, a flash memory joint venture between AMD and Fujitsu that is primarily controlled by the chipmaker, aims to raise $600 million in a public offering, according to papers filed Wednesday with the Securities and Exchange Commission.
The most recent quarters, however, have been tough times. Spansion has reported two consecutive quarterly operating losses, has seen its market sharesignificantly and is starting to experience pressure from Samsung and other companies as the different wings of the flash market--NAND and NOR--blur.
"They have got to make it more appealing than they have so far," said Kevin Krewell, editor-in-chief of the Microprocessor Report.
Spansion has also had to delay products. If the unit had released a 256-megabit flash chip in the fourth quarter, as planned, it could have added $60 million in revenue to AMD's bottom line, which would have allowed the company to report a profit for the quarter, said Jim Handy, an analyst at Semico. Instead, AMD reported a $30 million loss last quarter, and the chip didn't come out until this quarter, Handy said.
Spansion's difficulties lie in part in the hard realities of the flash memory market. Unlike processors, which can command relatively high prices, memory chips are subject to intense price competition. Companies eke out profits by turning out large volumes rapidly.
"The way you play in the memory market is having a bigger factory than the other guy," Handy said.
, Spansion was one of the highlights for AMD. In the second quarter of 2004, revenue for the memory division came to $673 million, while the operating profit for the unit tripled from $14 million to $45 million during a three-month period. By contrast, processor revenue came to $554 million, while operating profit declined $9 million from the previous quarter.
Spansion had also passed longtime rival Intel in market share and was making inroads into the cellular market, a traditional Intel stronghold. In AMD's announcement of second-quarter earnings, flash news was mentioned ahead of processor news.
Intel, however, was already recovering from earlier missteps at the time. It had lowered its prices and was rapidly recovering market share. By the third quarter of 2004, Spansion's operating income had dropped to $15 million. Then, in the fourth quarter, it reported a $39 million loss. In the just-reported first quarter, its operating loss expanded to $110 million.
Intel has also been trying to raise doubts in the market about Spansion's MirrorBit memory. MirrorBit holds two bits of data per cell, thereby letting a single chip act like two. Intel has a technology with a similar function, called StrataFlash.
The architectures of the two chips, however, are vastly different. StrataFlash is in reality a single cell that acts like two. In MirrorBit chips, there are two physically separate cells.claimed in February that AMD has been having trouble shrinking the size of MirrorBit chips, thereby reducing the volumes it can manufacture.
AMD hotly disputes this, stating that MirrorBit chips actually cost less to make than StrataFlash processors. However, Handy noted that MirrorBit chips have been out for two years and still represent only a relatively small fraction--a low double-digit percentage--of Spansion's sales.
"They should be able to take MirrorBit down in price to Intel's pain threshold, but so far they haven't been doing that," Handy said.
To top it off, semiconductors are out of favor in the IPO market at the moment, said Bill McLeod, co-director of capital markets at Thomas Weisel Partners, at a presentation at the VentureOne conference in San Francisco this week.
Splitting the company should buoy some of the investors who mostly look at AMD as a processor company, Krewell said. But they should remember the company's past, he added: "People forget that when the processor business was hurting, it was the flash business that kept them going."