GlobalFoundries CEO Doug Grose Declares Sky-Rocketing Costs Slow Product Innovation
The far-reaching effects of the global economic downturn could make it easy to blame the IC industry's problems on short-term market effects, but Doug Grose, CEO of GlobalFoundries, believes the fundamental problem confronting the industry is economic, not technical. According to Grose, "Only a handful of providers can afford the costs and keep up with Moore's Law...The IDM (Integrated Device Manufacturer) model is on a slow, dying path."
The numbers behind Grose's rhetoric lend credence to the man's opinion though. When AMD announced its own plans to split into two separate companies, it provided the following diagram:
Expect to see that diagram again, and in the very near future. According to the cost-per-process conversion curve, it'll cost AMD almost twice as much money to move from 45-32nm than it did to shift to 90nm-65nm. Fab costs rise at a similar rate; the tremendous up-front cost of building a foundry virtually guarantees that only a handful of companies can afford to construct or sustain these behemoths. If we guesstimate a bit (and consult a handy timeline), AMD introduced 130nm technology in 2003, 90nm in 2005, 65nm in 2007, and 45nm, as we all know, came along in 2009. The growing cost of moving from node to node in an effort to match Intel, combined with the debt the CPU manufacturer was already carrying, is part of why we now speak of the two as two separate entities.
One of the major bones of contention in AMD's antitrust lawsuit against Intel, for those of you who care about that sort of thing, is that the allegedly illegal/monopolistic controls Intel imposed on the market during the 2003-2005 period (give or take six months) prevented AMD from making anywhere near the kind of cash it would've raked in had these constraints not been artificially imposed. Intel has no comment on these allegations to date, beyond saying that it believes its actions have benefited consumers while the company itself has remained within the letter of the law.
If we grant AMD its premise—strictly for theoretical reasons—the company has a real point. AMD held a real competitive edge against Intel in the desktop and server markets during those years. AMD would argue that its share of those segments (and subsequent profits) would've both been much higher, and that income would've been used to drive the adoption of next-generation technologies like 45nm. Intel would undoubtedly counter that AMD did gain market share as a result of its improved product line, and that it doesn't get to pick an arbitrary number out of the air and claim it "would" have gotten this, if not for Intel's interference. Expect the five billion+ purchase of ATI to come up, too.
Whether you believe AMD's side of the story or not, the trend Grose is decrying is very real.Semiconductor manufacturers have been allying themselves together in increasing numbers, whether it's through joint licensing agreements with IBM or GlobalFoundries' buyout of Chartered. This new-found cooperative spirit among the various manufacturers probably won't vanish, even when the dollar signs begin edging back into the black. In good times or in bad, "going it alone" is quickly becoming a bleaker road to travel.
The numbers behind Grose's rhetoric lend credence to the man's opinion though. When AMD announced its own plans to split into two separate companies, it provided the following diagram:
Expect to see that diagram again, and in the very near future. According to the cost-per-process conversion curve, it'll cost AMD almost twice as much money to move from 45-32nm than it did to shift to 90nm-65nm. Fab costs rise at a similar rate; the tremendous up-front cost of building a foundry virtually guarantees that only a handful of companies can afford to construct or sustain these behemoths. If we guesstimate a bit (and consult a handy timeline), AMD introduced 130nm technology in 2003, 90nm in 2005, 65nm in 2007, and 45nm, as we all know, came along in 2009. The growing cost of moving from node to node in an effort to match Intel, combined with the debt the CPU manufacturer was already carrying, is part of why we now speak of the two as two separate entities.
One of the major bones of contention in AMD's antitrust lawsuit against Intel, for those of you who care about that sort of thing, is that the allegedly illegal/monopolistic controls Intel imposed on the market during the 2003-2005 period (give or take six months) prevented AMD from making anywhere near the kind of cash it would've raked in had these constraints not been artificially imposed. Intel has no comment on these allegations to date, beyond saying that it believes its actions have benefited consumers while the company itself has remained within the letter of the law.
If we grant AMD its premise—strictly for theoretical reasons—the company has a real point. AMD held a real competitive edge against Intel in the desktop and server markets during those years. AMD would argue that its share of those segments (and subsequent profits) would've both been much higher, and that income would've been used to drive the adoption of next-generation technologies like 45nm. Intel would undoubtedly counter that AMD did gain market share as a result of its improved product line, and that it doesn't get to pick an arbitrary number out of the air and claim it "would" have gotten this, if not for Intel's interference. Expect the five billion+ purchase of ATI to come up, too.
Whether you believe AMD's side of the story or not, the trend Grose is decrying is very real.Semiconductor manufacturers have been allying themselves together in increasing numbers, whether it's through joint licensing agreements with IBM or GlobalFoundries' buyout of Chartered. This new-found cooperative spirit among the various manufacturers probably won't vanish, even when the dollar signs begin edging back into the black. In good times or in bad, "going it alone" is quickly becoming a bleaker road to travel.
