It doesn't take long to fall from grace in Silicon Valley.
Advanced Micro Devices is hurtling toward what could be another disappointing period as the second quarter draws to a close next week. Wall Street expects the company to have lost 84 cents per share for the June quarter on $1.26 billion, up 4 percent from the same quarter a year ago. That compares with a loss of $611 million or $1.11 per share in the first quarter and a net profit of $88 million or 18 cents per share in the second quarter of 2006.
AMD needs a fix, and fast. A $2.2 billion debt offering will help keep the lights on in Sunnyvale, Calif., and Austin, Texas--the chipmaker's primary U.S. locations--as some financial analysts had been worried about AMD's cash-flow situation. The company is pinning its near-term hopes on Barcelona, a quad-core server chip that AMD hopes will allow it to shore up the eroding prices of its server chips, the source of most of its pain this year.
But is Barcelona, which is expected sometime in the third quarter, a Band-Aid or a tourniquet? The company's problems could be broader in scope and more than one good product launch can fix.
"AMD really needs to round out their product portfolio," said Jim McGregor, an analyst with In-Stat. "If not, a lot of the market share gains they made they are going to give back. They've got a critical 12 months ahead of them."
Intel is expected to keep the pressure on with server price cuts around the time of Barcelona's arrival. The new realities of chip pricing could force AMD to re-examine its manufacturing strategy. And its ambitious project to integrate the PC processor and the graphics processor is years away from fruition.
An AMD representative said the company will have more to share about its future road map and its near-term plans for becoming profitable once again. The company will report its second-quarter results in July and hold an analyst day shortly thereafter, he said.
AMD's current mess starts with tough competition from chip king Intel. A price war initiated by Intel last year, ostensibly to clear out older products ahead of the Intel Core launch, has persisted into 2007, as AMD has grappled with the market's response to Intel's quad-core server chips launched in late 2006.
But Intel chose a quicker route to four cores by building server chips with two dual-core chips together in the same package. AMD, on the other hand, chose a more complicated design in which all four cores live on the same piece of silicon. The company believes this will deliver a performance advantage over Intel's approach because all the cores can communicate directly with their neighbors.
And supply-chain problems made it even harder for AMD to compete as the calendar turned from 2006 to 2007. The company was already in the process of aggressively discounting its processors when it was forced into even steeper discounts when one of its customers, believed to be Dell, left it stranded with a bunch of unsold chips.
Analysts believe that Barcelona will likely outperform Intel's Clovertown generation of quad-core chips from 2006. The problem for AMD is that Intel is preparing to launch a second generation of quad-core server chips in the fourth quarter of this year. While it's too early to ascertain the performance of those processors, they will likely be extremely competitive with Barcelona, and expected price cuts on the older Clovertown processors will give Intel's customers two options for quad-core technology.
Plus, Intel's plan for 2008 is to mimic the design characteristics that made AMD's Opteron chip a winner. The Nehalem generation of processors will come with integrated memory controllers and point-to-point interconnects, design characteristics that allowed AMD to enjoy a significant performance advantage from 2003 until last year.
"One board maker we met with suggested that their tests indicated that AMD's 65-nanometer Barcelona only bridges the gap between Intel's 65-nanometer Clovertown," Citigroup's Glen Yeung wrote in a recent research note. "The release of 45-nanometer Harpertown (Nehalem) is expected to re-establish the lead for Intel, emphasizing the importance of any delay by AMD."
AMD will face a number of challenges generating revenue over the next 18 months while up against a solid (on paper, at least) road map from Intel. And AMD CEO Hector Ruiz and his management team might have to re-examine the cost side of their company to turn things around.
In April, Ruiz began dropping hints about changing AMD's cost structure in what has come to be known as "asset light." He hasn't said much more beyond that, but a few scenarios seem most likely.
"Intel got a sense of urgency, and AMD needs the same sense of urgency."
--Jim McGregor, analyst, In-Stat
AMD could expand its use of third-party chipmakers. It already uses Chartered Semiconductor's factory to make some processors, but could expand its use of other factories such as Taiwan Semiconductor Manufacturing Co. or United Microelectronics Corp. in Taiwan, which make the graphics chips AMD acquired along with ATI Technologies. Citigroup believes that AMD is almost ready to announce that TSMC will manufacture certain low-end processors for AMD, freeing up capacity at AMD's own Dresden, Germany, facility to make more profitable chips.
The company could also enter into new manufacturing partnerships like the one it has with IBM. Figuring out ways to make ever smaller transistors is a dauntingly expensive proposition. The joint partnership with IBM allows AMD to face those problems with a little help from its friends. This could involve a company like IBM taking a partnership stake in one of AMD's Dresden plants or a proposed facility in upstate New York, nearby IBM's own East Fishkill facility.
It's very unlikely, however, that AMD would get out of the chipmaking business altogether, McGregor said. Notwithstanding company founder Jerry Sanders' famous quote--"Only real men have fabs"--there's a cost advantage to in-house manufacturing, so long as the factory stays full, he said.
AMD's biggest strength at this point in its history is that it's no longer just an upstart chip company best known among gamers. AMD can now count the four largest server makers among its customers, and companies across the Fortune 500 have shown a willingness to put AMD products at the heart of their systems.
But while the company has spent much of the past few years emphasizing its power-efficient designs, it must finally back up all that talk with a true mobile processor design if it wants to pull off its ambitious Fusion project, McGregor said.
AMD has unveiled plans for this mobile processor, code-named Griffin, but it's not expected until the middle of next year. AMD's current Turion mobile processors are really just power-optimized versions of its Opteron chips, while Griffin would follow Intel's approach of designing for low-power applications first and increasing power for systems designed to handle larger amounts of power consumption--like servers.
"Fusion is only going to be successful if they have a new mobile core to go with it," McGregor said. AMD wants to integrate graphics technology from ATI into its processor designs with the Fusion project. The obvious upside is for PC gamers, but the high-speed processors used today mostly for graphics could have all kinds of other applications should programmers be able to find a way to take advantage of that horsepower.
Before AMD gets to Fusion, however, it has to repair the current state of its business. Executives spoke about a restructuring plan in April that would help the company regain its footing, but few details about that plan have emerged.
That doesn't change the fact, however, that Ruiz needs to light a fire under AMD--and soon, McGregor said. "Intel got a sense of urgency, and AMD needs the same sense of urgency."