Semiconductor Equipment . . . A Fahnestock analyst has published a note providing a "cautious" outlook on 2nd quarter orders in the semi cap equipment space. In the note, the analyst cites concern over recent rally in the semi space as a contributing factor to rich valuations despite lackluster chip demand.
Semiconductors . . . Merrill Lynch expects Microchip to be slightly more conservative in its outlook during its mid-qtr update next week. The firm cuts 2004 estimate to $0.82 from $0.85 (consensus $0.85), and lowers their revenue growth forecast to 13% from 16%. The firm sees fair value at $22.
Semtech announced it would be supplying power management chips for AMD's next generation 64-bit processors. These chips will be included in AMD's PC and server markets.
Semi Comments . . . This quarter is bearing a watered-down resemblance to 1st quarter 2002, when many specialty chip companies' business was tracking slightly ahead of plan -- while customers sounded hesitant, at best. It is a little troubling when even weak companies and weak product lines within strong companies are seeing an improvement in business despite no pick-up in end markets. Up until this week, thought that the relatively positive (albeit muted) data points coming from several specialty chip companies were a combination of a few factors, including 1) low expectations, 2) positive seasonality, and/or 3) company-specific factors. We are now adding a fourth factor: there is probably a mild inventory effect helping specialty IC suppliers after customers had (in hindsight) overly compressed inventories in 4th quarter 2002.
To be clear, we are not saying that all of the recent improvement is due to inventory effects -- it is one factor out of several. In addition, this is a pale echo of the inventory-driven circumstances of a year ago. That said, most analysts think that it is appropriate for investors to apply a discount rate to the positive updates emanating from the chip portion of the supply chain.
This note is comprised of two sections, company-specific comments plus larger macro themes (i.e., 'Spring Hopes Eternal', Customer Inventory Analysis and Comm IC market estimates).
Similar to the quarterly business updates from peers Altera and Xilinx, #3 PLD supplier Lattice also positively revised its 1st quarter sales guidance on Thursday after the market closed. Management's initial outlook for the March quarter forecasted revenue to be flat to slightly up on a sequential basis. Based on quarter-to-date business trends, the company now expects sales in March to be at the higher end of that range.
#4 PLD supplier Actel also provided its update after the close on Thursday, and reiterated its guidance of flat to very slightly down.
Realtek Andrew Lu reports from Taiwan that Realtek's business has bounced very strongly in March. Among the reasons is Realtek's WLAN chip business, which is set to triple MoM in March. Specifically, Realtek expects to triple its WLAN 802.11b chipset shipments in March to 150,000 from nearly 50,000 a month in Jan and Feb. In addition, Realtek reiterates its WLAN IC shipment target of 330,000 per month in 2nd quarter 2003, versus Andrew's estimate of 200,000.
By the way, any chip investor that hasn't taken a serious look at Taiwanese fabless firms like Realtek and Mediatek had better get acquainted with them. These companies are important technology firms and are serious competitors to many U.S.-based suppliers such as Broadcom, LSI Logic, Marvell and others.
Among other data points, Agere has been citing its 802.11b design win in the Dell Latitude notebooks as evidence that it is not losing its grip on the WLAN market. This week's announcement that Dell is planning to use 802.11a/g chipsets from Broadcom in the Latitude D-family is a sign that Agere's lack of '11a' and '11g' products is hurting its position with major customers. In addition, the fact that a company with the stature of Dell is comfortable pushing a pre-standard wireless technology (i.e., '11g') suggests that adoption of '11g' could happen rapidly. The faster '11g' ramps, the more negative it is for Agere.
Another factor to consider about Agere is its roughly 30-35% sales exposure to the hard disk drive (HDD) market. As we describe in the next section about Marvell, there is reason to be concerned about recent developments in the HDD market. Given Jon Joseph's concerns about the desktop PC supply chain and the fact that Agere's HDD sales are predominately to desktop- oriented manufacturers such as Maxtor, this is a fundamental area to watch for Agere.
Agere's current quarter is on track, but the combination of HDD and WLAN issues could become a factor by 2nd quarter. For Agere's stock, we believe these concerns do not carry the same weight that they might have for other stocks. Nevertheless, shareholders are advised to be aware of the potential risks in this situation.
According to a story Thursday on Briefing.com, hard disk drive (HDD) stocks started the day under pressure due to reports of aggressive IDE pricing for 80-120G by IBM (it is probably multiple culprits). According to a web-based price source that we track, pricing on 80-Gig drives dropped $1 overnight on Wednesday to $77. Prices on the site were in the $82 range two weeks ago. Another site listed 80-Gig drives at $75 -- although analysts have not been tracking that source in the past, so tough to know the relative change.
Since the beginning of the year, we have been wary of a HDD channel inventory correction affecting Marvell's business. So far, that has not materialized. Given the recent declines in HDD pricing and the recent slowdown in the PC motherboard business (see Kirk Yang's note from earlier in the week), it is possible that this risk is beginning to take form. As a semantic matter, we wrote a week ago that we believe that Marvell's claim of 100% market share at Western Digital for 80-Gig platters may be overly aggressive. To be more precise, we believe that Marvell currently has 100% of the read channel chip business for WD's 80-Gig platters at this early stage in the product cycle, but that state of affairs is unlikely to last.
According to multiple industry contacts, WD is evaluating other sources (i.e., Agere and ST Micro) for read channel and SoC devices. In addition, WD is unlikely to use 80-Gig platters in 100% of its drive platforms anytime soon. That is, while the company has stated that it will be capable of shipping 100% of its drives with 80-Gig platters by the end of June, that does not imply that it will literally be doing so. For example, it may make economic sense for drive capacities of less than 80-Gigs to use lower density platters.
Bottom line: we believe that it is a miscalculation to multiply WD's unit shipments by Marvell's ASPs to estimate the amount of business Marvell can derive from WD. Separately, Briefing.com reported on Wednesday that a marketing executive had left the company. At the time, we the story referred to a VP of marketing in the communications business group who had left a m nth ago.
Upon further checks, it turns out that two additional marketing managers responsible for the Prestera switch products have left the same group, one for a start-up and another for Broadcom. We have not written much about Prestera because we do not see the product as pivotal for Marvell. As such, do not see this development as big news. Those who might have had greater expectations for Prestera might take this news as a negative data point.
On to a third subject, analysts sometimes get requests for our estimates about the shape of Marvell's revenue with Intel since it is generally recognized that Intel will be moving away from Marvell over time. Intel will peak as a customer for Marvell in the April quarter and fall fairly steeply from that point forward. Marvell's management believes that its business with Intel will not be affected until the January 2004 quarter. To be blunt, the company could be right and we could be too aggressive about the timing of the Intel effect. Marvell is counting upon the failure of Intel's CSA which is an assumption we are unwilling to make. Springdale-based PC announcements over the next weeks and months will help to define the success or failure of CSA.
This quarter is beginning to bear a watered-down resemblance to 1st quarter 2002, when many specialty chip companies' business was tracking ahead of plan – while their customers sounded more hesitant. In 2002, the first half of the year was helped by inventory replenishment at customers - that quickly dissipated. Given the provisional relative improvement of many of the same chip companies (e.g., Cypress, PMC-Sierra, Xilinx, Altera, and Maxim at this point, it is worth considering some of the similarities and differences between this year and last.
Median and total sales grew 5-6% sequentially in 1st quarter of 2002. This year, analysts are forecasting a 2% overall decline in 1st quarter (up 1% on a median basis). IF there is some further upside reported at the end of 1st quarter 2003, it is unlikely that median sales will be up more than 2-3%.
Customers were more optimistic about a 2nd half recovery a year ago, while there is little talk of that today. As a result, there is lower risk of inflated backlog or inventory buildup by customers. Customers still sound fragile - in fact, business for a few major OEMs (such as Cisco sounds more problematic than a year ago (and more problematic than a month or two ago).
The slight edge over guidance that many specialty IC companies are experiencing seems to be fairly widespread. Analyst find it a little troubling when even weak companies and weak product lines within strong companies are seeing an improvement in business despite no pick-up in end markets. There is early evidence of a potentially widening discrepancy between monthly foundry sales and out-sourced assembly and test (OSAT) trends.
Fundamental differences in factory cycle times between foundry and OSAT companies account for much of the difference between the two supply chain segments. Foundry cycle-times are about 4-8 weeks depending on geometries and metal utilization. Cycle times through OSAT lines are about one to two weeks. The total cycle time from wafer start to completed chip is therefore 5-10 weeks.
Thus, semiconductor companies must start wafers for new and existing products in advance of demand due to production cycle-time considerations. However, many chip companies hold wafers in die bank (the step between the fab and assembly) until demand actually materializes. A 'delta' graph is a good tool for tracking the relative rate of change in die bank - and rising die bank is an indication of optimistic forecasts that have not yet come to fruition.
In the 1st half 2002, there were fairly large increases in wafer demand from companies within coverage universe - beyond the rate that was being seen by OSAT companies (as implied by the huge spike in the graph one year ago). Some of the increases were based upon new product ramps at companies like Xilinx and Marvell. Other increases were simply driven by apparent improvements in end demand that was witnessed across most of our coverage space last spring.
However, chip demand fell off again in the summer and fall of 2002, wafer orders dried up, and we saw a large negative delta between wafer production and OSAT growth. The wafer production improvement last year at this time was a false positive leading indicator for semi demand. Currently, wafer production appears to once again be out-growing OSAT production. At this point, this may simply be a rebalancing of the supply chain after a rather steep fall-off in wafer production in the second half of last year. However, advise investors to watch the relative foundry production changes against OSAT changes. A widening gap may spell trouble.
Paradox: How can chip suppliers positively benefit from inventory effects when customers are still bleeding down inventories? At the moment, two paradoxes exist in the specialty IC space: 1) business trends appear to be improving gradually at several specialty IC suppliers, yet end-market conditions remain challenging -- at best; and 2) How can chip suppliers positively benefit from inventory effects when customers seem to have no inclination to build inventories?
It is possible that a few chip customers may have a slight rise in inventories this quarter as some might have run things a little too lean at the end of 4th quarter. However, expect aggregate customer inventories at the end of March to decline for the eighth quarter in a row. With chip delivery lead times short and chip price trends favorable to customers, there seems to be little incentive to build component inventories at this juncture. In any case, this climate cannot last -- either end-market conditions get better or the improvement in Specialty IC trends will stall. Yet there is a fundamental reason why the paradox exists at this point in time: the Specialty IC inventory depletion cycle has nearly run its course, resulting in an 'order rebalancing' by customers. Inventory effects are by nature transitory, and the idea is not a sustainable business driver. Nevertheless, believe it is important to analyze the fundamental underpinnings of inventory effects to better reconcile rising Specialty IC sales with at best flat near-term equipment sales.
This situation bears some resemblance to a year ago. In fact, the first half of 2002 for IC companies is largely remembered as an inventory restocking head fake. Yet, go back and look at the data: there does not appear to be a single customer that built inventories one year ago. Here is the answer to the riddle: if customers overall slow their inventory depletion rate, it can seem like better business for chip companies. The following analysis is narrow in its scope to the communications OEM and EMS areas, but the general principles can be applied to most end-markets. In the nearby graph, we compare intake and consumption levels for the top communications OEM vendors and the EMS companies. Use the terms 'consumption' and 'intake', which are not traditional financial terms but best describe the material flow through the supply chain. 'Consumption' is roughly OEM or EMS COGS but also adjusted for written-off inventories that were actually used. 'Intake' is the sequential dollar change in OEM or EMS inventories (adjusting for write-offs) plus consumption. So, when customers reduce inventory levels, intake falls below consumption, while intake rises above consumption when customers build inventories. (The consumption line would mirror the intake line when inventory levels remain unchanged).
Intake rates have remained below consumption rates since AMJ/01as low as 79% of consumption. The closer intake is to the rate of consumption, the slower inventories are burned. Over the last several quarters, intake rates have been slowly rising to 92% of consumption in OND/02. We estimate that the intake gap will close to 96% of consumption in JFM/03. To put it another way, for every dollar of material that a communication OEM or EMS manufacturer used, suppliers shipped $0.79 in AMJ/01, $0.92 in OND/02, and an estimated $0.96 this quarter. So while the end demand picture has not recently improved, the intake gap to consumption has continued to narrow and specialty IC sales have shown a modest improvement relative to end-market demand in the current quarter.
Since several specialty semiconductor companies are guiding to sales growth for March, signs of rebalancing activity are already evident. Belief is that communications OEMs and their EMS partners, while not increasing their own inventories, have increased chip orders to bring intake closer into balance with consumption. At an estimated 96% of consumption in JFM/03, inventory rebalancing has almost run its course, however, and further improvements in Specialty IC sales must be fueled by end-market growth.
Expect that communication IC sales of all types will rise roughly 5% in 2003 versus our previous estimate of 8% (made in August, 2002). Most of the change in forecast is driven by lower expectations for infrastructure-oriented chips. As a partial offset, Gigabit Ethernet and WLAN networking segments are tracking well above our expectations of 8/02, despite on-going price pressures. Note that without the anticipated growth in mobile phone chip sales, communications sales would be lower by a few percent in 2003. Overall, communications accounts for roughly 21% of the total semiconductor industry.
As a point of reference, our bottom-up models for the specialty IC suppliers indicate that the underlying communications IC sales growth implicit in our models is down 6%, predominately because of the relatively low mobile handset exposure in our specialty IC coverage.
Split the communications IC space into several major segments, including wireless terminals (roughly 51-54% of the total mix), Enterprise/SOHO networking (about 21%), wireline infrastructure (about 9-11%), wireless infrastructure (6-7%) and 'other' communications (9-10%, including terrestrial and satellite broadcast equipment, cordless phones, pagers, et cetera).
Wireless terminals (mobile handsets) Wireless terminal IC growth forecast of 11% for 2003 is largely driven by the market forecast of about 11% unit growth and our expectation that a rising mix of 2.5G and color handsets will allow the dollar content of chips per phone to remain roughly flat for the year.
There is concern about potential production overruns and inventory corrections in the handset market, particularly in China. Significant portions of the market appear to be adopting 'no-frill' phones, which could be damaging to ASP trends.
The rising importance of '2nd tier' handset suppliers in Asia is creating opportunities for a new wave of chip entrants (e.g., Broadcom, Silicon Labs) to gain market traction without waiting to crack one of the classic 'majors'.
Service providers appear interested in pushing 'data oriented' 2.5G phones regardless of the generally lackluster acceptance of wireless data services.
Wireline infrastructure (telecom equipment) is expected to suffer another year of overall decline. Component inventory corrections have subsided, however, so the component business should more closely mirror end-market trends rather than under-performing the end market as in the last two years. Based on these considerations, anticipate a 10% IC revenue decline this year.
Will it ever end? Some ILECs are threatening to cut spending following the FCC's recent UNE-P decision. Expectations are low. Pockets of spending could pick-up in key bottlenecks or geographic regions without an overall recovery in place.
The largest determinant of enterprising networking revenue is IT spending trends. Despite a flattish IT spending environment, however, a few strong product cycles (i.e., WLAN and Gig-E) should help the Enterprise/SOHO networking IC sector grow slightly this year. Although Broadcom and Marvell are possible beneficiaries of these newer markets, other concerns limit enthusiasm for these stocks.
Boxmakers . . . Growth in the worldwide PC market is expected to slow in 2003, due to accelerating declines in public sector spending, according to market intelligence and advisory firm IDC. In the U.S., consumer spending on PCs remains strong and business spending is improving incrementally, but severe budget cuts and a poor fourth quarter performance is expected to constrain government and education spending on information technology, IDC said. In the fourth quarter, worldwide PC shipments rose 3.7 percent over the year-earlier period to 38.4 million units for a value of $46.9 million, which was within 1 percent of IDC's forecasts.
2020insight.com
Thanks for the good laugh Cary!
RtS |
