The crash of 2008: When it comes to semiconductor downturns, DRAMs are the canary in the coalmine. Major disruptions in the semiconductor market have often been most visible in the DRAM market. One of the worst crashes came between 2007 and 2008. What was different about this crash was that it came from the economy and was not a semiconductor supply-induced cycle.
No financial or industry analysts saw it coming before May of 2006, including VLSIresearch. In fact, little was foreseen of the coming storms by industry executives even in early 2007. Here are some of the general semiconductor forecasts in the public record from the SEMI Industry Strategy Symposium held early in January of 2007, where semiconductor executives gathered in Half Moon Bay, California:
- DRAM revenues up 16.1% in 2007 and +4.1% in 2008 — Klaus Rinnen, Managing Vice President for Dataquest’s Semiconductor Manufacturing Group, Gartner
- DRAM ASPs down 27.9% in 2007 to $4.27 per 512Mb and down 32.6% in 2008 to $2.88 per 512Mb — Klaus Rinnen, Managing Vice President for Dataquest’s Semiconductor Manufacturing Group, Gartner
- Semiconductor market to grow +9.2%. — Klaus Rinnen, Managing Vice President for Dataquest’s Semiconductor Manufacturing Group, Gartner
- IC Market to grow 7% in 2007, ASPs -3% to +2% — Bill McClean, President, IC Insight
- Semiconductor market to grow +7% Jim Feldhan, President, Semico
- Semiconductor market to grow +5% — G. Dan Hutcheson, VLSI Research
Here are some of the general semiconductor forecasts from the SEMI Industry Strategy Symposium Europe held early in February, where semiconductor executives gathered in Zurich:
- Semiconductor market to grow 6.5% in 2007 and 8.7% in 2008. — Jean-Philippe Dauvin, Chief Economist, STMicroelectronic
- Semiconductor market to grow 12% in 2007, with units up 8% — Malcom Penn, President and CEO, Future Horizon
- Repeated, IC Market to grow 7% in 2007, ASPs -3% to +2% — Bill McClean, President, IC Insights
Even early in 2008 the issues were not recognized by many. In early January, Dale Ford of iSuppli showed an estimate of a 3.95% revenue decline for 2007. Once all the data was in, the market was off 7%. Not being able to get the history right for 2007 in January of 2008 demonstrated just how difficult it is to pin down pricing in this market. His forecast of DRAM growth revenue for 2008 was a positive 0.2% — off on 2008’s downturn by well over 100,000% — and that was in January of 2008, well after May of 2006.
At the macroeconomic level, most of the world’s most respected economists, investment banks, and central banks did not foresee the collapse of housing prices and derivatives that would threaten the global financial system in 2007 and 2008. They missed the subprime mortgage disaster, the derivatives mess, the Bear Stearns bankruptcy, and the Lehman collapse. Hence, there is no reasonable expectation that industry analysts could have foreseen it and then factored all of these events into DRAM prices.
Prior to the 2007/2008 financial crisis there had not been two back-to-back years with DRAM price-per-bit declines of 50% or greater since the Asian Financial Crisis in 1997-1998, ten years earlier. Before that, at no time in the previous history of DRAMs had there ever been two back-to-back years with price-per-bit declines of 50% or greater. Moreover, excluding the Asian Financial Crisis, all downturns prior to 2006 had lasted only a year.
Financial crises had an extraordinary effect on the DRAM market in part because of the latter’s linkage to the general economy (which drives consumer and IT purchases of PCs) and in part due to the heavy capital burden born by DRAM makers to keep up with Moore’s Law. Access to capital dries up fast in a financial crisis and if a DRAM maker cannot afford tool replacements to stay on the leading edge, their costs will not fall with Moore’s Law. This problem is compounded by the fact that DRAM prices fall deeply in downturns.
DRAM makers expect prices to fall deeply in downturns. Healthy companies with good balance sheets can easily weather the normal single year downturns. This is especially true when there is easy access to capital to get them through the normal market fluctuations. But the unexpected financial crisis negates this, forcing producers out of the market.
The effects of financial events in 2007 and 2008 rippling downstream into the DRAM market can easily be seen. DRAM prices had been trading tightly in 2006 at a slightly negative 2.9% CQGR. Once 2007 got underway, subprime lenders started to file for bankruptcy. DRAM prices followed the news with a sharp decline. They stabilized briefly in the summer.
But DRAM prices started to fall again in August 2007, when the Interbank funds market froze. This event meant that things were so bad and trust so low that banks would not even loan money to each other. Credit was drying up. The electronics and DRAM industries do not operate in a vacuum. Executives are very attuned to macroeconomic events. This is especially true with DRAMs due to their price volatility.
Downstream, the Interbank funds crisis meant that business could be expected to cut investments in IT and consumers would hold off buying PCs. PC OEMs reacted by cutting DRAM orders and by pushing inventory into the market. DRAM makers reacted by unloading as much product as possible. Both OEMs and DRAM makers wanted to convert inventories into cash, since falling prices meant the value of DRAM prices would decline. It is just like a run on the stock market. So, both demand and supply shifts in DRAMs were amplifying each other to drive prices down.1 Between August and December of 2007, DRAM prices fell at a 32% CQGR.
2008 kicked off with another tightening of DRAM prices. From January to August, they only declined at a 4.0% quarterly rate — back on the 2001 to 2005 trend of tight trading. And this was with a return to normal bit production levels. Then, Lehman Brothers filed for bankruptcy on 15 September 2008. DRAM prices would fall at a 40.1% quarterly rate through December — more than ten times the 2001 to 2005 trend rate achieved through August of that year. It was 15 times 2006’s rate of decline.
It’s hard to overemphasize how deep a 40% compound quarterly decline is. This is equivalent to an 87% annual decline and 98% over a single Moore’s Law two-year cycle. At this rate, in just two years, price-per-bit would be 2% of its starting price.
Weekly spot prices lend more evidence, pinning the biggest drop of 2008’s DRAM price collapse to the exact week following the Lehman Brothers collapse. Lehman announced their bankruptcy on Monday, September 15, 2008, the same day we published spot market average medians for the prior week. Prices collapsed 21% in the week following the announcement. You have to go back to 8 October 2001 before you hit another week in which prices dropped greater than 20%.
1: It’s also why there is often significant bit growth in downturns when prices crash. Bit growth only measures what is sold by DRAM manufacturers. So there can be a surge in bits when PC demand is soft. Sometimes the bits wind up in inventory and other times it goes into PCs at deep discounts to hold prices up.
Author: G Dan Hutcheson