Skip to main content

IBM, Nvidia In Line To Buy AMD

Opinion - The past 19 months haven’t been easy on the company from Sunnyvale, CA, but in this analysis we’re going to try to shed some light into what’s really going on over there, and provide a reasonable insight into the current state of, "To merge or not to merge, that is the question."

At the moment, AMD is a very interesting target for acquisition - and there are two obvious candidates. After looking for a target for acquisition, AMD turned into a bride (can you imagine AMD CEO Hector wearing a wedding gown?) Then again, our own Charlie Demerjian fell through a rabbit hole for betting on Intel, so nothing is surprising these days.

AMD’s success...

Before Intel managed to kick ego-boasting, sleeping-on-laurels AMD with its brilliant Core 2 architecture, a lot of people thought that AMD’s success was all up to current management, headed by ex-Motorola guy and advisor to George W. Bush, Mr. Hector De Jesus Ruiz and the rest of the gang.
Truth of the matter is that all these guys did was execute on a brilliantly positioned situation, set by no other than legendary Jerry W. Sanders III (a.k.a. Colonel) and Atiq Raza. After losing out on the Microsoft Xbox deal, and with K8 in advanced development, Sanders felt that they needed to bring a strong seller and a good politician to the team, and old ex-Motorola guy, Hector J. Ruiz was selected. This was followed by the arrival of Henri Richard, who introduced company into sport sponsorships, creating a successful relationship with legends of sport such as Ferrari, Lance Armstrong, and footie clubs in Europe.

From a company that was losing out heavily in the mid-90s, AMD delivered a major coup with the K7 design (more well known as Athlon), a chip that trounced Intel’s Pentium III. The K7 design was the baby of a team led by Dirk Meyer. This was the first time in history that AMD managed to produce a better part than Intel had in store (and is going to have for years to come). The result was that even companies from Taiwan and China started to warm towards the Green field instead of Blue river. The success of AMD in the early 2000s is especially visible today, when you go and visit the factories that manufacture graphics cards, USB sticks and other components. In most factories that we’ve visited, AMD Athlon 64 systems are the "weapon of choice" for qualification and testing process. This is a fairly small thing for number of sold units, but huge in terms of influence.

...and its fall

AMD was always preaching the "customer centric" mantra, but in reality, the company was changing processor sockets and making incompatible design calls almost every 12-18 months. AMD’s Athlon 64 debuted with Socket 754, but when company learned of its mistake, it created Socket 939. Athlon 64 FX was an interesting story, changing Sockets as marketers wanted: original Socket 940, then 939, then back to 940 (AM2, AM2+, AM3 also use 940 pins, but are incompatible with the first S940). Also, AMD kept on promising something that it could not deliver: backwards compatibility. SI’s had ton of issues with Santa Rosa Opterons, and the whole debacle over Phenom upgrades as tested over at Tom’s Hardware just leaves questions. While Intel keeps silent and uses the same socket for years now, few dared to openly criticize the so-called "Chipzilla", with the company’s policy of keeping the same socket, but shunning the upgrade option (e.g. you cannot put a new CPU on an older motherboard, due to "some XY incompatibility").

There are numerous examples where AMD went wrong in 2007, from treating the press like kids to events in Lake Tahoe and Warsaw, forgetting the workstation market at the K10 Opteron launch (this was an answer given to us by Dirk Meyer in Barcelona), mistreatments of the channel, and losing the already won HPC contracts with its inability to deliver thousands of Opterons for HPC. And yet, at events such as Analyst Day, you see analyst hot-shots nodding to the mantra given by AMD leadership, and not asking real questions.

At the end of the day, AMD screwed up badly in 2007, but this time it was so bad because now the focus was on them. Back in 2002, and in the 1990s, nobody considered AMD to be a serious player. Now, it is in the limelight, and obviously making wrong choices every step of the way. Exodus of high-quality management and engineers after the takeover of ATI was just the silent witness that something went terribly wrong.

And where we are now? We’re left with the prospect of AMD/ATI ending up bought by either IBM or Nvidia. The relationship with Nvidia is a very interesting one, so we’ll first analyze SNAP.


SNAP - Strategic Nvidia-AMD Partnership

As you may or may not know, the relationship between AMD and Nvidia has been a long one, starting with the first Xbox, which unleashed nForce to the world. It is not a secret that AMD had Xbox all lined up, with Microsoft buying a special version of the CPU which was somewhere between Duron and Athlon - if we recall correctly, AMD’s XCPU was a Duron design with 128 L1+128KB L2 cache, instead of Duron’s 128KB L1 + 64 KB L2 cache combo. Intel pimped AMD at the last minute with a Pentium III design (actually, this design was more Celeron, but let’s not go into that), and all of the AMD presentations where AMD was praising Nvidia and Microsoft relationship, went into the toilet.

Given the fact that SNAP was ousted out of Xbox, AMD and Nvidia continue to closely collaborate on bringing the Xbox design onto a PC, and nForce 420 was launched in June and delivered in September of 2001, ahead of Xbox console. It was nothing but an original nForceX design for Microsoft, with original support for EV6 FSB (that AMD used) instead of Intel’s GTL. A year later, this was followed by the legendary nForce2 chipset, which sold in the tens of millions (quite a high number for an enthusiast chipset), and Intel was constantly shunned for an Intel-based nForce. Nvidia was disappointed with the nForce3 design, since it was completed a year ahead of Athlon 64, and that gave the company time to create nForce 4 SLI for Intel, only to end up screwed by Intel’s Pentium D 820 design, which was incompatible with the boards (a desperate bid to avoid utilizing overclocking capabilities of the chipset). This is the reason why Nvidia is extremely careful in its dealings with Intel, and the recent nForce 680i-Intel 45nm Penryn fallout only reminded us of the fragile relationship between the two.

SNAP continue to take place even after the AMD-ATI buyout, with a war that engaged AMD people versus ATI people. This led to ATI’s roadmaps being leaked to Nvidia and various departures from key engineers harmed the product cycles. To us, this is also the biggest failure of Hector Ruiz as the CEO of AMD. He did not voice out clearly what would happen with ATI and why the acquisition was necessary, and when we heard stories that engaged various people from AMD Human Resource department - questions like "Why can’t we fire more ATI people?" and various internal e-mail communication still stays vivid within this author’s head.

AMD started attacking Nvidia in summer of 2007, making bundle deals that included AMD’s processors and ATI graphics and chipsets for the "back to school" period. Thus, a whole year was lost and in that time, Nvidia gained tremendous market momentum.

Failed AMD-Nvidia merger, take I

However, with the recent talk about Nvidia buying ATI, it would be only logical to go back at that point in time when AMD wanted to buy Nvidia. In fact, Nvidia was AMD’s first choice for acquisition. At the time of negotiations, AMD’s market cap was $23B, and nV was worth somewhere in $11-13B range, or just about dead-even between merger and acquisition. We’re talking about the second half of 2005, and AMD was looking into ways to respond to upcoming threats named Nehalem and Larrabee. But at the end of the day, it all fell apart because of fallout between Hector Ruiz and Jen-Hsun Huang. Huang wanted the CEO position, a position that Ruiz did not want to seize in any way, and the merger did not happen. If that merger had gone through, we would probably have a monster semiconductor company right now, with Athlon 64, Phenom, Opteron and GeForce, Quadro and nForce dominating the market, regardless of the strength of weakness of some components in the package.

AMD turned its focus onto ATI, and merger talks between AMD and ATI begun at the very end of 2005. The deal was set to happen by March 2006, and it was publicly announced on July 24th, 2006. The Inquirer wrote an excellent analysis on the subject, but the fact remains that ATI was a second pick. If AMD was smart enough, with an acquisition of Ageia the company would actually have gained a foothold in the physics segment, instead of constantly making two steps forward and one step back.

AMD-Nvidia merger, take II

Fast forward to November of 2007. AMD is in trouble and the natural call for Jen-Hsun Huang (one of the most skillful CEOs in the industry) - was to go back to scene of the crime. Approaching Ruiz and the gang with the attitude "what goes around, comes around" and "I want to buy you now" did not go well, but business is business and there was no reason why the deal would not have gone through, if Nvidia had enough money to take the company AND the debt. A repeat of the 3dfx charade is not an option here.

However, in order for that to happen, the only way to go was to get support from partners from AMD’s eco-system and this is where the trouble happened. While Nvidia is achieving record success with sales, it has done so by weakening its partners. From a company that started selling GPUs with memory back in GeForce3 days (and threw Guillemot/Hercules out of the race), Nvidia turned into a company that now sells a complete card, and partners are nothing else but "sticker stampers". This crashed the profits of all the companies involved, and many fell out from the race. With Nvidia controlling the AIB market for both ATI and Nvidia products, Taiwan would rush into Intel’s arms. Memories of what happened with 3dfx when ex-Siemens CEO took over the company are still livid in world of graphics card manufacturers.

The second unfavorable factor is the debt-to-equity ratio, which is the reason why AMD had $10M lower market cap than ATI at the time of purchase. Nvidia has a lot of money in the bank, but to eat up a company like AMD, it would have to cough up somewhere in tune of $10B, since AMD is 5.4 billion dollar in debt.

But with AMD on-board, any rumors that Nvidia is toast would go forever, and Intel would have to face Nvidia on almost every front company that is competing. Nvidia can easily diversify into a networking company (it bought 3Com team ages ago, which is the reason for the brilliant "plug’n’play" connectivity on nForce chipsets), handheld company (APX 2500 is much more important than you might think at first), desktop (Phenom+nForce+GeForce), mobile (Turion+nForce+GeForce), workstation (Opteron+nForce Pro+Quadro), server (Opteron+nForce Pro+Tesla), HPC (Opteron+nForce Pro+Tesla)... and of course, with the recent purchase of Ageia, nothing would stop the company from implementing PhysX in every pore of its DNA.

Sadly, Huang’s deadly fault is the fact that he believes in the value of a single company. Just like the company destroyed the 3dfx line-up and did not use the world’s second IT brand (Voodoo, just after Pentium) to its full potential, we have no doubt that Nvidia would cannibalize ATI and basically screw up chipset development, as well as development of future Radeons. Instead of creating synergy, the company would probably lose valuable time in preparations for the arrival of Nehalem and Sandy Bridge from the CPU side, and Larrabee from the cGPU side.


Why IBM?

IBM is a logical choice to buy AMD. For starters, AMD cannot be bought by just anybody, since there are limiting factors with the license for the x86 architecture. If AMD would be acquired by a foreign company, that license would be lost, and that would diminish the value of AMD as a company. Thus, Nvidia and IBM. are the only two real players are in the game.

When it comes to IBM, the matter of market cap is very simple - AMD is currently valued at $3.88B, and Nvidia goes for $13.44B, while IBM is worth just shy of the $150B mark ($148.88B). Thus, AMD is pocket money for IBM, and the potential gains with AMD under IBM’s wings are numerous.

For starters, IBM has a ton of manufacturing partnerships and is a member of a manufacturing alliance that involves AMD, IBM and Chartered from one side, and Toshiba and Sony from other. Let’s not forget the fact that with AMD/ATI (or just Daamit) on-board, IBM would rule the world of video game consoles - all three consoles feature IBM processor, and the two leading consoles have ATI GPUs inside. The next generation of consoles could look much different, if IBM would have such design teams to develop complete solutions for Microsoft, Sony and Nintendo. And the market of 200-300 million consoles is nothing to be sneezed at.

There is also a small fact that AMD relocated one engineering team from Austin, TX to East Fishkill, NY, to work on a new manufacturing processes on the spot. With a lot of AMD engineers working closely beside IBM on next-gen designs, opening all secrets would create an incredibly flexible semico giant that would be able to seriously cut down Intel’s market share.

This would also propel IBM into worlds that are currently closed for the company, such as the consumer electronics segment, where ATI Imageon chips go head to head against STMicro, Thomson, Philips, Silicon Optix and others. The handheld market is also an interesting segment where ATI and Nvidia ship hundreds of millions of chips per annum, and we won’t go into lucrative HPC deals where IBM usually competes either with its Power, Cell or Intel chips. Going all-IBM in the x86 arena, as well, would open a new world for IBM’s services, and this is what the company wants to influence. With ATI on board, IBM would enter the world of visual computing, which is set to dominate the world in decades to come (holographs, HUDs, monitors everywhere we go).

GPGPU also comes to mind, and this is the reason why both IBM and Nvidia want AMD very, very badly. AMD has been quite vocal about integrating GPGPU-friendly improvements, and GPUs would gain CPU-friendly features.


With Nvidia not having any support from partners to buy AMD, and the fact that the debt-to-equity ratio could prove deadly, Nvidia is effectively out of the running for the company. A deal could still happen, but if AMD is for sale, IBM is the logical choice.

As some Intel insiders have told us, this would be the move that scares Intel’s leadership beyond belief. There is a saying inside Intel: "IBM is a vampire that never dies". You can give it Tom Cruise’s face, but the fact of the matter is that IBM has been around the block before the block even existed.

A combination of IBM and AMD would leave x86 licenses safe and intact, and that is a matter of utmost importance. From another side, the manufacturing alliance should not be overlooked.

A third option for AMD is to weather the storm and stay on course alone, surrounded by a ton of partners, and faith that K10.5 and K11 will bring out competitive CPU parts, to complement graphics and chipsets. Only time will tell how this story ultimately unfolds.