Joanne Itow's blog

On February 25th, 2012, Altera and Intel announced an agreement to build next-generation, high performance FPGAs on Intel’s 14nm Tri-Gate technology.   This is significant because Altera has used TSMC as their sole foundry for years.   Altera has never before been swayed by price or lured away by the promise of technology.  But, as the saying goes, “never say never”.  As of today Altera has two foundry relationships.

This announcement comes less than a week after Achronix announced the successful rollout of their Speedster 22i which is being built using Intel’s 22nm Tri-Gate technology.  Achronix used TSMC for their 65nm and 40nm product development.  It was somewhat of a risk, but Achronix left the TSMC technology track at 40nm, skipped 32/28nm and went straight to Intel for their 22nm Tri-Gate technology.   Was Altera watching in envy?

During TSMC’s Q4 2012 financial conference call, Morris Chang, TSMC’s CEO, was asked at what technology node and for which end markets does TSMC view Intel as becoming a more fierce competitor?   Morris Chang is not new to this business.   He stated that he already viewed Intel as a fierce competitor.

Do periods of economic downturn create more pessimists, or do they just get more attention during tough times?  Economists and politicians love to alarm us into believing economic growth is now on a slippery slope to permanently low levels.  Everything has already been invented, baby boomers are on the decline, and productivity will never experience the kind of improvements that were seen in the 1950s.  Robert J. Gordon, Professor of Social Science at Northwestern University wrote a paper for the National Bureau of Economic Research (NBER) in August 2012 in advance of his book release Beyond the Rainbow: The American Standard of Living Since the Civil War.  He points out that the computer and Internet revolution has resulted in a much smaller economic impact than that of the combustion engine.  He then argues that future innovations will have a diminishing effect on economic growth because of six headwinds¹.

SEMI recently released their silicon shipment forecast for 2012-2014.  Total wafer shipments are expected to reach record levels in 2013 and 2014.  Semico’s Wafer Demand model concurs with their forecast.  Wafer demand is expected to grow at a compound annual rate of 11.7% over the next five years.

The wafer demand pie keeps getting bigger but all the pieces are not growing at the same rate.  The pie looks a lot different than it did 15 years ago or even 10 years ago.   The chart below presents a few of the product categories that have traditionally utilized the most advanced technologies.

Figure:  Wafer Demand by Product as a Percent of the Total Wafer Demand

Source: Semico Wafer Demand Model Sept. 2012

Wafer demand for microprocessors has grown 65% over the past 10 years, but, as a percent of the total, MPU remains at less than 3% of the total industry wafers.  To a certain extent, this is due to the production efficiencies and improved designs that the MPU manufacturers have implemented.

There’s been a lot of attention focused on MEMS in the past couple of years and rightfully so. In 2011 when total semiconductor revenues grew by only 1.3%, MEMS revenues grew by over 34%. MEMS have been activating air bags in our cars and projecting images on DLP screens for years, but it wasn’t until the accelerometer in smartphones when mainstream semiconductor manufacturers decided they wanted a piece of the action.

No doubt, MEMS is still in its infancy. But will this market experience an inflection point that will trigger a meteoric rise in sales causing shortages and a market imbalance? And if so, when will that happen?

There are MEMS products available today which offer more efficient timing devices, displays, and microphones. MEMS are also opening the door to new applications with innovative developments in energy harvesting. Yet, many of these products are still playing second string to the traditional solutions currently available. What will it take for all these products to enter the market in volume? There are several variables that come into play, but the most important are cost and availability.

The ConFab 2012 highlighted sessions on the semiconductor industry’s blockbuster topics impacting advance technology manufacturing such as the transition to 450mm wafers and the increasing importance of 3D integration and advanced packaging. But this year the conference also allocated time to a discussion revolving around legacy manufacturing. Unlike finFETs and 450mm wafers, the challenges faced by mature production facilities are seldom in the headlines. However, as Sanjay Rajguru, Director at SEMATECH/ISMI, pointed out, over half of the current fab capacity comes from facilities that are more than 10 years old.

The challenges faced by older production facilities include equipment obsolescence, skills obsolescence, availability of parts, software and support and equipment capability extension and tool re-use. Maintaining “More than Moore” fabs is a major concern to semiconductor manufacturers as these operations reach 20 or even 30 years old.

At the ConFab 2012 Executive Roundtable representatives from Sematech/ISMI, IDMs, OEMs, equipment dealers, and industry consultants gathered to have an open discussion on concerns, roadblocks and possible solutions.

If you cover a long enough time period, the small ups and downs of a graphed line can look very smooth.  Semico’s semiconductor wafer demand data goes back to 1991.  When graphed from 1991 to 2016, wafer growth appears to have a very steady upward trend with only a few minor interruptions.  Wafer demand grows at a compound annual growth rate of about 8-9%.  When graphed against semiconductor units it looks very tame.  Looking at Figure 1 you may say to yourself, “The forecasting business can’t be that difficult.”  What’s so difficult about forecasting wafer demand?  Looking at the first graph, one might conclude that capital investment in fab capacity should be very successful if a company stuck with an 8-10% investment rate every year.

Figure 1: Semiconductor Unit Sales and Wafer Demand

Source: Semico Research Corp. & SIA

Reality sinks in when we look at the percent change in wafer demand since 1991.  (See Figure 2)  Wafer demand growth is anything but stable.  There is almost a 50-point spread between the peak growth year to the worst year of decline.  Even if we throw out the worst year (-23% decline in 2001) and the best year (25% growth in 1995) the change in wafer demand growth rates range from -2.7% up to 24.9%.

Figure 2: Percent Change in Wafer Demand

Source: Semico Research Corp. & SIA

Most semiconductor companies have completed their 4^th quarter 2011 earnings calls and announced their capital expenditure budgets for 2012.  Current capex plans indicate a weaker spending year for the industry compared to 2011.  Based on current announced plans, capex spending would fall 4.5% to $59.8 billion in 2012.  Semico believes some companies may be miscalculating the market. In 2011, semiconductor capex reached a record $62.7 billion, an increase of 25.6% over the 2010 level.  The growth was not nearly as high as the over 140% increase in capex that was spent in 2010. Top Ten Spenders in 2012:  Announced Capex Plans

As usual, CES was huge and overwhelming …too much to see, too little time.   Actually there may have been enough time, but my feet gave out before I could get to everything.  The Qualcomm booth wasn’t really a booth; it was a mini-city with 2-story briefing rooms.   From what I did see, the big trend in consumer electronics this year will be in touch, voice and motion sensing.   All the same products were on display, i.e. TV’s, tablets, and smartphones, they’re just doing a lot more than they did last year.

It all started for me at the Qualcomm morning keynote.   Banking on a commitment from Microsoft, Qualcomm is supporting Windows 8 along with ARM technology.  Their move into the x86 arena was showcased on stage with a demo tablet powered by Qualcomm’s Snapdragon S4 running Windows 8.  Furthering Qualcomm’s expansion into the home, Lenovo showed off their SmartTV powered by Snapdragon with voice activated remote control.  Can they truly get an instant on device?

Walking the floor there were many booths displaying motion sensing software.  Soon our online shopping experience will be improved as we virtually try on clothes using motion sensing devices that allow us to view the item’s fit as well as the color.

Last week I attended my first Semicon Japan.  This is typically one of the largest Semicon events, but considering the lull in equipment sales since this summer combined with the Japan earthquake in March, I was uncertain what to expect.  Seating at the opening keynote and welcome presentations were standing room only, but the foot traffic on the show floor was far from overwhelming.  (Even during the peak lunch hour, a SEMI aficionado directed me to some delicious food booths with essentially no lines.  I wish I could get those yummy noodles at Moscone during Semicon West.)

There was still a lot of talk relating to the earthquake impact and recovery but overall the mood was upbeat.  On the advanced technology front, vendors are preparing for more 450mm activity.  TSV and the move to FinFET transistors are providing plenty of opportunities and challenges for equipment and material suppliers.  Add to that a host of More Than Moore innovations for the mature fabs, and the topics for discussion were abundant.

Recently Semico released a report touting the above average growth rates in the analog market as well as the healthy revenue per wafer for analog products.  Smart money is moving into power management and other analog applications which have gained in popularity in our mobile electronic world.

Over the next five years, semiconductor units will grow at a CAGR of 9.6%.  The analog market will grow at a CAGR of 10.1%.  Specific analog product categories such as power management devices will grow even faster, logging in a 13.2% CAGR over the next five years.  Companies are gearing up for this growth by offering increased fab capacity, more efficient manufacturing and innovative materials and process advances.

Although a majority of the analog products are run on 200mm and 150mm wafers, there is still a fab running 75mm wafers and of course, there is the one 300mm fab now being operated by Texas Instruments.  The flexibility of analog manufacturing is also exemplified by the fact that many of these fabs are capable of running two different wafer sizes.