by James Montgomery, News Editor
As the industry watches memory firms push out orders as prices and profits evaporate, those waiting for foundry spending to mount a late-year comeback to pick up the slack might want to avert their eyes from the scoreboard. Warning signs in the memory sector — and the dawning realization of how much of the industry’s capex is tied up in the most price-sensitive segment — were major topics discussed at a recent SEMI breakfast panel of analysts near Boston.
Comparisons to the dark days a decade ago echoed in the conference room. Bob Johnson, research VP at Gartner Dataquest, showed slides illustrating that memory spending peaks in both 2007 and 2008 will account for more than half of all wafer fab equipment spending, something that hasn’t happened since 1995-1996, and that was followed by a couple of the worst years in the history of the semiconductor industry. Having NAND flash today as an offsetting factor to DRAM helps, he noted, but general conditions “are surprisingly similar.” Memory capital intensity has been >40% for three years in a row now and is also approaching record levels [see chart, above] comparable to both 1996 and 2001, he said.
“We’re back in the 1990s again,” agreed Bill McClean of IC Insights, pointing to multiple examples of wild capex growth from the likes of Powerchip, Nanya, Elpida, Qimonda, and Samsung and Hynix. Even Micron, which he noted has “been through all the price wars” and should know better, has been swept up in the “irrational exuberance,” he claimed. McClean also chastised what he called the “voodoo financing” behind Sandisk’s multibillion-dollar partnership plans, pointing out that the company’s SEC filings to report fab financing are more than 400 pages long.
Although DRAM and NAND flash producers are using silicon more efficiently, their revenue productivity is declining [see charts, below], Johnson pointed out, and that has a direct correlation to profitability pressure. Firms are building and investing in capacity too much, he warned, and ultimately that will put pressure on pricing and profits that is best withstood by larger firms with deep pockets. “If I were a second tier supplier, I’d be worried,” he said.
Memory firms account for half of capital spending (seven of the industry’s top 10 spenders are memory firms) but represent just 20%-30% of industry revenues, Johnson noted, and with only a portion of the rest of capital spending coming from foundries, they “can’t save the industry if memory capex tanks.” In fact, Gartner’s projections don’t show much of a foundry spending rebound at all this year — three of the four quarters in 2007 show 11%-20% declines in total foundry spending, somewhat concentrated in 2Q-3Q, and just 4.1% growth in spending for the entire year. Foundry capex is expected to gain momentum in 2008, but right now the overall picture for 2007 is unclear, he said.
Capital intensity is the reason foundries have emerged in the industry, and why they won’t ride in like “a knight on a white horse” to save this year’s equipment segment, Johnson noted. Foundry as a % of total capex has gradually declined from a peak of 24% in 2002 to 13.5% today, and with better capital investment efficiency it takes less spending to keep up growth rates and smaller investments to ramp up the business again, he explained.
Part of foundries’ issues are that their inventories have been building for seven quarters prior to 1Q07, while fabless firms stayed roughly flat — “it looks like the fabless buys are pushing back,” noted Johnson. Combined, foundry/fabless inventories are between the peaks of 2000 and 2004, but inventory days were significantly reduced in 1Q07 and should keep melting in 2Q. The speed of the decline will influence the timing of a revival in capital spending from the foundries, he said. — J.M.