Analysis: How AMD Asset Light will reveal MAD AMD

Share on facebook
Share on twitter
Share on linkedin
Share on whatsapp
Analysis: How AMD Asset Light will reveal MAD AMD

Sunnyvale (CA) – It’s easy to get lost in the continuous wave of bad news surrounding AMD, competitive threats and predictions that the company is on its way out. Over the past few weeks we spent time with AMD employees and partners in and around AMD’s ecosystem. Not everyone is happy anymore and AMD’s often praised talent how to treat partners is showing dents. Time for a look at AMD’s internal struggles, corporate culture as well as our prediction how AMD will cope with the current situation and possibly emerge as a much stronger player in the not too distant future. 

It does not take much to see that Barcelona, ironically the product that was supposed to save AMD from Intel’s Core 2 Duo assault, unveiled all the weaknesses in AMD. Originally, the K10 architecture (Barcelona/Agena) was supposed to enter volume production in February 2007. Then, the product was delayed to August 2007. In the end, we all know what happened – the Translation Lookaside Buffer bug delayed the volume shipment until Barcelona B3 showed up – in April of 2008. We don’t want to blame everything on Barcelona, but the fact that AMD was unable to compete this during 2007 brought the company more than $2 billion in losses. No matter how much money you have, $2 billion is steep.

The problems for delay did not surface in February 2007, nor was the delay the result of some weird technological issue from The Twilight Zone. AMD’s sales executives were quite happy with the fact that the Opterons were selling like hotcakes. In fact, they were happy enough to decide that 65 nm development could be delayed to make sure the company could get out as many 90 nm wafers as possible. Delaying of the process pushed back the Barcelona development and we ended up with a rushed product. The TLB bug contributed to a grand total of 14 months of delays. If you remember, the original Opteron was also delayed for close to two years. But back then, technology caused the delay and not strategic decisions.

AMD SludgeTM

If we compare AMD to Nvidia or Intel, it is easy to see what is actually going wrong with the company. AMD’s culture was explained to us as the sludge, often seen in countries such as Spain, France, Italy, Croatia, and Greece: Distributing the responsibility for past, current and future projects to a level where it is impossible to tell who was responsible for the project that what went wrong. The result: Scapegoats get pink slips.

Over the past two years, we listened to numerous partners complaining about the projects that were started, effort (time+money) that was invested, but suddenly stopped. “Sludge” was the term we heard from several industry partners, from the AIB/OEM/Distribution industry to the Software/FAE side. We listened to partners complaining that the representatives that they were working with who aren’t reachable simply shut down their computers during important hours. Just one of the results of a scenario with virtually no flexibility is that deadlines will be missed. Problems aren’t surfacing on a 9-5 schedule.

One of the people we asked, and who requested to remain anonymous, is still an important partner for AMD and had to say this about the Radeon HD 3800 launch: “We worked hard on launching a product that would cater to the enthusiast market segment. However, seven days before the launch, we were still waiting to receive something from them. No blueprints, no dummy boards, nothing.  How can we compete with goblins [Nvidia], when their partners launch enhanced products on week one?” Another partner commented on one office in EMEAI (EMEA plus India) region: “I cannot believe that they hired a MarCom person that does not speak English. What is going on in there?

The Radeon launch was subject to a lot of criticism from a lot of vendors, but it got silenced by the Barcelona TLB bug and the cancellation of the Phenom 9700 part.

AMD executives jumping off board

The issue with “The Sludge” spreading inside the company caused several AMD executives to leave the company would span across several pages, but we’re going to name just a few:

Henri Richard. One of first things he did at AMD was to establish a partnership with Ferrari. Due to this partnership, the Mubadala Abu Dhabi investment fund (UAE) bailed out AMD with a $608 milion investment and now owns 8.1% of AMD. It is the same fund that owns 5% of Ferrari, with logos located on Ferrari F1 cars and drivers. Henri left the company when the TLB-gate disaster unfolded and  now heads sales and marketing at Freescale Semiconductor.

Dave Orton. probably one of most soft-spoken executives you’ll ever meet. Dave was CEO of ArtX, which got acquired by ATI, took the lead of ATI, turned them around, established a partnership with Microsoft, and expanded the company into the desktop, mobile, console, handheld and consumer electronics space. After failing to acclimatize with the new corporate culture, Dave decided to leave AMD in the summer of last year.

Phil Hester. The most recent departure, Phil was brought from IBM to head the AMD Fusion project. He left after the whole Barcelona TLB bug debacle and the delay of the upcoming “K11” core to late 2009 (Fusion processors: Falcon/Swift will feature 45 nm K10.5 core, not K11 – Bulldozer/Bobcat), a key part of AMD’s APU (Accelerated Processing Unit) strategy. He was replaced by Mike Uhler (who, however, does not carry the rank of CTO).

Other departures included VPs such as Peter Edinger, Rich Hegberg, Gianluca Degliposti, and Chris Talago, just to name a few. Overall, the number of “leaving AMD” e-mails we’ve received so far was greater than from any other company we’ve seen. We have mentioned only executives here, but a lot of people left the PR department as well: Andrzej Bania, Chris Evenden, Lorenzo Martone, Lars Weinand (former Tom’s Hardware Guide editor), Bubba Woolford, Jagoda Zieleznik and many more. When executives and PRs are jumping ship, there is something seriously wrong.


Read on the next page: AMD’s Asset Lite is the path to AMD 5.0 – defined by a “MAD” investment, Timeline and Conclusion

AMD’s Asset Lite is the path to AMD 5.0 – defined by a “MAD” investment

In its current configuration, AMD needs about $2.0 billion in revenue per quarter to turn a profit, and the company right now cannot deliver more than $1.5-1.7 billion. With the reduction in workforce, the company will save $40-$60 million per quarter, which should help the final bill. In a best case scenario, at least one quarter per year should turn in a profit, if sales and margins stay the same. Your guess is as good as ours, as to whether AMD in fact can deliver its promise to become profitable again in H2 2008.

But there is a more dramatic change in the works at AMD: Asset Light. Before to get to that, let’s look at the history of AMD.

AMD 1.0: 1960’s, 1970’s. A memory semiconductor company, reverse engineering of Intel’s 8080 processor, the bit-sliced microprocessors Am29K series, experiments with graphics and EPROM memory. Sex, drugs and rock’n’roll are all parts of an industry legend how company founder Jerry Sanders signed some of his first customers.

AMD 2.0:  1980’s to mid-1990s. A lot of AMD employees like to skip the first part, given the wild-60’s and 70’s nature of the business, so they like to say that AMD actually started with a so-called age of “second-source”: IBM wanted a second source for x86 CPUs, and Intel had no choice but to give an x86 license to an … unnamed company, in a way to be able reverse engineer the CPU: The first semi-serious challenge to Intel was the AMD 386DX/40, a 386 CPU with an integrated FPU, working at 40 MHz (Intel’s 386 peaked out at 33 MHz). Truth to be told, 386DX/40 appeared when the i486 was gaining ground. However, the finances were questionable and the company struggled to survive.

AMD 3.0: mid-1990s to 2006. AMD made its first serious acquisition when the company was bleeding badly – with 5×86-PR1xx and K5 processors suffering from dismal performance in floating-point operations. They acquired NexGen, brought Atiq Raza on-board and started the creation of the first truly competitive architectures: K7 and later K8. This era resulted with the success of the Opteron and other signs that company had grown up enough to start working on mission-critical infrastructures.

AMD 4.0: 2006-2008. The acquisition of ATI came at the very end of a time when AMD was sitting on its laurels, enjoying the utmost success of the K8 architecture, challenging Intel on all fronts: desktop, mobile, server and workstation. However, after the acquisition, the company slowed down to a crawl, while the competition stepped up the game and left AMD/ATI offerings behind.

AMD 5.0: Asset Lite era, 2008-2012. During and after cutting 10% of its, AMD’s business units are migrating into more flexible divisions. By Q1 2009, AMD announces an additional Mubadala Abu Dhabi investment for its business, which we rename from “RealMenHaveFabsTM” to “OnlyArabsHaveFabsTM”. AMD is set to adopt a semi-fabless model. The biggest challenges remain: Corporate culture and market conditions.

Mubadala Abu Dhabi is an investment fund owned by Abu Dhabi Government and serves as a direct investment company for United Arab Emirates. Mubadala invests in the aeronautics & automotive industry, electricity and real estate. Technology-wise, MAD owns two telecommunication companies: YahSat (100%, satellite phone) and du (20%, fixed line provider).

What is Asset Lite?

There have been various conclusions and thoughts about what AMD’s Asset Lite is: In short, we are talking about the creation of an influential giant that links USA, Canada, Germany, Russia, Emirates, Singapore, China and Korea into one melting pot. It is important to understand the fact that AMD’s vision for the future is not delivering chips, but rather complete platforms. My ex-colleague Charlie Demerjian delivered a spot-on analysis a while back, but what he omitted is just how “AMD Foundry” would operate.

AMD is splitting into two entities: Daamit Inc. (“original” AMD+ATI), AMD East Fishkill, AMD Saxony, facilities in Malaysia, Singapore, China and upcoming AMD New York will going to form into “AMD MAD LLC”. By some odd chance, the short name for Mubadala Abu Dhabi would be MAD. So, we have a MAD AMD in the making.

This is the very same model as FASL LLC (Fujitsu-AMD), a joint venture that was renamed into Spansion. FASL LLC started its life as Fab 25 in Austin (TX), which was converted from CPU to flash manufacturing; it took capital from Fujitsu and ended up being spun off. This reduced the headcount by 8400 employees and you can expect that AMD MAD could do the same.

MAD AMD LLC is an interesting combination: A split ownership between AMD and the Mubadala Abu Dhabi investment fund. We’re talking about a more than $5 billion investment for a minority stake in this joint venture, or a golden ace from Hector’s sleeve. Ironically, according to our sources, this new company will be “worth” almost twice as much as AMD’s current market cap.

Just like in a case of Ferrari S.p.A, Mubadala Abu Dhabi is apparently happy with a minority stake, while AMD has to control the company: The x86 license terms between Intel and AMD state that the percentage of non-AMD manufactured wafers cannot exceed more than 20% of x86 market share. MAD AMD LLC will be owned by AMD, case closed.

According to numbers that found their way to us, AMD Saxony LLC (Fab36+38) has to undergo a $2.5 billion investment for a complete conversion to a 45 nm 300 mm wafer plant and the ability to produce 50,000 wafer starts per month (resulting in a manufacturing capability of more than 100 million CPUs).

At the same time, the New York Fab was supposed to cost $3.2 billion. However, due to the continuous devaluation of the US Dollar and the rising cost of oil, the New York Fab is now expected to cost around $3.6-4.2 billion, plus an additional $1.1 billion from NY State (originally, it was $900 million). Thus, MAD will have to invest between $4.5 and $6 billion. Our bet is on $5.5 billion as the original sum, but we are very deep into speculative territory here.

The headcount for AMD Inc. will decrease by 4000 (Saxony) and 3000 (Malaysia, Singapore, China). With the completion of the New York Fab, MAD AMD LLC will employ around 10,000 people across the globe, according to industry insiders.

Daamit Inc. is nothing else but the current company without the manufacturing headaches. This will tremendously help to reduce Capex and enable AMD to turn in revenue with much better margins. In numbers, AMD’s staff count will first decrease from a current 16,800 to around 15,000 (10% cut) and an additional 7000 – 8000 will be spun off into MAD. With approximately 7000 – 8000 employees, AMD should be much more flexible in terms of designing new products. However, the inherited danger of The Sludge will remain.

Timeline

According to our sources AMD has started to migrate from version 4.0 to 5.0. This will happen in following stages:
1. Workforce reduction: 10%
2. Reorganize regions in order to cover the market better. We’ve seen EMEA becoming EMEAI (Europe, Middle East, Africa, India), and the trend is set to continue.
3. Asset Lite announcement: changing the business model into AMD Inc. and MAD AMD LLC. Manufacturing will receive a cash injection high enough to build the New York Fab and pay for tooling machines for Fab 38 and Fab 36.

Of course, each of these moves can result in a positive, but also in a negative way. The inherited traps are:
1. Keeping the obedient, pink slips for creative individuals. During the first wave of firings at AMD/ATI, we heard that people who kept their head down stayed on the job, while people that were trying to change the status quo and threatened the sludge were given good-bye letters. If AMD loses its creative soul, the company is destined to remain on the outskirts of competitiveness.
2. Expanding the regions by too much will result in poor coverage due to lack of understanding for specific markets (we heard numerous complaints about how all three major players mistreat Africa, middle-East and India/Singapore).
3. AMD siding with middle-Eastern companies is guaranteed to cause a lot of turbulences, especially in the case of the Luther Forest Technological Park. Arabs owning the largest allowed number of shares for a factory close to The Big Apple is a tough one to crack.

Conclusion

AMD is not going down any time soon and even after the AMD + ATI vs. MAD AMD LLC split, the cooperation with IBM, TSMC, Chartered, ANGSTREM will still be in place. In fact, it may expand into another alliance. The current corporate climate has to change, otherwise AMD will continue to be just an occasional challenger to industry heavy-weights such as Intel and Nvidia. This is, according to our information sources, one of primary reasons why the deal with Mubadala Abu Dhabi fund has not been announced yet.

One thing is certain: Doomsayers claiming that AMD is dead forgot to check the facts. Just like they forgot to check actual facts of Ferrari in 1993, Apple in 1997, Airbus SAS in 2006, Nvidia in 2002, and Microsoft in 2007. This is big business, and big changes do not happen overnight. And success or failure of one product cannot change the destiny of a company.

Author