by Anton Shilov
07/12/2006 | 09:32 AM
All the companies in the emerging industries begin with very relatively specific products. Eventually, those specific products may become popular and widespread, or may not. Some of those companies become conglomerates, like International Business Machines, some become nothing. Those, who become successful in their primary markets, usually begin to address adjacent markets, those, who are not, begin to concentrate on one type of products and may either stagnate for years, or cease to exist.
Intel Corp.’s concentration on platforms proves that the company is enough successful on its core microprocessor markets, however, has capacity and ability to deliver products for new markets, challenging different rivals and achieving different goals. Recently the world’s largest chipmaker vowed to invest $900 million into a provider of wireless Internet access, a move, which will help the company to deploy many WiMAX networks in the
ATI Technologies, which started as a maker of graphics cards, but also used to produce communication equipment (modems) and other products, eventually gained ground in the markets of consumer electronics, handheld products, chipsets and game consoles.
Advanced Micro Devices used to clone x86 processors in the eighties and the nineties, but then had been gradually increasing forces in the CPU biz in the late nineties, just in order to fight the performance crown from Intel in 2003 and quit flash memory business in 2005.
Nvidia Corp. has always been the supplier of graphics processors, but in 2001 it started to produce its core-logic devices and in 2003 acquired MediaQ, a producer of graphics processors for handhelds. While Nvidia has been very aggressive on the markets of GPUs, chipsets and even has managed to win a couple of game console deals, it has never tried to seriously penetrate other markets, such as consumer electronics with dedicated products.
Recently X-bit labs has learnt from industrial sources that Advanced Micro Devices is going to acquire ATI Technologies, thus, to change the strategy and market behaviour of both companies. The rumours concerning the possibility began to circulate in early June, when Apjit Walia, an analyst with RBC Market Capitals, wrote in a note to clients that he would expect AMD to acquire ATI Technologies, as he thought it was consistent with AMD’s plans to expand capacities. The sources of the information may be reliable and may be not; the information may be correct and may be not. In this short editorial we will try to summarize the known facts and investigate on the matter of causes and consequences of the deal, which should be considered potential just now.
Intel Corp, the world’s main maker of microprocessors, has always been very aggressive across all markets, trying to become a leader in any segment. But the company has been very pragmatic too: it decided to leave the market of standalone graphics chips in 1999 and it basically exited the flash and special purpose XScale processors businesses in 2006.
But Intel knows that without further development there comes stagnation and, according to sources familiar with the matter, is going to either enter the market of discrete graphics processors, or, at least, put nearly extreme efforts into the development of built-in graphics cores to make them much more competitive against discrete solutions.
Intel 740 graphics card from Legend QDI
Intel has been re-creating itself from a chipmaker into a provider of platform solutions for 1.5 years now and even changed its logotype to emphasize that Intel is not a provider of something “Inside”, but everywhere. To highlight the changes and to stress its rapid technology development, the world’s main producer of microprocessors introduced “Leap Ahead” motto and campaign early in 2006. In addition, Intel started to enter the market of consumer electronics by bringing-in Viiv platform, which effectively takes Intel brand-name to living rooms.
By the end of 2008, Intel may have 7 fabs producing chips using either 45nm, 65nm or even more advanced process technologies using 300mm wafers. While Intel has not provided any guidance regarding its manufacturing capacity growth or actual manufacturing abilities (thousands of wafers per month), 7 fabs is a formidable amount of high-tech manufacturing facilities, which will allow the Santa Clara, California-based chipmaker to fabricate astonishing quantities of CPUs, core-logic sets and other chips.
All in all, Intel is trying to sell as many chips for one computer as possible: a processor, a core-logic set, a wireless network controller and so on. Why not produce a discrete graphics processor and enter a rather lucrative market of graphics cards?
Even though nothing is known about Intel’s possible strategy on the market of discrete GPUs, entering this market might allow the company to achieve several goals:
What the world’s largest chipmaker – and other companies in the industry – knows for sure, is that the market of discrete audio processors has been essentially destroyed by Intel, when its integrated audio capabilities reached the level that satisfies casual consumers. While the demand towards the number of discrete graphics processors gradually rises, their market share shrinks and at some point in future developers of high-end discrete graphics processors may find themselves in a tough situation: they may sell very advanced chips, but they will spend too much on the development of technologies used in them. The latter means that those companies will be balancing between profits and losses, like the creators of luxurious sport cars today, for example, Bugatti.
Certainly, designers of graphics processors may develop their chipsets and chips for other markets, consumer electronics, handhelds and so on, but this means that they must refocus their companies and pay less and less attention to the discrete graphics chips businesses.
The main question right now is whether it makes sense for Intel to enter the discrete GPUs market or not. The answer is yes, provided that Intel will be able to offer a lineup of discrete processors, if not, the company would only be able to capture very niche markets among its devoted partners. Certainly, if the company launches its discrete GPUs, it will have the aforementioned advantages. On the other hand, there is always a threat that the market of standalone GPUs will erode and while R&D investments will not vanish, profits will be negligible.
Advanced Micro Devices has a history of producing dynamic random access memory (DRAM), flash memory, special purpose processors, clones of Intel’s processors and acquisitions of third-party designers to boost its CPU division.
Until late in the nineties, AMD had usually been a step behind Intel, as it either cloned its chips, or used micro-architecture not as technologically advanced as Intel’s one. But everything changed in 1999, when AMD introduced its Athlon processor, which was faster compared to the Pentium III in loads of applications. It took Intel nearly three years to recapture performance lead with its Pentium 4 “Northwood” processor in 2002, but AMD introduced its Athlon 64 chip in late 2003 and managed to grab the performance crown for three more years, which allowed the firm to significantly improve its profitability.
But the other side of AMD success on the microprocessor market is the company’s full focus on CPUs: the firm quitted flash memory business, the company ceased development of chipsets and the firm sold its special-purpose Alchemy processors division. AMD’s usual rhetoric concerning its concentration on the CPU biz have so far been the claims that the AMD64 processors fit perfectly into eco-system developed by third parties.
“AMD has long been a leader in providing solutions for networked computation and communications. Going forward, our efforts will focus on providing communications-enabled solutions at the personal computer platform level , where we can leverage our chipset design and systems expertise and the AMD Athlon processor. The interests of our customers, shareholders and employees will be best served by our action,” said W.J. Sanders III, chairman and chief executive officer of AMD back in 1999, when the firm announced sell-off its communications division.
Jerry Sanders, AMD's chairman emeritus
Mr. Sanders stressed that his company could offer the same functionality as the communications products division with the chipsets, but his successor – Hector Ruiz – held different opinion and after the AMD 8000-series chipset was released in 2003, the company has generally withdrawn from the market of core-logic solutions.
Mr. Sanders indicated back in 1999 that the company does want to be a provider of computer platforms and a creator of advanced chipsets. Currently the firm cannot actually afford it: manufacturing capabilities of AMD are generally limited and when the situation gets significantly better, the company will be in a situation when it has no expertise in developing advanced core-logic designs and has no abilities to offer a platforms for its customers due to the lack of any non-CPU technologies.
While currently AMD enjoys being only a maker of microprocessors, the company plans to rapidly increase its manufacturing capabilities.
AMD's fab 30 in Dresden, Germany
By 2010 the Sunnyvale, California-based chipmaker will have three foundries producing chips on 300mm wafers with total output up to 67 500 wafers per month. Considering that 300mm wafers are 2.25 times larger compared to 200mm ones produced today at Fab 30 (30 thousand per month), by 2010 AMD’s manufacturing capacities will be about five times higher compared to today’s levels. According to IDC, in 2006 there will be 230.2 million computers sold, while in 2010 – the shipments will reach 333.7 million, about 45% more compared to today. At the same time, it is hard to believe that AMD will have five-fold increase of processors shipments by that time: either AMD maintains its current market share and the computer market grows five times (not 1.5 times anticipated by IDC), or AMD starts to rapidly gain presence in the market of microprocessors (all the way to about 75% by late 2010).
It is a known fact that modern foundries should be loaded nearly fully to stay profitable. Therefore, increasing capacities very rapidly either means that the firm plans to wipe-out Intel out of the CPU business in several years, or follow the strategy outlined by Jerry Sanders several years ago and sell “computing platforms”, something, which would allow the firm to sell more chips in general and load the fabs to the maximum level in particular.
Even though the recent success allows AMD to make acquisitions and build fabs, the company is begin to facing the threat of Intel’s Core 2 micro-architecture, whereas Intel’s rapid transitions to newer process technologies may leave AMD significantly behind: if Sunnyvale, California-based maker of CPUs starts to volume production of its 65nm processors production in Q4 2006 - Q1 2007, it will be over a year behind Intel in terms of 65nm manufacturing process. Transition to 90nm process technology took AMD about eleven months from official first revenue shipments to official claim about the completion of migration to more advanced technology and about sixteen month since the 90nm production began.
A slide from a presentation by AMD
AMD is very optimistic about its 65nm ramp, yields as well as 90nm and 65nm product crossover in Q1 2007. Moreover, the company is confident that it would begin volume 45nm production of chips in mid-2008, about one and a half years after the first commercial 65nm chips. The past track record of AMD does not allow us to believe AMD without hesitations. First 65nm products will emerge 2.5 years after the first 90nm chips from AMD. Also 2.5 years passed between the introduction of AMD’s first 180nm processor and the company’s first 130nm chip. Development of 130nm silicon-on-insulator (first chips shipped in April, 2003) as well as 90nm SOI process technologies (first chips went to clients in July, 2004) went a bit faster, however. Still, if AMD is enough successful with its 45nm process technology, it will only be a couple of quarters behind Intel, not that significantly.
The world's second largest supplier of x86 chips has downplayed Intel's ability to launch new process technology ahead of its smaller rival, claiming that AMD still has performance crown. But now that Intel has a very promising Core micro-architecture, AMD needs new process technologies to pack more features into its CPUs, or just increase clock-speed potential while keeping the power consumption low.
According to Core 2 benchmark numbers on the Internet, the Core 2 Duo chips are generally faster compared to AMD Athlon 64 X2, whereas according to Intel Corp., its new Xeon 5100-series chips for dual-processor servers and workstations are considerably faster than AMD Opteron 200-series chips. AMD may retain leadership in multi-processor server market for several quarters from now, but that business is unlikely to bring really huge volume numbers. Potentially, this may mean that AMD would stop to increase market share. Nevertheless, analyst Nathan Brookwood of Insight 64 believes that AMD may continue to gain market share, even when Intel Corp.'s new micro-architecture is on the market.
The impact of Intel’s new Core micro-architecture based processors will have on AMD’s ability to gain market share is overestimated. Look how well the company did last year in the mobile segment, where it had no technology advantage (and a real deficit) vs. Intel,” Mr. Brookwood said.
Hector Ruiz, AMD's chief executive officer
Moreover, AMD’s current K8 micro-architecture resembles that of the K7 in terms of general principles. Given that the K8L will also be generally based on the K8 and K7, it is nearly fair to say that AMD will use the same core with serious tweaks for nearly a decade in 2008. This means that AMD needs some new blood from outside of the company to bring-in some new ideas.
Additionally, the IT equipment becomes more and more complex and soon the industry will require competitive platforms for certain devices, not an ability to assemble many components from different developers and manufacturers and hoping they would work together. Intel is promoting computing platforms, Philips is preparing an ambitious amBX project – a platform for a high-tech living room. AMD should follow the same road too, it seems.
“In order to stay competitive, AMD needs to keep up with Intel’s technological advances, however AMD cannot match Intel’s R&D spend that is around 6 times as large as AMD’s R&D budget. To avoid lagging too far behind Intel, the best competitive response for AMD would be to seek out smaller companies that provide key technologies and buy them out,” said said Martin Kariithi, an analyst with Technology Business Research.
To sum up, AMD may find itself in the situation akin to 2H 2002 – 2H 2003, when its processors were lagging behind Intel in terms of performance. The situation may get even worse, as AMD sells nothing but CPUs: if the sales drop significantly, the company will have to spend its cash on survival, which makes no sense for the firm, if it can buy itself an additional source of revenue ahead of a potential core business crisis.
ATI Technologies was founded in 1985 and is currently the oldest and most experienced graphics chip designer on the planet. But the company is no longer a designer/producer of graphics cards or graphics chips only: it develops a very broad portfolio of multimedia products. Unlike AMD, the company has always said that it is tailored for product portfolio expansion and currently its GPU business accounts for about 50% of the revenue. Still, the company is greatly focused on the PC segment: 77% of revenues come from chipset and GPU businesses, while about 23% are the earnings of consumer electronics division.
While generally the business of ATI grows stably, there are a lot of risks associated with it. The main one today is definitely Intel’s possible entry into the market of discrete graphics processors as well as Intel’s ability to increase its production capacities extremely, which will allow the company not only to supply enough CPUs and chipsets for everyone, but may also guarantee that the firm may enter an additional market with significant volumes of products.
Already now Intel dominates the market of chipsets and graphics devices (with its chipsets containing built-in graphics cores, which are sometimes referred as IGPs, or integrated graphics processors) and if the company presents competitive lineup of standalone GPUs, significant part of ATI’s discrete GPU business will be affected, which will mean that ATI’s and Nvidia’s revenues will go down.
On the other hand, ATI faces competition from Nvidia Corp. and tremendously needs any advantage it could have over the rival. The main advantage ATI does already have is that a huge chunk of its business has no relation to the PC market performance and, given that it is not obvious that Intel will be able to offer competitive GPU lineup, the company does not need to be acquired by anybody. Yet, ATI feels tremendous pressure on its margins now: the company’s gross margin in the most recent quarter was 30.1%, while Nvidia’s exceeded 42%, primarily because ATI’s margins in the desktop discrete segments deserve to be better: they were “not in the PC gross margin model in Q3,” said Patrick Crawley, ATI’s chief financial officer during the most recent conference call.
“Despite missing the streets revenue target in Q1 2006, ATI’s long-term outlook is positive. As a result, there is a risk that the purchase could command substantial goodwill. This would dilute the profitability of the deal or force AMD to incur a significant write-down at some later point,” said Martin Kariithi of TBR.
“ATI does not need the acquisition. Nonetheless, AMD is rumored to be strongly interested in ATI,” said Jon Peddie, the head of Jon Peddie Research.
And ATI truly does not seem need the acquisition in the short-term, as everything should play in favour of ATI, Nvidia and their desktop discrete businesses, revenues and margins up. Intel releases its Intel Core 2 chips on July 27 and it is highly likely that the launch will catalyze sales of desktop graphics boards. Additionally, more graphics cards are sold in the second half compared to the first half, which is another positive sign for ATI and Nvidia. Nevertheless, Intel is planning to release its DirectX 10-supporting integrated graphics core this Fall, which may drive the demand for desktop discrete components down.
“Nearly all the variations in the market share of discrete graphics have to do with seasonality (consumers buy more discrete, and that increases in the 2H of the year), inventory adjustments or the age of the currently available integrated chip sets (as they grow older, the graphics is less appealing and discrete connections go up) These two – seasonality and chipset age – are far and away the dominant factors in discrete graphics shipments,” said Dean McCarron, principal analyst at Mercury Research.
Additionally, ATI is working with many digital TV (DTV) makers and continues to improve its Xilleon series of products aimed at DTV devices. The Xilleon business continues to set records for the company.
“Our DTV business set record revenue and unit shipments in Q3, now six sequential quarters of record revenue. On the strength of our Xilleon 240 processors, we expanded our position in nine of the top 11 LCD OEMs. Two key areas for innovating in DTVs are displaying superior image quality, and system on the chip integration, both areas where ATI is excelling in the very fast growing market,” said Dave Orton, ATI’s chief executive officer, during the most recent conference call with financial analyst.
Moreover, ATI’s aggressive branding campaign, which started a couple of years ago in the PC market is now expanding on the consumer electronics, something, what even Intel would like to see with its Viiv!
“Our next generation Xilleon is ramping with designs hitting the shelves this fall. This new Xilleon will redefine image quality benchmarks and deliver features reserved in the past for separate, specialized chips. Image quality is such a critical criteria now for digital TVs that OEMs are now asking to use the ATI logo on upcoming designs as a competitive differentiator ,” Mr. Orton said during the call.
On the other hand, if AMD acquires ATI now, it will ensure that its revenues are high because of GPUs and DTV chips even when Intel is king of the hill and will also benefit, when it invents something, which causes standalone GPU sales to rise. One thing AMD will not have for sure is Intel chipset business (in Q1 2006 ATI shipped 4.3 million Intel chipsets and 3.9 million AMD64 chipsets), which was largely advertised by ATI as well as very close collaboration with the world’s leading chipmaker, also hyped by ATI.
“In [FY2006] Q4, the powerful combination of CrossFire, the 1900, and Intel’s
The key areas, where ATI Technologies has put a lot of strategic efforts to in the last few years are the following:
Let us try to think whether ATI has truly succeeded in achieving the goals:
The key operating goal of the company was to:
While ATI has been executing itself fairly well, it has not become an indisputable leader across all the markets it serves. Obviously, AMD is not an indisputable leader on the CPU market either. The chances that both ATI and AMD will be pressured by Intel on their respective markets are there and in this case collaboration of resources might help the merged company.
The impact of AMD-ATI merges on the industry, if the merge happens, will obviously be tremendous.
“It will be shocking to the industry. Nvidia and AMD have had a long-term and profitable relationship and that will change. AMD will be able to offer a complete solution (CPU, chipsets, IGPs, and GPUs) which will be broader than Intel’s (only CPU and chipsets currently) However, the duopoly of AMD and Intel and the duopoly of ATI and Nvidia will be greatly disturbed and that may not be tolerated by the FTC,” said analyst Jon Peddie.
“Both companies appear to be doing fine by themselves, and each has an advantage in being independent. If one takes both a very long term view and also believes that graphics and processor functionality will integrate over time, then a merger would make sense. Outside this line of thinking a merger seems less favorable as it undermines AMD’s ecosystem by alienating current partners and also creates an awkward choice for ATI as it would have to choose between supporting Intel potentially to the detriment of a combined AMD-ATI, or choosing to support only AMD and declining the benefits of a larger market and revenue base,” said Dean McCarron, principal analyst of Mercury Research.
If not keep the “platform strategy” “unveiled” by Mr. Sanders seven years ago in mind, the only strategy of AMD, after it either got rid, or simply withdrawn from telecommunications, chipsets, flash and special-purpose processors businesses, was to sell x86 central processing units. When asked about possible acquisition of ATI, AMD’s president Dirk Meyer told analysts: “our mission statement is to pursue standards, that’s core for us and key for us, nothing else.”
“So if that [statement] is true, then how do they rationalize ATI? It’s not a ‘standard’. Maybe AMD thinks it could make it one, but I doubt it,” said Jon Peddie in a conversation with X-bit labs.
But the acquisition of ATI will allow AMD to follow the “plan” of Sanders and the path of Intel. By acquiring ATI, the world’s second largest maker of x86 microprocessors will find itself being a company, strongly oriented on the PC and entertainment industries, as it will be able to:
“The lack of a chipset solution puts AMD at the mercy of third-party vendors pricing which hurts the company’s margins. Acquiring ATI would allow AMD to be more competitive against Intel in pricing and post better margins. Currently AMD relies heavily on Nvidia for high-end chipsets for the gaming and corporate PC market. This is a lucrative high-margin business; acquiring ATI would allow AMD to get a bigger piece of the profit pie in this market. Nvidia has a strong brand-name in the high-end gaming market, AMD can however make inroads into this market by offering a strong integrated chipset solution, mirroring Intel’s 'platform' approach,” said Technology Business Research analyst.
ATI has more than 3400 employees in the
“ATI is a pretty large company for AMD to absorb. ATI’s Q1 2006 revenue of $652 million is more than half of AMD’s $1.3 billion in revenue during the same period. Any integration problems resulting from the deal could place significant pressure on AMD’s financial results,” added Mr. Kariithi.
While ATI has always addressed the market of OEMs very precisely, the company has strong orientation of building the brand recognition. AMD has always lacked proper publicity and its brand-name is hardly as known among CPU consumers as compared to ATI in the market of GPUs.
One of the reason why ATI may be interested in being acquired by AMD is the fact that the latter has its own fabs, which are generally ahead of TSMC or UMC. Potentially, this means that ATI will be able to make its processors at AMD's fabs and ensure that the chips have some advantages over competing offerings.
AMD has process technology similar to that of IBM, where ATI’s arch-rival manufactured the NV40 and NV45 (GeForce 6800 Ultra and GT products) graphics chips, but without much success: the chips could not achieve clock-speeds necessary to dethrone ATI’s R420 and R423 chips (Radeon X800 XT). So, to say at least, manufacturing at IBM did not give Nvidia any tangible advantages over TSMC. It is not evident, that very close collaboration with AMD helps ATI.
A worker at AMD fab 30 in Dresden, Germany
AMD’s process technologies are tailored for AMD’s processors, whereas ATI’s engineers are used to work with TSMC and the tools it requires to create the chips. Even though AMD claims that “dozens of product types and speed grades to be manufactured in a single AMD fab using the same tool sets and production lines”, it remains to be seen whether such manufacturing is really efficient for both AMD and ATI.
If it is possible to create the GPUs more efficiently on your own fab, then, ATI is likely to offer higher performance products and rival Nvidia may find itself in a difficult situation. Moreover, ATI will save pretty much on prototyping costs, if it has access to AMD’s fabs eventually. On the other hand, if AMD does not execute itself well, ATI will not have enough manufacturing capacities for all of its chips.
At press time, ATI’s market capitalization was $3.94 billion, plus, the company had $518.113 million in cash. AMD’s capitalization was $10.9 billion and the chipmaker had $2.63 billion in cash.
It is unknown whether AMD and ATI have actually reached acquisition agreement, but some Chinese web-sites reported late last week that the agreement had been reached. Theoretically, this would mean that AMD may announce the acquisition of ATI as early as on Thursday, July 20, when its Q2 FY2006 results are to be announced, or after market close on the 14th of June, on Friday.
It is obvious, that AMD is unlikely to buy ATI for cash, as it does not have so much resources and it is uncertain whether it is able to take credits anywhere. But AMD may either get the graphics company for the shares of the merged firm, or using a combination of cash and stock.
One have to know, that any potential acquisition deal means stock prices shifts and the fact that quite some sources spread the information about possible AMD-ATI deal may be a result of someone’s desire to manipulate with shares.
“Each time I’ve tried to trace back the source of the rumors and when I have been successful in doing so it's gone back to someone with a financial interest in spreading them,” said Mercury Research principal analyst Dean McCarron.
While our sources generally tend to be correct, chances that no acquisition is going to take place, or the companies find different ways to collaborate, or just a part of ATI will go to AMD, while the rest will remain independent, still exist.
Even though there are a lot of advantages for AMD and ATI to unite, only one technology analyst believes that the deal will happen, while the rest three polled by X-bit labs remain skeptic.
“I remain highly skeptical, and will believe them only if/when I see an announcement from AMD and ATI,” said Nathan Brookwood of Insight 64.
“It could very well happen though as there several factors in favor of the deal. I think it is likely that AMD will pursue the ATI acquisition. If it fails to materialize the most likely reason would be that the goodwill demanded by ATI was too rich for AMD,” said Martin Kariithi, an analyst with Technology Business Research.
“I'm quite skeptical until the companies themselves say otherwise,” said Dean McCarron of Mercury Research.
“It really doesn’t make sense for AMD to acquire ATI, I just don’t get it,” said Jon Peddie of Jon Peddie Research.
Obviously, if the acquisition takes place, it will be more of a strategic merge, rather than a pure takeover, as AMD does not have enough resources to proceed with hostile takeover or pay for ATI significantly more than its market capitalization is. Therefore, the deal should have great strategic sense for both companies.
If the takeover happens, AMD will become a provider of a broad portfolio of solutions overnight, but it will be tough to organize the new company for the management. ATI’s and AMD’s corporate cultures are completely different, moreover, ATI is focused around rapid development of graphics processors, chipsets and technologies, whereas AMD is concentrated on creation of successful x86 micro-architectures and successful fabrication of them at its foundries.
Let us make no final conclusions, but just try to summarize, what the potential deal means for both companies.
Advantages for ATI:
Disadvantages for ATI:
Advantages for AMD:
Disadvantages for AMD:
Impact on Nvidia:
The changes AMD-ATI merger can mean for the whole industry can be even more dramatic. Our readers are welcome to share the thoughts on the matter in the comments, or in the forum.