The Business of Tech: Breaking Up Is Hard to Do - AMD Goes Fabless

We have periodically written about AMD's poor financial standing for some time now, as the company has struggled for nearly 2 years now to keep their operations profitable and their debt manageable, a historically difficult task for the company. Back in July, we noted that AMD may need to sell some assets to raise cash in the face of these problems, and this has since come to pass. August saw the sale of their digital TV unit (responsible for the OCUR CableCard receiver, among other things) to Broadcom, and today we've seen something much bigger.

We just got off of a conference call on AMD, where they have announced that the year and a half speculation about their "asset light" strategy is true. AMD is going fabless. Effective immediately, AMD will begin to spin off its fabrication facilities into their own company, simply called the Foundry Company. There are a lot of details to cover, so let's dive right in.

Calling a fab "expensive" is an understatement these days. With the cost of a brand new, state of the art fab exceeding $3 billion, building one requires a great deal of capital that immediately begins depreciating. Once a fab is online, a company is looking at roughly 2 years of top-of-the-line production, after which a newer manufacturing process will come online and what was once a top fab is now a generation behind - and with it goes much of the profit that the fab will ever make. Fabs will continue to produce for several years after they're out of date, starting with the continuation of older parts and winding down later with less complex parts (e.g. chipsets), but ultimately being in the fab business means that you're building or retooling (which is equally expensive) a fab every 2 years.

AMD can no longer afford to play the fab game. Even with generous tax breaks coming from Germany for fabs 30 and 36 AMD's declining revenue, rising debt, and the increasing costs of building fabs has pushed the company to the point where they can't get the capital needed (internally or externally) to keep building fabs. In recent months the situation at AMD has become so bad that the company has been forced to halt Fab 30's latest upgrade and plans to build further facilities in New York, all of which would be necessary to keep parity with Intel in the fab game in future years. Complicating matters, AMD doesn't even have enough business at the moment to run all of their Fabs at full capacity, which represents capacity they paid for in building the fabs and which they aren't recovering the costs of.

As such, AMD has thrown in the towel: AMD and its fabs must break up. AMD cannot afford to keep building fabs, and their fab business cannot afford to be limited to just building chips for AMD. AMD has begun to spin off its fab operations into their own company - the Foundry Company - and if AMD's shareholders and the United States government approve the deal (we'll get to this in a moment), it will be a done deal either at the end of this year or at the start of the next.

The Foundry Company

However, AMD is not just throwing out its fabrication operations and telling them to fend for themselves; there is a great deal more than that involved. Spinning off their fabrication facilities is only one of many things that need to be done. The Foundry Company still needs money to resume upgrades and build new fabs, and AMD still needs money period. This is where the investors come in, and the financials get thick.

The emirate of Abu Dhabi, through its investment company Mubadala Development, previously made a large investment in AMD in 2007 in return for 8.1% of the company. Abu Dhabi will again be the investor AMD is turning to, and as a foreign investment this is why the United States government gets to review the matter. Abu Dhabi will be making investments in both the Foundry Company and in AMD itself. For the Foundry Company, nearly $6 billion over several years is being put directly into the fledging foundry through the Advanced Technology Investment Company (ATIC), in order to give it the capital AMD never could. In return ATIC will own 55.6% of the foundry and will have half of the voting rights. AMD will own the remaining 44.4% and will have the other half of the voting rights. AMD is also giving up Hector Ruiz, who will be leaving AMD to become the Foundry Company's new chairman. The complete details are a bit more complex than this (spinoffs always are), but these are the most important bits.

AMD meanwhile will be receiving another round of investments from both Mubadala and ATIC. ATIC will be paying $700 million to buy that 55.6% share of the Foundry Company from AMD, and Mubadala will be putting in another $300 million for 58 million new shares, increasing their share of AMD to 19.3%. Furthermore, AMD will benefit from an immediate debt reduction, as $1.2 billion of their last reported $5.3 billion in debt will be transferred to the Foundry Company, since that debt is against the fabs themselves. The net result will be a reduction of debt to around $4 billion, AMD will have more cash, and AMD will be free from the immense future commitments to build and upgrade fabs.

Going forward, AMD and ATIC have lofty goals for the Foundry Company: they want to be #1. The Foundry Company will be continuing AMD's long-standing partnership with IBM through at least 2015, giving them access to IBM's manufacturing technology and allowing the two companies to further develop new technology. Ultimately, the Foundry Company is trying to surpass current AMD partners TSMC and Chartered, along with UMC, SMIC, and perhaps one day even Big Blue itself. AMD's Fab 36 will be the Foundry Company's first fab, followed by Fab 30's retooling finishing 2009, and beyond that their new fab in Fishkill, New York. It's a lofty goal to say the least, but given AMD's success in Dresden and the amount of raw capital ATIC is throwing at the process, it's entirely attainable. Once the spinoff is complete, AMD will be the Foundry Company's first customer, although we should note that this will initially continue to be processors, while other AMD products (chipsets, GPUs, etc.) will continue to stay at TSMC for the time being.

Fab 36 at its opening 3 years ago

It bears mentioning that there is a great deal of government involvement with this deal at all levels. Abu Dhabi is investing as part of their own long-term strategy to diversify and drive their economy away from oil. It's likely that their latest investments in AMD are part of that strategy, with the goal of ultimately buying up the company outright. Furthermore, if the Foundry Company does well and needs to build additional fabs, those fabs will be built in Abu Dhabi. Meanwhile Germany is quite interested in creating and keeping high-tech manufacturing jobs and facilities in the country, so the resumption of the Fab 30 upgrade is a big deal for them, although in the long run they stand to lose if all production ends up in Abu Dhabi. Finally, New York has thrown a great deal of money through tax-breaks to get the Foundry Company's next fab and a research facility in Fishkill, largely for the same reasons as Germany. Hector Ruiz's first act today will be to head to New York in order to ensure that all of the previously promised deals with AMD to get the facilities there will be extended to the Foundry Company in light of this spinoff.

Last but certainly not least, we have the future of AMD to discuss. Jerry Sanders - AMD's co-founder and enigmatic CEO for many years - was quite fond of AMD's fabrication abilities and believed them to be absolutely necessary for the continued survival of the company, going so far as to once quip that "real men have fabs." Of course this was well over a decade ago, and since then the costs of a fab have increased dramatically. AMD's current CEO, Dirk Meyer, in acknowledging this has said that "we feel like we're still pretty manly at AMD," along with justifying this move based on the aforementioned costs. Sanders was right at the time, but it remains to be seen whether Meyer is right now. For the sake of the CPU and GPU industries, we hope he is.

It has been a long held belief that for effective CPU manufacturing, the same company needs to be the designer and the builder of the CPU. As these devices are practically built down to the transistor, an in-depth knowledge of the manufacturing process is needed to best design a processor, and again to integrate feedback from manufacturing based on how such processes evolve. The billion-dollar question right now is whether AMD can still achieve that synergy if they are no longer in the manufacturing business. Certainly farming their manufacturing out to a third party won't completely inhibit them from manufacturing CPUs, but it may keep them from manufacturing CPUs as well as they have in the past. The situation is not particularly comparable to GPU manufacturing; while GPUs have just as many (if not more transistors), they do not have the high clock speeds of CPUs, and GPUs are still developed along a functional unit level rather than a transistor level.

To AMD's credit, they do have some experience with third party CPU manufacturing. During AMD's boom years earlier this decade, they enlisted Chartered to manufacture Athlon 64 CPUs on their 90nm and 65nm processes, which to the best of our knowledge worked quite well for the short period of time thereafter that AMD's boom lasted. Yet at the same time, AMD was still a fab owner and had developed the Athlon 64 initially with their own fabs, so it's not an entirely comparable situation. Although AMD does not subscribe to Intel's tick-tock strategy, historically this is similar to how they have operated (Barcelona was introduced on their existing 65nm process), so the loss of their own fabrication abilities could be a significant blow. The flipside is also possible, however: with freedom to evolve separate from AMD's CPU business, the Foundry Company could bring newer process technologies online more rapidly, giving AMD and others the chance to upgrade as the need arises. Which scenario will occur is impossible to say right now, but short term it is more likely than not that this will hurt the AMD's ability to be competitive. All we can do is wait, see, and hope.

Moving forward, AMD is in a tough spot. Nehalem's launch later this year robs AMD of their ace in the server market, HyperTransport, and this has been one of the last highly profitable areas where they've been competitive. The 45nm Barcelona successor Shanghai (and its derivatives) is still set to ship this year, so these should at least help AMD keep the same gap behind Intel. Given the cycle times on CPU design and manufacturing, the true effects of this spinoff will not be felt until AMD's new microarchitecture, Bulldozer, hits the market in 2010.

For now, this deal leaves AMD with more cash which they greatly need to continue operations, and the aforementioned $4 billion in debt. Unfortunately, what it doesn't leave them is much in the way of assets; their fabrication facilities were most of the company's assets, so their debt to asset ratio just got a lot worse. Without a doubt this move has saved AMD; there is no longer a question of whether there will be an AMD in 2010. Had AMD not done this spinoff, they very well could have gone insolvent before the next decade. Beyond 2010 this does not mean much, however; the lack of fabrication facilities could kill their ability to be competitive at the high-end, it could simply kill them entirely… or it could even be beneficial if they can adapt to the new design and fabrication process.

AMD will be announcing their Q3'08 results on October 16. We'll have more on the company then, including a fresh look at their revenue, cash reserves, assets, and debt. Stay tuned.