These are old articles, nevertheless, they provide interesting insights.
I guess it is not easy for AMAT in managing its client interests - afterall AMAT does know the "secrets' of its clients - and its clients are relying on AMAT's products and services to outperform each others - not a bad problem to have though.
Wondering if the "discount" given by UMS to AMAT in 2012 has to do with "rebate" to Samsung - haha !
(vested)
_____________________________________________________________________________________________________________________________________________________
Beyond 22nm: Applied Materials, the unsung hero of Silicon Valley
By Sebastian Anthony on December 1, 2011
In the shadow of major semiconductor manufacturers, a little-known but crucial company called Applied Materials has plied its trade for almost 45 years. Inside every Intel, GlobalFoundries, and TSMC silicon chip foundry, there are Applied Materials machines performing some of the finest handiwork known to man. It isn’t just chips, either: Applied provide the systems that enable Samsung and LG to make LCD displays, and Suntech and JA Solar to create photovoltaic solar power cells. Basically, if a company deals in silicon wafers, it’s almost guaranteed that Applied Materials equipment governs most or all of the fabrication process.
“The vast majority of chips in operation today have been through an Applied’s machine at least once. This includes memory, logic and other types of chips produced by integrated device manufacturers and foundries,” CTO Klaus Scheugraf told us in an interview.
To put a finer point on it, when Intel announces that it has reached 22nm, it has done so using Applied Materials equipment. This isn’t to say that Intel’s achievement is any less awesome, but it’s important that we draw a line in the sand before we go any further. Applied Materials provides the tools, but it is down to Intel to make the most of them through chemistry tweaks and chip design. It’s also important to note that AMD, Samsung, and TSMC all have access to the same tools. Applied Materials works closely with manufacturers to customize their solutions, but more on that later.
How is a silicon chip made?
If, like me, you thought that making a CMOS circuit is mostly a matter of lithographically etching paths onto a silicon wafer with a laser, you’re in for a surprise. Yes, lithography is a vital part of chip fabrication, but with modern processes (45nm, 32nm), and especially with new “3D” FinFET (22nm) designs, it is really just one step out of hundreds. Lithography is like laying down the concrete foundations of a skyscraper.
Geek.com has a fairly comprehensive guide on how a chip is made — from melting down sand, to etching, to testing, to binning — but basically, this is what you need to know: Once the chip’s pathways have been lithographically etched, the transistors and copper wire interconnects are grown using a combination of electroplating, ion implantation (doping), chemical vapor deposition, atomic layer deposition, and more. Furthermore, most of these processes can be broken down into tens of steps: ALD, for example, builds up the layers of a FinFET transistor atom-by-atom.
It’s not like all of these processes happen inside a magical, all-in-one, Willy Wonkaesque machine, either: To create a single FinFET transistor, for example, Applied Materials provides no less than seven different machines. The silicon wafers must be moved between each of these machines, usually inside an extremely tight vacuum. When you’re dealing with high-k dielectric layers that are just 10 atoms thick, a single atom of a contaminant can ruin the chip.
To top it all off, in a major foundry, this entire process will be almost completely automated — using software provided by Applied Materials, of course. Just so you get some idea of the scale of the process, too, bear in mind that a 300mm fab, as operated by the likes of Intel or TSMC, costs in the region of $4 billion to build. Each Applied Materials machine is around the size of an office desk or large chest freezer, and costs between $2 and $6 million each. Taking a median price of $4 million, that means there could be 1,000 separate machines in a large silicon foundry — all located within a cleanroom that is probably acres in size.
Tweaking
If every semiconductor manufacturer uses the same Applied Materials hardware, then why is there such a huge difference between the processes used and the chips produced? Why is Intel pushing ahead with 22nm FinFET chips, while GloFo still languishes behind with 32 and 45nm SOI?
The reason is twofold: First, as we alluded to earlier, Applied Materials provides the tools, but it is down to the chip maker (Intel, AMD) or the foundry (TSMC) to decide on the exact processes that are used to create a silicon die. Intel might have a six-layer process, while AMD only has five. Intel might have discovered a variant of hafnium oxide that provides a better high-k dielectric than TSMC. To this end, every chip maker retains some of the best CMOS chemists in the world to secure an edge over the competition. It is their job to really understand the specifications and limitations of Applied Materials machinery, and cajole the best chips out of them.
Second, while Applied Materials equipment is generic — you could buy a full set for $50 million and produce your own 45nm chip today! — the company works with individual manufacturers to tweak each of the machines. If Intel needs a nozzle changed or an piezeoelectric actuator upgraded, Applied Materials is happy to oblige. Like a souped-up car with go-faster stripes and fatter rims, each machine is fundamentally the same but customized to the exact needs of the client.
The great divide
In short, then, Intel couldn’t reach 22nm without working very closely with Applied Materials’ nanoengineers and chemists — but likewise, Applied also knows the deepest, darkest secrets about AMD’s latest chips. As you can imagine, this is potentially a huge conflict of interest. What if an engineer accidentally gave GloFo Intel’s FinFET secret sauce?
To get around this, Applied Materials actually has separate teams that work with each of its big clients — and an internal network with a very stringent permissions system in place. In theory, an Applied-Intel engineer never sees the Applied-AMD files, and vice versa. “We have strict policies and procedures designed to protect our customers’ confidential information. Protecting customer’s intellectual property is critical to our business culture,” Scheugraf told us, though we didn’t find out whether this segregation is enforced in the physical world as well.
Does Applied maintain separate cafeterias and common rooms for each of its teams? Is their Santa Clara campus split in five, with lookout towers, razor wire, and death strips between the buildings? We may never know. Considering Applied has been doing this for more than four decades, though, it presumably knows what it’s doing.
With this level of confidentially, foundries and Applied can enter into billion-dollar capital equipment and cross-licensing agreements that bring to bear their full, combined arsenal of corporate secrets and patents. It is only then that the first brood of 22nm, or perhaps even 14nm chips, are born.
To 22nm, and beyond!
While we’re certainly pushing the limits of silicon-based CMOS — the gap between two silicon atoms is 0.5nm — multiple patterning, immersion lithography, and 3D Tri-gate (FinFET) transistors mean that chips built with 16 and 11nm processes should be possible within the next few years. As the transistors and interconnects get smaller, so do the tolerances — and the cost per wafer, as you would imagine, increases. “If we look at the foundry transition from the 45nm node to the 14nm node, their overall spending per wafer is almost doubling in this transition due to increasing complexities associated with the three dimensional nature of the transistor, added steps for the interconnect, the increased use of double patterning and the advanced wafer-level packaging occurring at the advanced nodes,” Applied’s CTO told us.
Even so, Applied is doing its bit to increase yields and drive down costs. In a modern CPU there can be more than 60 miles (100km) of copper wiring, and through leakage and heat resistance it increases the total power consumption and heat dissipation of the chip by around 30%. Just this week, Applied Materials announced the Onyx System, which uses an atomic-level treatment to strengthen the dielectric (insulator) around those 60 miles of copper wire, increasing the number of usable dies (yield), and reducing overall power consumption of the chip by 3% — not bad, when one of the most important metrics in today’s computing climate is battery life. Scheugraf told us, incidentally, that some of the big players are already using the Onyx System on their new chips.
So there you have it: Applied Materials, the unsung hero of Silicon Valley that has been guiding the bleeding edge of CMOS since before Intel released its 4004 CPU in 1971. Judging by its entrenchment, expertise, and technology portfolio, Applied is in good stead to see us through to the physical limits of silicon at 11nm… and beyond, to graphene!
http://www.extremetech.com/extreme/10689...con-hero/2
_________________________________________________________________________________________________________________________________________________
Samsung Electronics, Applied Materials Settle Row Over Leaked Chip Secrets
By Jun Yang and Victoria Batchelor
Nov 30, 2010 2:24 PM GMT+0800
Applied Materials Inc., the world’s largest maker of chipmaking equipment, agreed to sell gear to Samsung Electronics Co. at a discount for three years as part of an accord to avoid lawsuits related to industrial espionage.
Santa Clara, California-based Applied Materials will provide “volume-based rebates” on purchases of semiconductor products by Samsung and its affiliates, the U.S. company said in a statement yesterday. The agreement started Nov. 1.
South Korean prosecutors charged 10 employees of Applied Materials in connection with leaks of Samsung’s chip-making technologies to smaller rival Hynix Semiconductor Inc. through the equipment maker for more than four years, the Seoul Eastern District Prosecutors’ Office said Feb. 3. No charges were brought against the companies. Applied Materials said in February that its former South Korea head was among those indicted.
Samsung has “agreed to reconcile with Applied Materials related to technology leaks,” Nam Ki Young, a spokesman at the Suwon, South Korea-based company, said by telephone, declining to provide further details because they’re confidential.
The agreement does not affect criminal charges against individuals, including the ongoing proceedings against the current and former employees at Applied Materials’ South Korean branch pending in the Seoul Eastern District Court, the statement said
http://www.bloomberg.com/news/2010-11-30...onics.html