Continuing with my previous articles on semiconductor demand (See earlier articles on National Semiconductor (NSM) and Texas Instruments (TXN)) and its implications, I wanted to focus on the Semiconductor Assembly and Testing Services (SATS providers) space this time around. SATS providers are also called 'back-end' service providers in semiconductor industry parlance since 'assembly and testing' constitute the back-end processes in the entire wafer-to-chip manufacturing process, while wafer etching and other processes where the chip design is 'embedded' on the wafer constitute the 'front-end' processes.
Silicon Precisonware (SPIL) is among the largest players in the SATS space and I will focus on SPIL's recent performance and use that to show how the SATS industry could face testing times despite visible signs of recovery in semiconductor chip volumes, their primary revenue and profitability driver.
Semiconductor demand can be volatile and history shows that
A bit of history is always useful to get a better perspective of the current scenario and so here we go:
As per Semiconductor Industry Association (SIA) data, the total semiconductor capacity (measured in wafer-starts in thousands per week-WSpW* and all numbers are normalized to 8" equivalent wafers) had increased from ~1300 WSpW in Jun03 to ~1700 WSpW in Mar06. This translates to a 30% increase over two years. Utilization remained in the healthy 85-95% range despite this capacity increase and SATS providers also expanded significantly during this period.
(*All WSpW numbers are in '000's through the note)
Exhibit I: Semiconductor Wafer-Starts (WSpW) from Jun03 to Mar06
(Bigger bar represents 'available' capacity and smaller bar represents 'used' capacity and the line graph represents capacity utilization.)
Now the capacity addition continued at a healthy pace from Mar06 till 2008 as utilization was still around the ~90% range and by Sep08, a peak capacity of ~2400 WSpW was reached. But the slowdown since Oct08 reduced demand and capacity utilization fell off a cliff.
Exhibit II: Semiconductor Wafer-Starts (WSpW) from Jun06 to Mar09
All SATS providers have their revenue directly related to the total chips fabricated and that is well reflected in the chart below which shows SPIL's revenue, gross profit and capacity (in terms of PP&E $'s) over the last six years 2003-08.
Exhibit III: SPIL- Revenue, Gross Profit and Capacity (in PP&E dollars) from 2003-2008
Source: Gridstone Research (Please note that PP&E values are plotted on RHS while Sales and Gross Profit are in LHS and scales vary. All figures are in NT$- New Taiwan Dollar)
The formula was simple: More chips assembled and tested means more utilization and more revenues and profits accrued. Since assembly and testing is not exactly a high-value add activity, volumes and effective capacity utilization hold the key to good financial performance. Revenue and profits are directly linear to wafer-starts as more wafers meant more demand for assembly and testing services. As a result, SATS players like SPIL added capacity to keep pace with increased chip volumes. This was a good strategy as long as the incremental capacity investments lead to sustained profits over the life of the capacity investments. But with the slowdown that has not happened in 2008 and looks set to continue in 2009 too.
Exhibit IV: YOY change in Revenue, Gross Profit and PP&E,net
Source: Gridstone Research
High fixed costs mean that returns have slowed down dramatically
Note how the gross profit figure has plunged with the decline in sales. With semiconductor volumes picking up since April 2009, should SATS players like SPIL feel happy? Well, they can feel optimistic but not happy yet. The table below shows the returns which SPIL has generated over the past five years. ROA (Return on Assets) increased during the 'boom' years of capacity additions and better utilization of increased capacity. But in 2008, ROA and ROE (Return on Equity) declined dramatically to approach the 2004 lows.
Exhibit V: Return ratios near historical lows
Source: Gridstone Research (ROA,ROE and PP&E have been shown in US$ to enable better reference for ADR investors though SPIL reports primarily in NT$-Taiwan Dollar)
Gross margins have also returned to the 2004-05 levels and unless utilization improves dramatically there is little scope for gross margins to reach the high 20's which SPIL achieved in 2006-07. Notice that capacity investments (in PP&E dollars) have increased over this period and with the current slowdown, the 'effective' utilization of such productive assets over their useful life is a big risk for SPIL going forward. The table below gives the quarterly returns generated for the last six quarters and clearly the challenge lies is improving the utilization of assets(while keeping pace with technology change through capacity additions) thereby enabling increased ROA and ROE. Asset utilization had fallen to abysmal levels in Mar09 quarter such that there was a real threat of 'idling' many of these productive assets for prolonged periods.
Exhibit V: Return ratios detoriated significantly in last few quarters
Source: Gridstone Research (ROA,ROE and PP&E have been shown in US$)
'Productive' Assets could still suffer shut-downs or offer lower returns as SATS providers don't have pricing power
Actual Wafer-starts at 1200 WSpW in Mar09 were at the lowest levels in the last six years and we need to go back to Jun03 to get such similar wafer-start numbers. However available capacity has almost doubled since Jun06 to the period of Dec08 and with many capacity shutdowns by various foundries and manufacturers, available capacity stood at ~2200 WSpW at the end of Mar09. The SATS players have all added 'assembly and testing' capacity over the years to cater to this demand i.e ~2200-2400 WSpW, but with demand in Mar09 quarter at ~50% of capacity, they are clearly having excess productive assets as of date. Only a 60% increase in chip volumes (which means WSpW increases to ~1920 WSpW and utilization could be ~85%) would result in effective utilization of such assets. As such, SATS providers have very low pricing power in this demand environment and thereby have to desperately seek volumes to generate sufficient returns on their assets. A price boost for services rendered looks highly unlikely (or rather impossible!!!).
It is anybody's guess as to when we will see such a sustained increase in volumes. Neverthless, the rough calculations above show that a 20-30% pick-up in demand from the bottom of Mar09 will just not do. Also, slower the pace of demand pick-up, the lower the utlization for prolonged periods which means that SPIL and other SATS players will have two options:
- Report very low ROA/ROE ratios for the next 2-3 years since demand may not be sufficient for effective utilization of assets or
- They should take a big-writedown in asset values (through capacity shutdowns) to boost the ROA/ROE in subsequent periods post such one-time writedowns. If the useful life of the 'assembly/test' equipment is the same, with decreased utilization, it makes sense to reduce 'productive' capacity. With depreciation accounting for ~15-25% of COGS (Source: SPIL 10-K dt. May 18,2009) or ~10-15% of revenues, the returns (ROA/ROE) will be much better once excess capacity is shut-down.
Industry Consolidation looks inevitable
Option 2 could also be pursued through industry consolidation. In such low-value add service industries, the only way to maintain some semblance of pricing power during such low-demand periods is to have lesser capacity and lesser service providers overall. This would also ensure that capex expansion across too many players simultaneously does not lead to the situation the current lot of SATS providers are in. The other possibility is for big foundries like Taiwan Semiconductor (TSM) which offer both front-end(Wafer fabrication) and back-end(assembly and testing) services to integrate such back-end only service providers.
And current surge in SPIL ADS price looks excessive
The challenge for SATS providers lies in the capacity-demand conundrum which tests the financial muscle of all semiconductor manufacturers. Capacity addition can happen in blocks only and is capital-intensive and irreversible. But demand can be volatile and the pace of technology change is rapid in semiconductors leading to periods of excess capacity (which affects bottom lines as much as the reduced demand). Dodging all these pit-falls and increasing investor returns in the process is a herculean task for SATS providers and in this context, the run-up in SPIL's ADS price in 2009 seems excessive.
With an Earnings per diluted ADS of $0.01 in 1Q09 (Mar09), $0.32 in FY2008 and $0.87 in FY2007, the current price($6.24, Jun 15 close) translates to a TTM P/E of ~26 when even semiconductor bellwether, Intel Corporation trades at TTM P/E of ~20 (based on Jun 15 close of $15.98). Though these are TTM multiples, SATS providers are at the mercy of chip OEM's for volumes and it is difficult to imagine a scenario where they would outdo IP owners like Intel in terms of financial performance or have better prospects in the near future. Only expectations of consolidation and a buy-out of SPIL by large foundries can justify such multiples. Hopefully, the industry players would also see it this way!!
Source: Google Finance
Disclosure: No positions