Disconnects in the Semiconductor Industry

There seems to be a misconception between semiconductor revenues and other business sectors of the industry, namely equipment, chemicals, and photomasks. In reality, all four components are loosely related at best.

Semiconductor Revenues – Revenues are a comparative measure of performance that will determine market share. It's not the revenues that count, it's a company's profits that determine stock performance. A company's capital expenditures (capex) to purchase equipment and build new fabs are loosely based on revenues, but are more closely tied to capacity and technology nodal points. A company will need to spend if they can't meet demand because they are capacity limited. Likewise, companies will spend to move to a tighter technology note (i.e., to 45 nanometers from 60 nanometers) to maintain leading edge product introductions and market share.

Semiconductor Equipment – Equipment sales are based on capex, capacity, and technology as discussed above. However, capacity is based on unit shipments, not revenues. In fact, the semiconductor industry has experienced six consecutive years of double digit unit shipments, averaging 10% per year. Since 1980, unit shipments of semiconductors only decreased once (in 2001), semiconductor revenues increased each year since 2001, and yet semiconductor equipment sales increased in only 2003, 2004, 2006, and 2007.

Chemicals and Materials – Revenues are based primarily on silicon real estate, price increases, and technology nodes. The more semiconductor wafers processed, the more chemicals used. So this sector is tied to unit shipments and it is not surprising that revenues grew from 2000 through 2007 each year except for 2001.

Photomasks – Revenues are based primarily on semiconductor design activity rather than sales volume and on technology nodes. There is no correlation with semiconductor revenues. There is some correlation with semiconductor unit shipments, but that is based on how unit shipments increase. For example, if more capacity was added to a production line, there would be minimal impact on revenues because the same mask set could be used whether there were 1,000 wafers processed or 10,000 wafers processed per month. However, if unit shipments were increased because of a die shrink (more chips on a die from smaller linewidths), new masks would be needed. Increased complexity plays another major role. A 45 nanometer mask set will cost much more than a 90 nanometer mask set. If capacity constraints necessitate the need for a new fab or production line, mask set purchases will increase because masks are fragile (made of quartz) and are not moved around the fab.

The discussion above is a bit simplified and generalized, but serves to illustrate these relationships. I spoke above about semiconductors as a whole, but the market according to the World Semiconductor Trade Statistics [WSTS] is comprised of Discrete Semiconductors, Optoelectronics, Sensors, and Integrated Circuits that include Analog, Micro, Logic, and Memory. So, if a supply-chain company has customer that makes NAND flash memory devices, which has tanked in 2008, then its revenues will be impacted even though the overall semiconductor market is slated to increase 4-5% in 2007.

I raise these issues because recently Gartner lowered its revenue forecast for the semiconductor industry from 4.6 percent to 4.2 percent over 2007 and from 7.9 percent to 7.8 percent over 2008. To me the differences appear in the "noise" range, and it amazes me that a firm can make an inexact science of forecasting (I've been doing it for 25 years) sound like an exact science.

Nevertheless, this almost monthly revision of its forecast raised eyebrows, impacting several supply chain companies, when in fact it should have only affect semiconductor companies and, as described above, only to a minor degree as P/E will be driven by earnings not revenues.

One supply chain supplier downgraded after Gartner's forecast was Photronics (PLAB). Again, photomask revenues are tied to semiconductor design activity rather than sales volume, so there is no correlation. Companies in the design space include Cadence (CDNS), Synopsys (SNPS) , Mentor Graphics (MENT), and Magma Design Automation (LAVA). In addition, there are hundreds of design houses in China, India, and Taiwan. Their designs will be built by foundries, each requiring a photomask.

Our analysis shows that for 2007, Photronics held a 16.8 percent share of the $2 billion global market for semiconductor photomasks, behind DNP and Toppan. In the past three years, Photronics continues to make substantial capital expenditures to meet advanced technological demands, which as mentioned above, will generate greater revenues than older technology masks.

There have been rumors that Samsung Electronics Co. is pursuing an acquisition of U.S. computer memory card maker SanDisk Corp. (SNDK). Samsung currently supplies NAND flash memory chips to SanDisk, which is based in Milpitas, Ca. A takeover would help Samsung widen its lead over Toshiba Corp. (TOSBF.PK) in the $15 billion market for NAND chips.

Samsung is Photronics's largest customer, with 25% of its sales in 2007. Samsung purchases the photomasks through PKL in Korea, which Photronics acquired in 2001. If the acquisition of SanDisk comes about, it will necessitate the need for additional photomask purchases by Samsung. In addition, Samsung has a habit of building a new fab every year, the envy of the IC world, which also required new photomasks from PKL. In fact, Samsung's capex are estimated to be over 40 per cent of sales revenue in 2008.

Disclosure: Long PLAB.

Comments

[BLOREAN:]

Since 1996, I am quite disappointed by regular wrong market predictions from these research firms. The semiconductor industry relationship is well explained here. What matters is how fast the technology is getting commoditized. The unit shipment might have grown enormously but not the revenues. The change current is fast and it's very invisible in semi market.

2008 Sep 11 01:42 PM