Just seven or eight years ago, it was multiple generations behind Intel in chip production technology. 15 years ago, it was a bulk producer of cheap commodity chips with no track record in high-performance silicon, and 30 years ago, it barely existed. So, how has TSMC come to rule the world?
Today, TSMC can claim indisputable world leadership in chip manufacturing with at least one generational advantage over Intel, arguably two. It produces roughly half of the global supply of customerdesigned chips and over 90 percent of those manufactured on cutting-edge processes. Its chips go into not just PCs and phones, but cars, military equipment, medical hardware, and everything else we depend on these days.
TSMC is based in Taiwan, an island territory operating as a de facto independent state without being formally recognised as such by the international community. For the past 50 years or so, China has viewed Taiwan as a rogue element and has officially maintained the unilateral right to repossess the island. As far as China is concerned, it and Taiwan remain a single nation.
Modern life as we know it depends to a large extent on the output of a single company based in a country that sort of doesn’t exist, itself situated on an island separated from China by a sliver of water that might just be the most dangerous place on Earth. How, exactly, did we get here and what does the future hold? If you’re a tech fanatic of a certain vintage, viewing everything through the prism of the PC, you’ll probably be familiar with TSMC. But you’ll also instinctively think of it as something of a budget outfit, operating slightly behind the curve. Sure, TSMC was critical – it enabled the creation of important PC components, such as graphics chips. Neither Nvidia nor ATI (before the latter was acquired by AMD) produced their own chips. They designed them, of course, but paid third parties, and often that was TSMC, to do the manufacturing honours.
So, TSMC mattered. But GPUs historically did not require the absolute best in chip production technology. Decent transistor density was certainly desirable. But GPUs were not designed for and did not demand the last word in clock frequencies, typically running at a third or less of the speed of contemporary CPUs. In short, TSMC simply didn’t compare with Intel when it came to the fine art of turning sand into ultra-high-performance silicon.
For AMD, increasingly that was a problem. AMD had been forced to spin off its own chip production facilities, also known as fabs or foundries, in 2009 in an attempt to keep the company alive. Building new fabs is a multi-billion dollar enterprise and AMD simply didn’t have the cash. The idea was that selling off its existing fabs would give AMD a desperately needed cash injection while allowing the newly independent manufacturing unit to acquire external customers and outside investment, both of which would bankroll the next-gen production units needed to keep up with Intel’s world-beating fabs.