While many industry sources see a handful of reasons why such a partnership could work, there are just as many observers see abundant reasons why it would not.
For one, AMD has never put much emphasis on owning a platform. It often touts the "freedom" system designers have when using AMD, citing its reluctance to team up with other companies to set mandated parts lists in its wireless and media PC efforts. Intel has done so with Centrino and Viiv.
Another concern over the pending acquisition is how well AMD can actually manage ATI's graphics chip business. "If AMD tries to integrate the whole thing too aggressively, they will suffer the same fate as Intel did," said Peddie. "You need to leave ATI alone."
GPU development cycles are typically much faster than those for CPUs6-12 months versus 18 months to 2-3 years. "Most integrated CPU-graphics products have either failed or been very short-term successes in the PC market with a bit more longevity in embedded," said Dean McCarron, principal analyst at Mercury Research. For proof, he points to Intel's Timna, and the MediaGX from the defunct Cyrix Corp.
In 2000, Intel hoped to position a heavily integrated CPU at low-end, sub-$600 machines, and some industry analysts predicted that the era of the system-on-chip PC was imminent. Then Intel canceled the project in late 2000 after running into technical challenges with memory controllers and graphics integration.
Some in the industry are estimating the price of the deal as high as $5.5 billion, while questioning whether AMD is positioned to make such a big acquisition work. The company is clearly enjoying good times. Its market share stands at just above 20 percent for the first time in four years, it has about $2.6 billion in cash and it's been making steady progress with its Operton line of server chips, most recently with a design win at Dell Computer.
But AMD also needs to husband its cash for fab expansion in Dresden, Germany, and the transition to the 45-nm node.