By Ed Sperling
Synopsys expanded its Discovery verification platform with a regression and analysis extension called CustomExplorer Ultra. The focus is analog and mixed signal, which is a hot topic these days with the momentum over 2.5D stacked die with interposers. Case in point: Texas Instrument’s purchase of National Semiconductor.

TSMC filed its annual 2010 report with the U.S. Securities and Exchange Commission. The report shows that 2010 sales hit a record $14.5 billion, up 68% over 2009. Net income was $5.68 billion, an increase of 84% year over year. The foundry noted in the filing that capital expenditures this year will total about $7.8 billion, an amount that may fluctuate depending on market conditions. It plans to add capacity to its 300mm and 200mm wafer fabs, develop process technologies at 28, 20 and 14nm, and invest in solar and LED production facilities. Also of note is that fabless companies accounted for 79% of 2010 sales compared with 80% in 2009.

Intel showed remarkable resilience in its earnings, as well. GAAP revenue for 2010 was $12.8 billion, up 25% over 2009. Net income was $3.2 billion, up 29% over the previous year. For the first quarter, Intel’s data center group showed the strongest growth, up 32%. Atom sales were up 4% to $370 million.

Those numbers coincide with strong sales growth at IBM, which got a boost from increased mainframe sales, and at Apple, which was bolstered by both iPhone sales over the Verizon network and new Mac sales. The iPhone uptick helped Qualcomm, as well.

By John Blyler

Is there any truth to the news that the semiconductor market is in the beginning of a cyclical bull market?

Taken in total, the available evidence seems inconclusive. On the plus side, the Philadelphia Stock Exchange (PHLX) Semiconductor Sector (SOX) index remains up more than 30% from its low in early March, as reported recently in The Wall Street Journal. The PHLX SOX is a price-weighted stock market index composed of 19 well-known semiconductor companies like Intel, Micron, National Semiconductor, Xilinx and others.

On the negative side, Intel’s outlook for the second quarter was less than impressive. Further, iSuppli has released several disparaging reports:

* Nearly 60% of China Chip Manufacturing Goes Unused in Q1 ’09

* Reports of a Memory Market Recovery are Greatly Exaggerated

Still, most of this information is backward looking based on existing chip inventory data. What is the future trend for the next six to nine months? While no one can be sure, perhaps the best indicator can be found by looking over the shoulder of today’s chip architects. What are the leading chip design companies investigating in terms of new chip projects – early trade-off analysis at the architectural level? Data from an aggregation of leading IDMS and EDA design houses, now over 70K unique chip projects since 2005, suggests that chip design investigations are on the rise (see Figure 1).

Figure 1: The number of architectural investigations for future chip projects picked up significantly in the first quarter of 2009.

Before discussing this data, one must agree to the overly documented fact that ASIC design starts – and hence investigations – continue to decline. This decline relates to the actual number of design projects, but NOT to die sizes or chip complexity, which continue to rise. Leveling out this slow downward trend in chip investigations and starts allows us to more clearly spot year-to-year patterns. We can “ignore” this overall downward trend in the same way engineers remove a DC bias from a signal. In other words, we can subtract it out.

Let’s return to the data trends represented in Figure 1. Both the actual number and rolling average of design investigations are on the rise. The latter is a trend predictor that shows the growth/decline tendency for the next month. The upward movement of both curves is good news. In fact, this recent growth over the first quarter of 2009 represents a 19% increase from the same quarter (Q1 of 2008) last year. Indeed, the first quarter of this year is starting to look much more like the first quarter of 2007 rather than 2008 (see Figure 2).

Figure 2: A quarterly look at architectural-level chip investigations suggests that Q1 ’09 may be more like Q1 ’07 than Q1 ’08.

This is encouraging news since it suggests that chip vendors are looking beyond the current economic crisis to six-to-nine month timeframe when demand for their products will increase. But which product segments are reflected in this growth of chip architectural investigations? Though not shown, the three most sited electronic market segments corresponding to the total design investigation data in Figure 1 include communications, consumer electronics and data processing (server) markets.

Knowing which market segments are likely to experience growth in the next six to nine months is useful, but it doesn’t answer the bigger question of how profitable a potential semiconductor bull rally might be. The answer to that puzzle lies in understanding the profit margins associated with each market segment. If current trends in electronics are any indication, then growth will occur in the embedded but not the PC space. This could mean high product volume but lower profit margin for manufacturers, because embedded chips sell below their PC counterparts. Still, a smaller profit is better than no profit.

Still unanswered is whether this increase in chip design investigations is due to a seasonal rise or whether it indicates the start of a cyclical semiconductor bull market. Only time will tell.

If you think things are bad, be glad you’re not in the Taiwanese foundry business—where the pain level is strangely uniform.

TSMC’s sales dropped 59.5% in February compared to the same month last year, and 7.5% compared to January. How many ways can you spell ouch? 

UMC’s numbers are down 56.9 percent in February 2009 vs. the same period in 2008. That’s pretty close. In fact, it’s remarkably close.

This kind of information is only available in Taiwan. SMIC, based in Shanghai, and Chartered, based in Singapore, don’t report monthly sales numbers.

Nevertheless, there was at least some encouraging news out of Chartered. It said that sales seem to be stabilizing and wafer starts appear to be increasing for Q2. 

There is evidence of this showing up in other parts of the market. U.S. retail sales, excluding big-ticket items like cars, show modest increases in areas like clothes and consumer electronics. Numbers were up in January and February. It certainly wasn’t a robust gain, but it wasn’t negative, either. That will translate into new design starts sometime in the next few months, which barring any more major drops will start this whole cycle rolling again.

Design activity has to begin at least six months prior to any turnaround, which means that if the overall economy is expected to show growth in 2010,  electronic designs have to begin by mid-year—perhaps even sooner.

None of this is perfect, however. Why, for example, did National Semiconductor just announce plans to cut 26% of its workforce? At least part of that can be explained by closing of an assembly and test plant in China and a fab in Texas. Too much capacity is expensive, and we wouldn’t be surprised if National ultimately begins outsourcing some of its work to foundries. Yes, it’s analog, but is it still more efficient to run fabs yourself, even if they’re fully depreciated, when TSMC and UMC are begging for business?

Meanwhile, in the FPGA realm, chip design is getting so complex that EDA vendors are finally beginning to find inroads. This is a market previously owned by tools from the FPGA vendors, which they readily gave away to customers at little or even no cost. That worked fine before the industry got to 90nm, and at 45nm it’s tough enough even with the best of tools.

Mentor introduced its Precision Synthesis Tool family for Altera’s Stratix and Arria families. Our guess is that you can expect to see a lot of activity in this market in the near future, and not just from Mentor. Synopsys’ purchase of Synplicity gives it a vested interest in the FPGA market, as well.

–Ed Sperling

Independent investor Jim Hogan talks about where the real value is and what companies need to do to survive in a changing market.

By Ed Sperling

Mountain View, Calif –Dec. 17, 2008—Solar cells may be one of the ultimate system-level design challenges.

While it’s well understood how to make solar cells far more efficient—gallium arsenide, for example would make solar cells 35% to 40% efficient compared with CMOS, which is about 20% to 22% efficient—the problem has to be overlaid on a cost vs. efficiency equation. Gallium arsenide would increase the cost of solar cell chips by about 10 times, rendering it completely useless to commercial adoption of the technology, according to T.J. Rodgers, CEO of Cypress Semiconductor and chairman of the board at SunPower.

The dark horse material is cadmium telluride, which is currently being researched as a possible replacement for CMOS. Unlike CMOS, which has an indirect bandgap, cadmium telluride has the same kind of direct bandgap as gallium arsenide, which makes it more efficient.

But the even bigger problem lies outside the packages of solar cell arrays. Rodgers says it costs $750 per unit for the solar cells, and about the same amount of money to install it on a roof. He said the technology wrapped around the solar cell needs to be dramatically simplified to bring down the cost.

“The installation is too high and electronics are too expensive,” he said. “All of those things will be improved. By 2012, the goal is to match PG&E’s cost for producing electricity. In a couple of years, the price of solar will be cheaper and you will start seeing exponential growth.”

National Semiconductor claims to have solved one piece of the puzzle. Brian Halla, chairman and CEO of National, showed off a battery management device this week that he claims will greatly improve the efficiency of existing solar installations.

“A string of solar panels today works like the batteries in a Maglite flashlight,” he said. “If one goes out, they’re all bad.”

He said the same applies to solar panels, where one may be partially shaded for part of the day and throws off the electricity generated by the others. By routing the power through the device, he said far more power is generated.

Both companies, as well as Actel, are experimenting with a number of new system-level designs that can either generate or conserve electricity. Among the technologies being tested:

1. Heat-generated electricity. A device being researched will wrap around the smokestack on diesel trucks is expected to produce 2Kw/hr. that currently is wasted. The same device can be applied to a hot exhaust pipe on a car to charge a battery. While it won’t fully run a hybrid car, it may increase the mileage by 10 percent.

2. Commercial fuel cells. These are expected to show up in commercial generating stations over the next several years.

3. Non-fixed solar farms. These are expected to begin operation over the next several years so they track the sun’s arc instead of fixed cells that can only produce electricity during part of the daylight.

John East, president and CEO of Actel, noted that the total savings from eliminating imported oil and avoiding wars fought to preserve oil could amount to about $1.5 billion a year, making it likely that such projects will become increasingly attractive sources of research and development over the next four years.