By Christian Gregor Dieseldorff, SEMI
SEMI’s World Fab Forecast tracks money spent on frontend semiconductor fabs that are equipping, including R&D and pilot fabs, fab construction projects, and capacity investments per fab. Unlike the SEMI Worldwide Semiconductor Equipment Market Subscription (WWSEMS) data, the World Fab Forecast and its related Fab Database reports track any equipment needed to ramp the fab, upgrade, expand, or change its wafer size regardless if it is new equipment, used equipment, or transferred equipment, while WWSEMS tracks only new equipment.
Money spent on fabs equipping: Lowest level in 10 years
A preliminary analysis of the World Fab Forecast (Dec. 16, 2008) reveals that spending on equipping of frontend semiconductor fabs (including R&D and pilot lines) is expected to be about $26 billion in 2008, a change of -31% to 2007. Given recent announcements since early December, the data predict 2009 equipment spending could decline by 30%-40%, and to drop down to the $16B range (see Figure 1) — resulting in the lowest spending level over the past 10 years.
Figure 1: Money spent on fabs equipping by quarter by region.
For 2008, Japan is expected to spend the most on fab equipment compared to other regions at $6.0B. This number is expected to drop by over 50% to less than $3.0B in 2009. Korea will be the second largest equipment market in 2008, though the market is expected to decline about 40% to $2.7B in 2009. The largest spenders in Japan in 2008 and 2009 are the flash JVs between Toshiba and Sandisk, Elpida, Toshiba, and Fujitsu. The Flash Alliance alone is the second largest spender for equipment, second only to Samsung in 2008 — even with cutbacks in capital spending. This is also projected to be the case for 2009.
Although most other regions show double-digit contractions in 2009 for capital equipment, Europe and the Mideast show flat to low positive growth. However, the absolute dollar amount is much less, about $2B compared to other regions such as Korea, Japan and the Americas (who are in the $2.7B-$3.0B range). But the contrast is still remarkable. Europe and the Mideast have had steady spending in the range of $2.0-$2.6B since 2007. The European semiconductor industry does not focus on memory, so non-memory makers — such as Intel with Fab 28 (Kiryat Gat, Israel) and AMD (Dresden, Germany), which recently received a financial injection from investors in Abu Dhabi — are expected to drive this positive growth rate.
Taiwan will experience the largest cutbacks in money spent on fabs equipping: from about $9.0B in 2007 to $4.5B in 2008 and an additional reduction of 53% in 2009 to reach about $2.0B-$2.5B on equipment. Equipment spending in almost all other regions is expected to drop by 20% to 50% in 2009.
Figure 2 shows spending on equipping of frontend semiconductor fabs on a year-by-year basis. 2009 is predicted to decrease in the range of -30% to -40%. This is measured against an already weak year 2008. If we compare the spending numbers of 2009 against 2007, we get a decline of 60%.
Figure 2: Money spent on fabs equipping by year.
Money spent on fab construction projects down 32% in 2009
In 2008, out of a total of 10 major construction projects, five are for new fabs and four are for 300mm fabs. In 2009, out of a total of at least 16 known planned fab construction projects, only six (four of which are for 300mm volume fabs and smaller lines) have a higher probability to begin construction by mid- to second half of the year.
Actual spending on major fab construction projects for frontend semiconductor fabs (including R&Ds and pilot lines) in 2008 is expected to decline by 41% year-over-year as more projects are pushed out or put on hold (see Figure 3). Construction spending in 2009 is expected to decline about 32%, to reach its lowest level in six years.
Figure 3: Money spent on fab construction projects by quarter by region.
The Americas and Japan stand out with positive growth rates in construction spending for 2009. Japan is expected to spend twice as much as the Americas, mainly driven by Toshiba, the Flash/Alliance JV, and Panasonic. Worldwide, Toshiba and the Toshiba/Sandisk Alliance represent the largest spenders on fab construction projects in 2009. In the Americas, AMD has announced the construction of its $4.0B fab in upstate New York, which is receiving an additional $1.2B in incentives from the state. This will be the only new high-volume fab project beginning construction in the Americas over the past three years. The last new high volume fab in the Americas was Samsung’s 300mm fab in Austin, Texas, which began construction in 2006.
Capacity growth at lowest levels since 2002
The World Fab Forecast tracks and forecasts out over six quarters capacities company-by-company and fab-by-fab. From 2003 through 2007, semiconductor fab capacity increased near or above double-digit rates annually, but as a result of global economic uncertainty capacity growth rates will be much lower for both 2008 and 2009. Wafer fab capacity is forecasted to reach less than 16 million wafers/month (in 200mm-equivalent wafer size) by the end of 2009 compared to 15.3M wafers/month capacity expected by year-end 2008.
In 2008, we expect overall fab capacity to grow by about 4.5%, and in 2009 to grow by about 3%-4%. These growth rates are the lowest since 2002. The growth rate for 300mm fabs in 2008 is the lowest since devicemakers begin to invest and ramp 300mm volume fabs, with the 300mm capacity ramp slowing even more in 2009.
In response to the economic crisis, oversupply, and falling average selling prices (ASP), most memory companies are closing their 200mm fabs. However, some companies maintain smaller but still positive capacity growth rates of their 300mm fabs. For 2009, foundries are expected to maintain the strongest capacity growth rate, about 7%-8%, followed by MPU and memory.
Foundries make drastic cuts in utilization
Utilization rates in 4Q08 for foundries are expected to reach the lowest levels in years. For example, UMC’s utilization rate may drop as low as 55% in 4Q08, the lowest since 2001. Chartered utilization rates for 4Q08 may drop to 58% (down from 85% in 3Q08). We expect utilization rates to remain low through first half of 2009.
As of mid-November 2008, most foundries reiterated their planned capex for 2008. Only UMC has announced cuts in its 2008 capex from $600M to $450M. SMIC has suspended expansion plans in 2009 and slashed its 2009 capex to $200M from $790M in 2008. At of the beginning of December 2008, Chartered cut its guidance for fourth-quarter sales to $343M-$353M million and a net loss of $76M-$84M, but the foundry has not yet announced any cuts of its 2008 capex plan (currently $630M). We expect that Chartered also will cut its 2009 capex. TSMC is expected to cut its 2009 capex by about 20%, and has also announced hiring freezes and asked employees to take unpaid leave to cut costs.
How the outlook changes
At the beginning of 2008 and even into mid 2008, the outlook for 2009 looked positive, but this changed very quickly (see Fig. 3) in light of the current economic conditions. In mid-December 2008, Gartner revised its 2009 semiconductor revenue growth to be -16.3%, down 14.1 percentage points from their previous 4Q08 forecast.
Our data tracking the money spent on fabs equipping follows the same trend over the past quarter (Figure 4), but the downward revision is more severe as money spent on equipping fabs is for future capacity growth. With extremely challenging economic and semiconductor conditions forecasted for 2009, device makers have significantly cut investments in new capacity.
Figure 4: Worldwide spending on equipping frontend fabs.
Since the burst of the “dot.com” bubble, the semiconductor industry has become more dependent on consumer products, and therefore on the consumer. The entire industry, and especially the memory segment, has experienced falling average selling prices since 2007, mainly due to oversupply. Yet, the memory segment outlook was still positive because of increasing demand forecasts due to high consumer confidence.
The collapse of the credit market and mortgage bubble in September 2008 changed all that. People who are losing their homes are not likely to spend money on the latest gadgets, so prospects for consumer spending have dropped, and with it, the demand of electronics and semiconductor devices.
What the future holds
As the economy continues to slide into a deepening recession, more and more semiconductor companies are posting losses and are forced to make cuts.
For the memory industry, the first half of 2009 may look even worse. Companies will be forced to execute different cost-control strategies such as layoffs, pay cuts, capacity reduction and employees unpaid leave. However, these strategies may not bring any benefits if the consumer confidence continues to be low and consumer spending stagnates.
Companies without large cash reserves are struggling and vulnerable. We can expect more mergers, such as Micron’s purchase of Qimonda’s share in Inotera (mid-October 2008); Panasonic’s efforts to acquire Sanyo (November 2008); and other rumors such as Toshiba buying Sandisk and Micron acquiring Qimonda. Mergers will create redundancy, which generally leads to more layoffs.
The Sandisk and Toshiba joint venture companies, Flash Partners and Flash Alliances, are facing severe price erosion and revenue decline. These joint ventures, among the largest flash memory makers and with the largest fabs in the world, watched revenues sink from 20% growth in 2007 to a 6% decline in 2008 (YTD 3Q08). On Dec. 16, 2008, they announced a halt in production of their Fab 3 and Fab 4 in Yokkaichi, Japan scheduled from Dec. 31 to Jan. 12, after which they will continue production at 70% utilization until market conditions improve.
All these factors may make the first half of 2009 even weaker than 2008 for the semiconductor industry. This downturn will last for a while. A recovery will most likely be in 2010, but it will be slower than we experienced before given the broader economic factors.
Light at the end of the tunnel
There is still activity for new fab construction and new fabs coming online in 2009. As mentioned above, six new fab construction projects are projected in 2009 with higher probability beginning construction, four of which are 300mm facilities. Out of 12 known fabs originally planned to come on-line in 2009, 4-5 are very likely to begin ramp (at least one is a 300mm volume fab), while the remaining projects will likely be delayed.
Two big companies in our industry still hold strong positions but are not unaffected by the current market. Samsung still has $7.6B in cash reserves as of October 2008, but they are not invincible. We expect Samsung to cut its 2009 capex by 30%-50% and have capacity growth at single digits if not flat. Intel cut its 4th quarter outlook from $10.1B-$10.9B down to $8.7B-$9.3B; job cuts may be evident in 2009.
But there is one company that suddenly stands out with a move that was unexpected for many at the beginning of 2008 — AMD, which made a surprising announcement involving outside investor ATIC from Abu Dhabi. AMD’s fab plans received a huge financial boost, which spurs investment at its fab in Dresden and in Luther Forest in upstate New York where the company plans to build new fabs. The Luther Fab alone, dubbed Fab 4x, has an estimated investment of $4.6B and will create over 1,400 jobs.
In times of struggling markets and economic recession, new ideas are born and new opportunities are created. The oil crisis sparked increased interest in renewable
energy and energy conservation, such as the solar industry. Additionally, the population continues to grow and consumables must be replaced, which means that there is a natural demand. Severe cuts in spending will create an economic vacuum which will need to be filled. Once we reach the bottom, we will see growth again. The semiconductor industry experiences severe slowdowns about every 5-8 years, especially after exuberant spending results in oversupply. At the moment all data indicate that we see growth again in 2010.
SEMI’s World Fab Forecast report provides high-level summaries and graphs; in-depth analyses of capital expenditure, capacity, technology and products, down to the detail of each fab; and forecasts for the next 18 months. These tools are invaluable for understanding how 2009 will look, and learning more about capex for construction projects, fab equipping, technology level, and products. Visit www.semi.org/fabs for additional information on these reports.
Christian Gregor Dieseldorff is sr. analyst and director of market research for SEMI’s Industry Research and Statistics Department, San Jose, CA