Search This Blog

Monday, August 30, 2010

This Rebound is Different

As the semiconductor capital equipment and materials market enjoys nearly unprecedented year-over-year growth, many industry analysts see signs of caution on the horizon. The most recent SEMI Book-to-Bill ratio reached 1.23 (meaning that $123 worth of orders was received for every $100 of product billed for the month) and the three-month average of worldwide bookings jumped 220.4 percent above 2009, but a packed audience of industry leaders heard mixed signals related to 2011-2012 growth at the recent Silicon Valley Lunch Forum.

Bob Johnson of Gartner sees capital spending up over 90% in 2010 and continuing to increase in 2011. He sees, “strength through 2012, then retrenchment.” After reaching $35.4 billion in 2010, Gartner’s current cap spending forecast reaches $37.7 billion in 2011 (6.6% growth) and $40.8 billion in 2012 (8.2% growth). Gartner’s current capex estimate for the 2013 “retrenchment” is $34.9 billion, a decline of 14.6% from 2012 estimates.

Gartner’s current data indicates market strength: IC unit volumes and equipment are surging, driven by strong end user markets, especially PCs and mobile phones. Johnson does acknowledge, however, uncertainty caused by recent company announcements. While global economic recovery is strong, US economic growth is apparently softening and uncertain. He also sees volume declines at the packaging, assembly and test houses, and some indication of inventory increases in the supply chain.

Bill McClean of IC Insights is quick to point out that the current weak US economic recovery is not representative of the overall global market environment. Modest GDP growth in the US, Europe and Japan projected for 2010 should not obscure the robust 7.3% growth elsewhere in the world. China’s GDP is projected to grow 10.6% and India 8.5%, and much of this rising prosperity is driving strong global IC demand.

But McClean does acknowledge, “This rebound is different.” Unlike any other year since 1983, after two down years in capital spending, this year’s capital spending surge will be in the first year of recovery, not the second. Capital spending will increase by a reasonably robust 9% in 2011, but far less than the 93% he projects for 2010. Every year in the modern history of semiconductors the largest increase in capital spending has occurred in year two of the recovery. Does this cap spending trend signal a more modest cyclical decline in 2013, or is the industry entering an era with new spending patterns? Typically capital spending will decline after a 50% increase in capex, but this year’s IC wafer capacity will only increase an estimated 1.6% after an unprecedented 7.6% decline last year (capacity has only declined one year in history). And, 300mm IC fab utilization will reach 98.4%, suggesting the market can easily accommodate the recent increase in capital spending.

Both analysts agree that future cycles will be primarily driven by memory manufacturers. Johnson estimates 41% of 2010 capex will come from memory fabs, versus 21% from foundries/OSATs and 17% from other IDMs. McClean ranks six of the top ten capital spenders in the memory segment, representing 55% of the total spending from these top companies. Memory firms have been most susceptible to the market forces that accelerate super-cyclical swings in capital spending. In addition to the traditional cycles influenced heavily by rising then falling ASPs and fab utilization rates, Johnson suggests that capital spending by memory makers will also not have sufficient net cash flow to fund future capex needs.

The cyclicality—or super-cyclicality—of spending patterns is one of the many challenging and unique attributes of the semiconductor industry. In one of the most demanding R&D industries in the world, the IC supply chain has had to endure radical swings in capital spending while keeping pace with relentless Moore’s Law and stockholder expectations. The rapid rebound from last year’s alarming drop in equipment spending (down to less than $17 billion from a 2007 high of over $44 billion) is a testament to the ingenuity, flexibility and resourcefulness of SEMI members. Whether this PhD-intensive industry can sustain another swift drop in spending remains to be seen, but for the next few quarters, a positive market environment looks certain. Beyond that, well, only time will tell.

Thursday, August 12, 2010

Manufacturing and R&D Go Hand in Hand

If you’re making an electronic product with a motherboard than needs stuffing and a final assembly, it doesn’t matter much where you make the product from the company’s perspective. Manufacturing certainly matters when you are talking about national interest and public policy. Here, governments have a strong interest in where large employers put factories, but private interests are just concerned with economic optimization.

For companies, you’re looking for the lowest total cost at guaranteed quality levels and there are plenty of qualified companies around the world capable of performing the function. The long term competitive advantage for most companies is not in manufacturing, but in innovation, design, marketing and other points on the value chain. Manufacturing is a function that can be outsourced, like payroll accounting, communications, insurance and the like.

For most products, there is very little interaction between the factory and product development process so the value of manufacturing location is slight. The calculation is concerned exclusively with the Bill of Materials, assembly costs, shipping costs, quality metrics, and other pure economic considerations.

ICs and solar PV are different. It does matter. R&D and manufacturing are intimately linked and losing one will probably accelerate the loss of the other.

Mike Splinter made this point at the Aspen Ideas Festival when he explained that in our business, “manufacturing isn't just where ideas are put into practice, but a key part of the innovation ecosystem.”

In ICs, how fast you get to the next node, how fast you debug and tape out, how fast you ramp, and how fast to yield will determine your success. Each of these steps requires manufacturing expertise of the highest order. Sure, a lot of chips are made in foundries, but manufacturing remains an enormous chunk of the value chain. There’s not too many fabless companies doing better than TSMC or ASE.

And its not just scaling that is sustaining the value of manufacturing in semiconductors. Since 1975 scaling has delivered around 8000 times improvement in the cost of ICs as transistor feature sizes were reduced from 4 micron to 45 nanometers. Manufacturing efficiencies accounted for another 500 X in cost reduction over this period. Some experts have claimed that technology shrink delivers 30-50% of the yearly cost reduction in chips, productivity improvement brings 10-30%, yield 10-15% yearly (wafer size transition brings another 30-50% cost reduction every 10-15 years.

In this environment, manufacturing capability in inextricably linked to profitability and competitive advantage. You don’t do productivity, yield and cycle time well, you don’t survive. And, if you don’t have R&D deeply engaged in manufacturing, you won’t be competitive.

Same is true for solar. Thin and crystalline solar technologies are similar process technologies to chips. Manufacturing optimization is critical to sustaining long term competitive value. Just a glance at NREL’s tracking of best efficiencies in solar sees a long, steady continuous improvement process. While the NREL data points are from the lab, the process is replicated in the fab; the ability to commercialize incremental efficiency improvements through manufacturing is how, in large part, solar manufacturers’ compete. Having a fab onsite, where R&D engineers can prototype new recipes, troubleshoot new processes, and tweak yield and productivity improvements is an advantage and provides real value.
Large, multi-billion dollar fabs also bring a clustering of best-in-class suppliers to the neighborhood. Specialty gas producers, machine shops, labs, scientific specialties, process technology firms, etcetera, ecetera always spring up around fabs. They innovate too.

So, when Apple builds the iPad at a Foxconn plant in Shenzhen, or HP netbooks in Taipei, is the US at risk of further losing their share of the mobile computing market? Will those companies shift a large portion of their R&D and marketing teams to Asia? Probably not.

But when fabs go up in Asia, PV and IC R&D is probably not too far behind.

Mike Splinter is absolutely right when he says the micro and nano manufacturing is an essential part of the “innovation ecosystem.” The question is whether any of US’s national leaders that flock to the prestigious Aspen Festival really understood him. I know leaders in China, Korea, Taiwan and Japan do.