Citigroup Investment Research
This Forest Might Be Dryer Than Meets The Eye - Semi Cap Beat Vol. 83
Timothy M Arcuri
timothy.arcuri@citigroup.com
Brian K Lee
brian2.lee@citigroup.com
Atif Malik
atif.malik@citigroup.com
January 29, 2006 SUMMARY
* Many investors we speak with feel equipment stocks may be coming to the end of their rope. Indeed, the bear call now seems to be summed up by "enjoy 1H:06, because this is all you get". We disagree for three key reasons.
* 1) When normalized for chip unit production, front-end mfg tools have been this "under bought" only twice (June-03 and Sept-98) the past 13 years.
* 2) Using our fab-by-fab bottom-up supply model, we estimate that of the 14% y/y wafer capacity add in C2006, 5% is in NAND Flash. Thus, the remainder of the industry (foundry, DRAM, IDM) is adding only 9% y/y.
* 3) Foundry (fdry) orders. Using TSMC as a proxy, our work suggests its capital intensity (wafer size adjusted) is at 5-year lows - approximately in-line w/C2003 trough levels. Together w/ramp of many key leading-edge applications in C2006, specter remains high for a capacity "panic".
* Big gross margin leverage (NVLS) or fdry exposure (AMAT) the best themes.
OPINION
With the front-end equipment companies reporting strong order growth in the
DecQ (avg. ~20%) and guiding another double digit sequential order growth (avg.
~20%) in the MarchQ, investors are concerned if the cycle has "overheated" too
quickly. Bears who have proven wrong so far on their a) memory spending
"falling off the cliff" and b) disappointing C2006 capex (down 5% to up 5%)
calls are now saying "enjoy 1H:06 because this is as good as it gets". We
disagree with the bears for three reasons.
* Normalized (adjusted for wafer size efficiency) front-end orders per IC
unit at low levels. With respect to front-end orders, we look at orders
on a per chip unit basis. Put simply, the more units the industry
produces, the more wafers are required and the more equipment is
necessary to process the wafers. In the front-end, however, because a
dollar spent on 200mm equipment can produce fewer chip units than a
dollar spent on 300mm equipment (tools are ~30% more expensive, but
produce ~2x the chip units), we must adjust for wafer sizes. Figure 2
graphs these ratios back to C1991. The trend shows that front-end
equipment has been this "under bought" only twice (June-03 and Sept-98)
in the past 13 years supporting our view that front-end orders have room
to grow beyond 1H:06 with the rising secular capital intensity as a
tailwind.
Figure 2. Normalized Front-End Orders per IC Unit Trend (Note: Front-End
Adjusted For Wafer Size Efficiency)
Source: SIA, SEMI, Citigroup Investment Research
* Capacity add excluding NAND roughly in-line with C2005. Our fab-by-fab
bottom-up capacity model shows that a big component (~5%) of the overall
capacity addition in C2006 (~14%) is being driven by growing demand for
NAND flash market. Excluding NAND contribution the capacity is roughly
in-line with C2005 ex-NAND capacity addition of ~9-10%. The latest
industry forecasts calls for semiconductor sales to grow by roughly the
same amount ~8% y/y in C2006 as C2005. Hence, we believe that the risk
of overcapacity from non-NAND customer types (DRAM+IDM+Foundry) is
fairly limited in C2006. Figure 3 shows the breakout of capacity
addition by customer type converted into WFE spending for C2005-06.
Figure 3. WFE Spending By Customer Type 2005-06E
Source: Company Reports, Citigroup Investment Research Estimates
* Foundries still holding onto orders. Despite the double digit order
growth in the Dec and MarchQ, the equipment companies are still seeing
very little order activity from the foundries. Indeed our checks
indicate that TSMC is planning to place ~10k wsm orders out of a total
planned year-end capacity addition of ~35k wsm by March/April of this
year- which implies that 2H:06 could see plenty of foundry order action.
Figure 4 shows that indeed TSMC's normalized capital intensity (capex
adjusted for wafer size efficiency in 8-inch equivalent divided by its
revenue) is at five year lows- roughly in-line with C2003 trough levels
and is ripe for a "panic" capacity add. Foundry orders in conjunction
with 2H:06 second tier DRAM spending driven by Vista and gaming cycle
could provide the legs to a strong 1H:06. Our checks suggest equipment
makers are anticipating large DRAM orders from Elpida (E300 Phase III),
Inotera Fab2 (Phase II) and Samsung in back half of the year.
Figure 4. Normalized Capital Intensity for TSMC (Note: Capex Adjusted For Wafer
Size Efficiency)
Source: Company Reports, First Call, Citigroup Investment Research Estimates
FIGURE 5. GROUP PERFORMANCE RELATIVE TO NASDAQ COMP (3-MONTH)
Source: Factset
FIGURE 6. SEMICONDUCTOR EQUIPMENT VALUATION
Source: Company reports, First Call, Factset, Reuters, Citigroup Investment
Research estimates
Note: all P/S analysis for KLAC, NVLS, and LRCX is based on SHIPMENTS and not
REVENUE.
Note: Adj. P/E calculated w/stock price stripping out net cash, EPS excluding
net interest income
B=Buy, H=Hold, S=Sell, NR=Not Rated, N/A=Not Available
*ASML and ASMI covered by Navdeep Sheera and Katherine Thompson
VALUATION AND RISKS
Applied Materials (AMAT--$19.75; 1H)
Valuation
We rate AMAT Buy/High Risk (1H) with a price target of $22.
With sharply reduced manufacturing cycle times at equipments OEMs allowing
swift corrective action by chipmakers in response to excess inventories, our
analysis suggests that the semiconductor equipment industry downturn will be
shorter and shallower than any in recent memory. As our analysis suggests that
most front-end equipment companies are thus unlikely to lose money at any point
in the downturn, we believe that investors should focus on a metric that
accurately captures cross-cycle earnings power. We thus focus on "normalized"
earnings on the notion that the downturn will be more muted than any such
downturn in recent memory. If we average peak rolling four quarter EPS of
$1.12 (FQ1:07) and trough (fully taxed) rolling four-quarter EPS of $0.65
(FQ4:05), we arrive at a normalized full-year EPS estimate of $0.885.
We then note that the average 10-year historical S&P500 multiple (excluding the
1998-2000 Asian crisis/market bubble time period) is approximately 18.4x
earnings and has ranged between roughly 16x-23x during this period. As
semiconductor equipment capital spending is likely to continue growing in the
high single digits % CAGR (or roughly 2x global GDP), we assign a 15% premium
to the market for bellwether AMAT, and arrive at a multiple of about 21.2x
earnings.
Applying this approximately 21.2x multiple to our $0.885 in normalized earnings
power, we arrive at a fair value of approximately $19. As this is based on
cross-cycle normalized earnings, we consider this to be "fair value" for the
stock.
On a book value basis, the ten-year historical average is 5.0x (ten-year range
from 1x-7x). Applying the historical multiple to an estimated book value per
share at the end of CY2005 of $5.40 yields a price target of $27.
As our analysis of industry fundamentals lead us to believe that the downturn
will be short and shallow, we believe the most appropriate valuation metric is
one that takes into account this more muted cyclical downturn. However, as
many investors are still concerned about the more traditional price to book
method of valuing stocks, we do not wish to completely discount this. We thus
assign a weighting of 75% to our normalized earnings method and 25% to book
value. Therefore we arrive at a price target of about $22, and we maintain our
Buy, High Risk (1H) rating on the stock.
Because this methodology and result should be somewhat consistent regardless of
the point in the cycle, we would view this as a barometer of "fair value"
whereby meaningful deviations above or below should serve as stock triggers.
Risks
We rate AMAT High Risk, primarily due to stock price volatility as well as
earnings volatility. The following are key risk factors:
Our valuation methodology is based on the assumption that semiconductor capital
equipment cycle will exhibit a shallower downturn than previous cycles. As fab
utilization and capital equipment orders are closely linked to stock price, any
material differences to our supply/demand model (e.g., demand drops suddenly,
or supply increases more rapidly than we predict) may cause our valuation
methodology to be inaccurate.
Applied has undergone changes in its senior management structure, even two-plus
years out after installing a former Intel senior executive as CEO & President.
While our limited exposure to customer reaction suggests a positive view to
this change, we recognize the disruptions such changes in senior management can
cause within a firm and the potential impact to operations and/or order flow.
Applied is currently undergoing an extensive cost restructuring program in an
effort to align their infrastructure with their perception of future business
levels. Disruptions associated with this cost initiative or a mis-estimation
of potential future business levels could create difficulties in Applied's
ability to meet our forecasts.
As the market share leader in multiple sub-segments of semiconductor equipment,
Applied is uniquely positioned to capitalize on inflections in industry cycles.
However, at their currently elevated share position, we believe incremental
share gains become increasingly difficult. While we believe share shifts are
largely stable at this stage, we recognize the competitive pressure Applied
faces from its smaller peers.
Additionally, although we view AMAT's efforts to expand their service business
positively, we believe that the expansion in core responsibilities risks
alienating some customers who may view AMAT as taking over too much of the
market.
If the impact on the company from any of these factors proves to be greater
than we anticipate, the stock will likely have difficulty achieving our target
price.
Novellus Systems (NVLS--$29.53; 1H)
Valuation
We rate NVLS Buy, High Risk (1H) with a $36 price target.
With sharply reduced manufacturing cycle times at equipment OEMs allowing swift
corrective action by chipmakers in response to excess inventories, our analysis
suggests that the semiconductor equipment industry downturns will be shorter
and shallower than in the past. We thus focus on "normalized" earnings on the
notion that the future downturns will be more muted than past downturns.
For NVLS, our model suggests a cross-cycle EPS earnings power of $1.47. We then
note that the average 10-year historical S&P500 multiple (excluding the 1998-
2000 Asian crisis/market bubble time period) is approximately 18.4x earnings
and has ranged between roughly 16x-23x during this period. As semiconductor
equipment capital spending is likely to continue growing in the high single
digits % CAGR (or roughly 2x global GDP), we assign a 10% premium to NVLS and
arrive at a multiple of about 20.2x earnings. Applying this 20.2x multiple to
our $1.48 normalized earnings estimate, we arrive at a value of approximately
$29.70.
On a price/book value basis, the ten-year average multiple is 3.78x (ten-year
range from 2x-8x). Applying this multiple to our estimated book value per
share at the end of C2006 of $14.24, we arrive at a price target of ~$54.
As noted above, we prefer to focus on "normalized" earnings on the notion that
the downturn will likely be more muted than any such downturn in recent memory.
However, as many investors are still concerned about the more traditional
price-to-book method of valuing stocks, we do not wish to completely discount
this.
We note the large discrepancy between the implied price targets from book value
and normalized earnings. Although we choose to take book value into account,
we note that it may be skewed due to past bubble valuations, and we believe
that normalized earnings will be the primary driver of the stocks moving
forward.
We assign a weighting of 75% to our normalized earnings method and 25% to book
value and arrive at our price target of $36 and Buy, High Risk (1H) rating.
Risks
We rate NVLS High Risk, primarily due to stock price volatility as well as
earnings volatility and uncertainty in terms of product market share at the
90nm technology node. The following are key risk factors:
1. Our valuation methodology is based on the assumption that semiconductor
capital equipment cycle will exhibit a shallower downturn than in previous
cycles. As fab utilization and capital equipment orders are closely linked to
stock price, any material differences to our supply/demand model (e.g., demand
drops suddenly, or supply increases more rapidly than we predict) may cause our
valuation methodology to be inaccurate.
2. NVLS' primary competitor is AMAT, the industry bellwether. While
Novellus has a long-term track record of share gain, fighting off competitive
pressure from Applied Materials is a daily occurrence. While we are encouraged
by Novellus' early lead and disproportionately high share in copper
electroplating, we recognize the likelihood of share gains by Applied Materials
with the introduction of its 2nd generation copper ECD tool, SlimCell.
3. Novellus recently acquired Speedfam-IPEC, a manufacturer of CMP
equipment. This acquisition comes on the heels of the acquisition of Gasonics.
While Novellus has been successful in integrating Gasonics and is now gaining
back lost market share, we believe the challenges facing the CMP division are
greater, as AMAT now dominates the CMP market.
4. Novellus has historically had one of the highest margin structures in
the industry, a result of the company's leading edge products and highly
outsourced manufacturing strategy. Particularly given the above-noted mergers,
we recognize some risk to the company's ability to repeat their peak margin
structure. While we still expect Novellus to show industry-leading margins,
investors may be disappointed in the inability of the company to improve upon
previous cycles' profitability.
Factors including, but not limited to, recovery in CVD market share,
incremental gain in PVD/CMP share, and/or improvement in operating margins
could all cause the stock to exceed our price target.
If the impact on the company from any of these factors proves to be greater
than we anticipate, the stock will likely have difficulty achieving our target
price.
APPENDIX A-1
I, Timothy Arcuri, research analyst and the au |
