by Bob Haavind, Editorial Director, Solid State Technology
The global economy will remain in the doldrums for at least a year, well into 2009 at least, analysts and economists agreed at the Industry Strategy Symposium in Half Moon Bay, CA. But views diverge on where it goes from there, ranging from a big boom starting in 2010 to a lengthy period of stagflation, with interest rates possibly hitting 16%.
The most upbeat of all was Bill McClean, president, IC Insights. “Every recession in the last 30 years has been followed by a boom in semiconductors, every time,” McClean exclaimed, showing how strong revivals for chips had followed five downturns over the past three decades.
While many speakers urged caution about the next few years because this recession involves an unprecedented breakdown in world financial markets, McClean ticked off a litany of factors suggesting a strong rebound ahead in spite of the current malaise. First, he cited the huge and growing fiscal stimulus: $850B in the US, $586B in China, $267B in the Eurozone plus $458B in the UK, $111B in Japan, and $8B in India. China is giving subsidies for people to buy cell phones and plasma TVs. “There’s even a stimulus package in Jamaica,” he added.
Aside from this, there are record low interest rates; oil prices are low; and pent-up demand is growing for automobiles and electronic products, he explained. He agreed that it is still tough to get loans, but data show that credit markets are unfreezing.
As a result, he sees at least double-digit growth for semiconductors in 2010, followed by a boom in 2011. But until then, he sees semiconductor revenues down 17% in 2009, with the second half of the year being 50% stronger than the first half, and a big fourth quarter for electronic products. Unfortunately, McClean said, while downturns tend to start in the first quarter, this one kicked in during the fourth quarter of 2008, the worst time of the year for the electronics industry.
The big issue in this downturn is pricing, not capacity, McClean said. He expects a record amount of production capacity to be taken offline this year (at least 3%), which will help raise ASPs. Capital spending will only be 15% of revenues, down some 30%, he believes, well below the trend line. The last time there was low capex in DRAM (in ’04 and ’05) it jumped 32% in the following year (2006). As a result of the underspending, McClean expects a big jump, at least a 20% increase, in capital spending by chipmakers in 2011. Foundries have been holding back spending on equipment to help boost ASPs, and as a result he expects revenue/wafer to increase after a steady decline over past quarters.
Worldwide IC wafer capacity changes (200mm equivalents)
Fixing the “broken” memory market
Several speakers, including McClean, see the memory market, which has been driving equipment sales over the past couple of years, as broken and in need of major restructuring.
“Natural selection doesn’t work in DRAMs. Hynix should be gone. ProMOS and Powerchip should be gone. Sixty cents for a 1Gb DRAM — Samsung is selling them for the cost of the crazy package!” McClean said. He noted, though, before chiding South Korea or Taiwan, that the US government is backing GM and Chrysler.
“Memory is not a business, it’s a hobby for Asian politicians,” Dan Hutcheson, CEO, VLSI Research commented. Too many producers with too much capacity, he said, is leading to “an ASP-driven sales decline and profitability collapse, with production below variable cost.”
The industry must rationalize the number of suppliers. In DRAM, the top four gained market share in 2008; in NAND flash, he said, none of the suppliers can make money. And after the top five there are none doing over $1B in business — yet it costs about $3.5B to build a memory wafer-processing plant, and it takes that much in revenues to pay off 20% a year for five years.
“Even Samsung doesn’t have a large enough market!” Hutcheson said.
Nevertheless, he cited a recent briefing given to VLSI by memory market expert Jim Handy with some hope for the future. Nine competitors had to split up a $25B market in 2008, he said, and the result will be that prices will have to rise in ’09. Capex is expected to decline sharply this year, and won’t get back to 2007 levels until 2012 or later.
Dan offered a prescription for the memory makers based on a quote from Don Valentine, a veteran venture capitalist in the industry: “Whenever you find yourself in a deep hole, quit digging.”
Randy Bane, VP/chief economist, Applied Materials, agreed that the memory sector needs restructuring. He sees cuts in 200mm capacity of 45%-50% by 2010 from 2007 levels. He believes that the whole industry needs restructuring. Semiconductor profitability erosion and capital destruction will lead to broad consolidation, he said, pointing out several joint manufacturing ventures in DRAM and flash.
“Memory makers have consolidated manufacturing at larger-scale sites,” he explained, adding that smaller sites also will have to consolidate.
Still, he did cite some hope of a revival ahead for memory makers. In DRAM, DDR3 (double-date rate) is being ramped and will far exceed DDR2 by the end of 2009, he said. DDR3 is far superior, he believes, with lower-cost, power, and voltage, and is faster as well. It will go far beyond server farms, becoming ubiquitous, and accelerating new applications.
“It will make memory profitable again,” he believes.
SanDisk sees emerging markets: SSD, slots
Sanjay Mehrotra, SanDisk president/COO, presented the view from inside the memory implosion. In 2007, he said, revenues peaked at $16B while there was $12B of capacity expansion. Capex rose from 40% to greater than 70% of revenues in 2007.
SanDisk has a 50-50 venture with Toshiba for two large 300mm fabs which are being gradually filled with production cut to 70% of capacity until 2010-2011, he explained. The focus will be on new technology and cost reduction. In 2009, there will be a 32nm production ramp, and a shift to 3 and 4 bits/cell implementation. Multi-die stacking also helps lower production costs.
Right now, he said, there are about 700M cell phones with flash slots, but this will rise to 1B by then as users increase music and video downloads.
In addition, he cited Gartner/Dataquest research indicating that by 2011, 1 in 5 computing devices will have solid-state disks (SSDs) instead of hard drives. Controller/system expertise is the key to this market, Mehrotra said, and multi-level NAND is required. There are important emerging markets in netbooks as well as notebook computers, with applications involving digital imaging. He sees a 117% CAGR over the next five years for SSDs.
SSD memory is 5× as fast as hard disk, boots twice as fast, can run applications as much as 10× faster, and is 400% more reliable, according to Mehrotra.
“Blockbuster alone moves 3 exabytes of movies annually,” he said, “and now they’re going to HD.”
Slot media will be widely available in stores like Walmart and Best Buy, and slot music cards will be capable of holding full albums of music. Slot radios with 1000 pre-loaded songs (selected by Billboard) will be available for <$100, with $40 for an extra slot that could hold another 1000 songs. This media can also be put into MP3 players, he noted.
The flash industry has been pushing ahead of Moore’s Law on circuit density, but he also cited the work on putting multiple bits per cell. SanDisk is moving from 2 bits/cell now to 3 bits/cell, and will go to 4 bits/cell in the future, providing 16-level advanced controller technology to write and manage data.
He agreed the technology is getting much tougher due to physics challenges with steadily fewer electrons per cell, and that’s why there is intensive research on phase-change nanomaterials and 3D. The ultimate 3D, he believes, will be the stackable cross-point diode array, with peripheral circuits at the bottom of the array, but this is still a few years off.
Weathering the downturn: Technology, not capacity
Bob Johnson, research VP, Gartner, laid out the conditions for financial viability for the semiconductor industry:
– Sufficient cash generation to fund needed investments,
– Funds to support R&D to stay competitive,
– Ability to weather major industry cycles, and
– Access to needed capital.
As for weathering the current downturn, Johnson feels the fabless companies, foundries, and Intel are in good shape — but all the rest are in trouble.
He tempered this by pointing out that although profits are tenuous for 2009, there will be high cash flow due to depreciation on tools bought in 2005-2007. He also expects a big jump in capex in 2010 and 2011 due to underspending the next couple of years. In recent times, Johnson pointed out, memory makers have provided 40%-60% of total capex, but it has been undisciplined investment. Memory has always gone through boom-bust cycles, but he sees it as different this time. Memory makers are now a group of stand-alone companies with revenue growth anemic at best, going into a recession in a condition of weakness. In the past, parent companies could prop them up, but now they have been spun off. With NAND weak and DRAM flatlining, Johnson sees the potential for running out of cash. He predicted that after a $150M infusion from Saxony, Qimonda had at most two more months to go without help from Infineon. (Only a week after ISS, Qimonda filed for bankruptcy to enable it to restructure.)
Although there may be some consolidation, Johnson believes there may be more cooperation rather than outright M&A activity.
A slow turnaround for the semiconductor industry may begin in 3Q09, suggested Klaus Rinnen, managing VP, Gartner, but he doesn’t see a sustainable recovery until 3Q and 4Q in 2010.
“There are no good signs for 2009 in any end-market sector,” Rinnen said, including data processing, communications (both wired and wireless), consumer electronics, and automotive. He expects all to hit bottoms in 3Q and 4Q of this year. The US led the decline, and he believes the US must lead the recovery.
Equipment makers will see some early turnaround in 4Q09, Rinnen believes — but again, he says, a sustainable recovery won’t begin until 3Q and 4Q of 2010. By then a new cycle of memory investment should start and foundry capacity utilization will rise again, triggering equipment buys. By 2011, he expects all sectors to be up again.
Rinnen showed a chart mapping the decline of revenue/sq. in. of silicon, peaking at $42 in 1995 and sliding down to $26 by 2012, about equal to 1986. He suggested that chipmakers may choose to limit supply to force up ASPs again, and he expects less buying of equipment for leading-edge chips.
Revenue per in2 of silicon — pressure on margin impacts below line costs.
The target markets have slowed from 17% growth in the ’90s to perhaps 8% now, Rinnen believes.
“How much capital intensity do we need?” Rinnen asked. With 17% growth, chipmakers were spending $0.22-$0.24 for every $1, so now at the 8% rate there has been overspending, he said. Memory spending of $0.45/$1 is unsustainable, he commented, and he pegged logic at about $0.17/$1 and foundry at $0.18/$1. He expects this capital intensity to decline going forward. As a result, R&D spending by equipment makers will drop $8B-$12B over the next four years — just when more R&D will be needed.
Rinnen had advice for both chipmakers and tool vendors. He believes chipmakers should focus on technology rather than just capacity. They must evaluate if Moore’s Law still makes sense, but he pointed out that technology leaders are also profit leaders. The industry must increase the rate of consolidation, he said, and it must seek radical improvement in manufacturing efficiency.
Capital equipment suppliers must prepare for survival in a slower-growth era, in which they will have fewer customers but with deeper pockets. Vendors should diversify into allied markets, and maximize R&D efficiency through alliances and collaboration. He suggested they consider licensing technology in some cases rather than creating it.
Madoff vs. Bernanke?
The most downbeat outlook of ISS ’09 was presented by David Townes, managing director/cofounder of Needham & Co. He emphasized that the views were his own and not of the company. Everyone knew bad news was coming when Townes opened with a WC Fields quote: “Start each day with a smile, and get it over with.”
Today’s stimulus will be tomorrow’s problem, he believes, as the government struggles to cushion a depression (he purposefully did not call this a “recession”). He believes in the next few years we will see hyperinflation in the US >10% annually and maybe reaching 16%; and he believes the US dollar may decline precipitously, and may be replaced as the global standard by a basket of alternative currencies.
Invoking Thomas Jefferson, Townes said that banking institutions are more dangerous to our well-being than standing armies. Since 1972, when the US closed the gold window, Townes believes the dollar has lost 98% of its value. “Even T-bills are highly speculative,” he suggested, and he believes that equities need to decline another 50%. He called the fiscal and monetary “debauchery” as bad as anything in the 1930s. Furthermore, US deficits of some 14% of GDP are now being echoed globally.
He suggested that while Madoff lost $50B, Bernanke may lose $10T!
The Dow Jones should bottom out about 4000, Townes suggested.
Is there any hope? In spite of the torrent of bad economic news, Townes said that just prior to ISS ’09, 325 companies had shown up at the Needham Growth Conference. It’s hard to keep a good industry down. — B.H.