities during the next 10 years. We believe that this is a unique and frankly - the best set of circumstances if you want to start a silicon foundry.
Looking at TSMC and others, it takes 15 years to come to leading edge… with our experience and a true world-recognized partnership with IBM, we have all the pieces we need.
BSN*: You mentioned alliances and the technology development. At GlobalFoundries, what alliances interest the company - do you plan to leave any existing alliances or expand your presence?
Tom: For starters, GF is a member of both IBM's Joint Development Alliance for both bulk and SOI silicon. We are the only foundry in the whole of IBM ecosystem to offer both SOI and bulk silicon. We are interested in expanding alliances, but expanding relationship with our future customers.
Our goal is to enable companies to expand their own IP. As it stands right now, you have to have pretty robust design capabilities if you want to keep on pushing the latest process technologies. We decided to hire the best of breed - recently we hired seasoned executives from other foundry companies and their respective partners, building a base of knowledge and management capable of handling future challenges.
I have to mention our APM [Automated Process Manufacturing http://www.globalfoundries.com/technology/apm ] set of technologies here. We are not a company that will jealously keep our secrets and send the invoice to the companies working with us.
We want to give [customers, Ed.] a chance to look at what's best in class and put out a truly 21st century solution that will put GlobalFoundries as the platform of choice for custom silicon. One of our challenges is that we are a truly independent company - we're not a part of conglomerate, but rather a force on our own. We have to deliver our bottom line to our Board of Directors and investors. In difficult times, this might turn to be our largest advantage, because we don't have a mother company to bail out on the expense of our customers.
BSN*: You mentioned APM. Now, in the past, this was considered to be one of secrets that AMD used to create extremely high yields and stay competitive with Intel.
Tom: As you know, we are not afraid to license our technology to partners that are open to innovation. A couple of years ago, Microsoft, IBM and Chartered found themselves in trouble considering low yields on the Xbox 360 CPU. We were asked for help and by analyzing the manufacturing flow, we ended up implementing APM 2.0 in the Chartered Fab in Singapore. The results were instant and the Xbox CPU achieved rewarding yields in mere months after APM deployment.
All APM specifics are now our key selling points. As I mentioned before, we're not an "invoice company". As our customers learn of their own designs we can come in and help them to optimize the designs to achieve highest yields and clocks. Our policy is "Our Fab is their Fab".
BSN*: With the facilities in Germany and this upcoming one in the USA, what do you see as your advantage and how are you going to cover the costs and pricing difference. It is no secret that most of foundries are located in the Far East - do you see yourself as the "black sheep" of the foundry industry?
Tom: [Laugh] We see that we have a big opportunity and the knowledge to overcome challenges that customers are now seeing for themselves. If you take a look at various manufacturers, only a few companies kept pushing on the development of smallest-scale transistors.
We differentiate ourselves with the decision to be in stable locations, staying neutral with both geographic and political conditions in places where our facilities are located. Dresden is a large technology hub and we have excellent cooperation not just with the TU Dresden [Dresden University of Technology http://en.wikipedia.org/wiki/TU_Dresden ], but with other leading universities in Germany. Our upcoming New York facility will benefit from our headquarters in Silicon Valley and cooperation with leading Universities in the US such as the local uni, the University of Albany. We want to become the company of choice for young engineers, solving issues we do not know they exist yet. The next decade will bring a lot of challenges to the semiconductor industry and with the locations in Sunnyvale [CA], Malta and Albany [NY] and Dresden in Germany [home of the autobahn http://en.wikipedia.org/wiki/German_Autobahns ] GF should prove very attractive for young engineers.
BSN*: Where do you stand with Fab 2… when do you expect to break ground?
Tom: Recently, we had an event where we showed the preparation work lead by NY State… we are planning to break ground this summer. But Fab 2 is more than another Fab. With the third generation of APM [APM 3.0], we are linking our factories together and making technologies inter-operable at multiple facilities. We are going to use this for all of our current and future facilities.
BSN*: Coming to future facilities, I believe you plan to have more capacity than Fab 1 M1/M2 and Fab 2 M1. Couple of years ago, I visited Fab complex in Dresden and Mr. Hans Deppe spoke about a possibility of building two more clean rooms. How big can GlobalFoundries actually become?
Tom: There is space in Dresden for other Fabs; there is space in other locations... Dresden and NY State are our potential first candidates for expansion, but they will not exclude any other location. What we are likely to see that turmoil in the countries that are trying to get manufacturing into their locations will bring competition. We might be attracted to [other countries, Ed.], but the ecosystem has to be there. After all, building a leading-edge semiconductor facility is an investment and job security of several thousand highly educated and skilled engineers.
To answer your question, Dresden has room for another clean room, while Malta facility could grow by couple of clean rooms.
BSN*: To sum up the numbers, if GlobalFoundries designs clean rooms ["Fab"] for 25,000 wafer starts per month capacity, does this mean that you could have six clean rooms, e.g. 150,000 WSPM?
Tom: Correct. If our customers will need all that capacity, we'll be happy to provide it for them [laugh].
BSN*: What are priorities for GlobalFoundries at this point in time…?
Tom: Right now, we have two priorities. Our number one priority is Fab 1 Module 2 and its ramp up with 32nm bulk technology. This is a key for swift transition to future manufacturing processes, such as 28nm "half-node" shrink. Our second priority is getting our initial Fab in place in Saratoga County.
Continued on the next page: Manufacturing Technology, Conclusion
Manufacturing technology – 32nm, 28nm, 22nm, SOI, Bulk, LP, SiGe...
BSN*: Moving on to manufacturing technology, let's discuss some of hot issues that surround GlobalFoundries and just about any other foundry business out there. Since GF can provide both SOI and bulk, can you give us more details about the future of SOI? As we all know, some companies have openly attacked SOI technology, while AMD and IBM pushed for it. Did anything change between AMD and GlobalFoundries?
Tom: When it comes to subject of Silicon on Insulator, we leave it up to design community to specify where they need to go. IBM and AMD used SOI, but I would say that we have to wait and see where SOI will develop.
There is a lot of interest in GF by getting bulk manufacturing up and running, and future SOI development will be decided. But to make the matters clear, there are fabless companies that wanted to go SOI, but were unable to find a foundry that can provide them with respectable yields. We have best-in-class yields on SOI wafers and the feedback we're receiving from the potential customers is very encouraging.
According to consolidated industry numbers, SOI will grow from a $500M/year to $800M/year industry in 2010. I don't want to negate the point that bulk technology is important. We have over 30 years of expertise with bulk silicon, before Opteron came - everything we manufactured was bulk.
Trend-wise, we see that RF-circuits will move to SOI in lots of different flavors - by bringing the GlobalFoundries foundry model covering both SOI and bulk silicon, we can now offer top notch manufacturing for all customers that could not order large orders from IBM - we don't want to speculate where the market can go, but our technologies are available for our customers.
BSN*: A quick one - what three key technologies are you offering to the consumers?
Tom: High Performance SOI for CPU & GPU, Performance Bulk for GPUs and Low Power bulk for the wireless chips.
Besides High-K touted by Intel, GF is also bringing new iteration of Low-K, dubbed Ultra-Low-K
I would also add that our wafer manufacturing capabilities cannot be described just with the process that we're using. For instance, we are already using SiGe [Silicon-Germanium] wafers with strained technology – we started with micro-strain at 65nm, then moved to macro-strain with 45nm SiGe and we are continuing to improve SiGe. We are on 3rd generation of the technology, and with 32nm we are also going to introduce both High-K and Ultra-Low-K technologies.
BSN*: What is the level of assistance you offer to your customers?
Tom: We can assist with to every customer design, not only for AMD, but to all manufacturers that are interested in it - we have proven our expertise with console CPUs, System-On-Chip designs.
BSN*: With 45nm process, AMD and IBM introduced immersion-lithography. What are any possible future developments?
Tom: Currently, We are on our 45nm using immersion-litho. We are also the biggest supplier of immersion technology on the market, unlike TSMC who was the inventor of immersion-litho. AMD has the history of firsts, just to mention Low-K, Copper interconnects, SOI… we are not first with High-K, but we are bringing something special to the table. Our 45nm process results in significant improvements over industry-standard 45nm process. As the recent success of Phenom II says, GlobalFoundries 45nm results in high clocks and high yields.
At GlobalFoundries, the 32nm process is available in two flavors
At 32nm we definitely see more advantages of High-K materials and we want to make sure we ensure the highest performing technology at high yields. That model proved beneficial in the past with AMD and we'll continue on the same path. We are also introducing a 32nm Ultra-Low-K material to address the other spectrum.
A couple of years ago, while the 130nm process was all the rage, all factories started to experience problems with the yields [percentage of correctly working chips on a wafer]. We perfected the technology and probably had the best-in-class and now working on developing techniques by reducing the amount of silicon needed to bring the technology to market.
Our competitors should be worried because we are attacking from the high-end, not from the low-end. We have leading edge technology. Being in alliance with IBM, we're looking at many ex-techniques in 22nm, 18nm technologies. We're really not disclosing much beyond our high-k efforts, and strained technologies. Just like everybody in the industry, we expect that Sin-Tec technology; 3D transistors will improve the overall performance and availability.
BSN*: You referenced to GPU technology several times. What are you qualifying for GPU production?
Tom: We're qualifying 32nm technology to be GPU-friendly, and customers in the industry are interested in our manufacturing roadmap. We see 32nm and 28nm as the point when we're getting into GPU side of business.
GlobalFoundries predicts a seamless transition between 32nm and 28nm, something that GPU manufacturers will appreciate.
We make all of AMD's products, but for GPUs, we have to go and compete with others. We are global and we're bringing tech that is serving not just the IBM ecosystem, but all the Top20 GSA companies.
BSN*: Tom, I wish to thank you for this extensive interview. We wish you all the best.
Tom: It was my pleasure.
There you have it folks… unlike AMD that was known for keeping a low profile and not coming out with bullish statements. This time around, GlobalFoundries simply doesn't care about being politically correct. These guys know that they have enough Middle Eastern capital to bring back jobs to the American continent and nobody can call them out for that.
Technology wise, GlobalFoundries will bring 32nm this year, volume production in 1Q 2010, with launching a 28nm half-node shrink just a quarter later [1H 2010, but we can say 2Q'10]. With all respect to high profile customers such as AMD, ATI and probably nVidia, but the key battle now is to take customers such as Texas Instruments, Analog Devices, Xilinx, Altera etc.
The business of chip manufacturing just got very, very interesting.