AMD sacks 1,400 to chase 'emerging markets'
Caught between a Xeon and an ARM
Chip maker AMD has had its share of woes in the past year, and now it will cut its workforce by 11.6 per cent to save enough money to invest in low-power chips and cloud computing.
As the market closed on Thursday, AMD announced that it was cutting approximately 1,400 people from its workforce, which stood at 12,019 people at the end of the third quarter – that's 11.6 per cent, which is more than the "approximately" 10 per cent it announced in its press release. AMD said the "workforce rebalancing" would help it "better address faster growing market segments." It wasn't clear if AMD meant by geography or product type – or both.
Some of the employees who are getting pink slips have already been notified, while others will get theirs before the fourth quarter is over.
An AMD spokesperson told El Reg that the restructuring would cut about $118m from AMD's costs in 2012, which is primarily driven by the job cuts across AMD's global operations and across all functions in the company. AMD is also doing a bunch of other things as part of the restructuring, including closing offices and facilities, and rejiggering software licenses with vendors.
Also helping free up some cash are a number of initiatives to increase operational efficiency initiated earlier this year by Thomas Seifert, AMD's CFO and interim CEO until Rory Read took over two months ago after being brought in from PC maker Lenovo. These initiatives are expected to bring down costs by another $90m in 2012.
AMD says that it will book charges of $101m in restructuring expenses in the fourth quarter, with $56m of that being related to the layoffs and continuation of employee benefits, $44m related to program termination costs, and $1m for asset impairments. AMD expects to take another $4m hit in 2012, $3m for severance costs and $1m for facility closures.
The other thing that AMD wanted to make clear – and what it did not say in its 8K filing with the Securities and Exchange Commission – is that this is truly a workforce rebalancing. The AMD spokesperson said that AMD wanted to make sure it had the right engineering talent for low-power chips and cloud computing and other emerging markets, and to have the right sales and marketing people in the right geographies to chase these opportunities. That very likely means more feet on the street in Asia, as well as more engineers moving offshore.
"The point is that we are expecting to rehire for some of those positions based on looking for a different set of skills," the spokesperson explained. AMD says it plans to reinvest "a significant portion of the savings from the restructuring plan" in emerging market opportunities.
Not to split hairs, but if you reinvest savings, then did you really save anything? There's no question that with SeaMicro selling Atom-based cloudy boxes, Tilera peddling its Tile64 chips in low-power servers, and a school of piranhas with sharp ARM processor teeth. Applied Micro, for example, launched its 64-bit X-Gene ARM processor last week, Calxeda showed off its EnergyCore processors earlier this week, and Marvell and others are coming up fast behind.
With these chips coming to market next year, AMD has as many problems at the low end as it does at the high end, where it does not scale as far as Intel's Xeon E7 processors in terms of sockets and main memory. ®
It is so much easier to just fire and hire new people instead of spending the same severance pay on re-training your existing engineers. In addition, an 11% RIF also greatly enhances loyalty from your remaining work force. /sarcasm off/
Seen this movie before - after a 10+% RIF, whoever was not let go will look around, and jump ship if given an opportunity to do so. Without the RIF, a lot less people would even dream of looking for other positions. And the people who leave, are usually your best performers, not the next in line at the bottom of the barrel, as they are the ones asked to pick up extra work for no extra pay and reduced job security.
AMD already has some nice products - their APUs are in very high demand, and there is a reason for it. So, instead of focusing on what works (like fixing performance issues in "Bull-dozer" CPUs and further improving the APUs, maybe adding ECC memory support to their video engines so they have a better chance of catching up with NVIDIA in HPC applications), they want to go chase a low margin, already crowded market, in which they have no prior experience, like ARM. BRILLIANT! I can now see that the new CEO is really earning his pay....
Re ARM-64 servers - they look more like a solution looking for a problem, given the current state of virtualization and capability of moving workloads on demand across physical nodes, then a highly desirable, high growth and high profit margin market. They will certainly find some buyers, but percentage wise, compared to Xeon or Opteron based servers, I doubt they will amount to much. You need a very compelling $$$ argument for a new architecture, and for ARM, cost savings are questionable - the delta in CPU cost is a very small number compared to the total hw and sw cost associated with a new server, and power savings are largely negated by the advances in virtualization, given the delta in processing power between ARM and Xeon / Opteron CPUs. If you can consolidate 3-5 ARM servers on a single Xeon / Opteron server, I highly doubt there are any price or operating energy savings, and you have more possible points of failure.
AMD making ARM CPUs for phones or tablets - it does not make any sense, it makes a lot more sense to push for even lower power APUs which are x86 compatible (which they already have on the horizon).
The only reason ARM is on the map AT ALL, are low power battery operated applications where you can get by with fairly small computing power (anybody remember RCA 1802 CPU? - low power is a niche market ignored by most CPU makers, with low volumes, at least until smart cell phones exploded) . Intel stronghold on x86 architecture prevented low power x86 alternatives to be developed by others, which gave ARM a chance. This does not mean that ARM is a good chip for high performance computing, and once you get into server territory, you have to use performance enhancing tricks already patented by established players. Cache coherency and parallel out of order code execution have enough patents around them to make life really interesting for anybody new to the field.
Pushing for higher end CPUs is a must for AMD, but the latest CPUs performed poorly compared to their prior generation Phenom II, as they have much longer pipelines (like Intel's Pentium 4). They need to either fix the architectural issue (kind of hard, short of major surgery) or quickly increase the clock speeds. They seemed to have the right focus on this, at least before this news set the organization in turmoil.
I really like AMD, and used their CPUs in all my home computers for 20 years (since their 286 CPUs), but I am highly skeptical of this new direction. They may loose focus as their best people will be shifted away from core products to chase money-loosing propositions dreamed up by marketing drones and a new CEO eager to show that he is "doing something to fix the company" (clue - AMD is not a dog or a cat, and "fixing it" may require a different set of skills).
Here's your P45. Merry Christmas.
"that's 11.6 per cent, which is more than the "approximately" 10 per cent it announced in its press release."
Erm, 11.6 is pretty bloody close to 10. Just how close would one have to get in order to satisfy _your_ definition of approximate?