Global chipmakers that have failed to follow the law over the past two decades have done so at their own risk. The reverse may soon be true.
The law of which I speak was posited by Gordon Moore, Intel founder and semiconductor pioneer. In 1975, Moore observed that the number of transistors packed into a chip doubled every two years thanks to the industry's incessant shrinking of circuitry .
TSMC has been pretty consistent in sticking to Moore's Law by introducing a new production technology on average every two years. This has allowed chips to become faster, smaller and more energy-efficient.
When the company released its third-quarter earnings Thursday afternoon, it once again wowed sell-side analysts with an array of crowd-pleasing numbers. Net income beat estimates and climbed to a record, while sales for the coming quarter look stronger than they'd hoped.
At the heart of TSMC's ascendance to become the world's largest contract chip foundry has been its engineers' skill in continually finding ways to reduce transistor size; its sales team's success in pitching that technology to clients such as Qualcomm, Apple and Mediatek; and its finance team's repeated ability to find the cash -- internally and externally -- to fund the increasingly extravagant cost of the equipment required to make it happen.
During TSMC's rise, observers speculated that Moore's Law was coming to an end because we'd no longer be able to keep up the pace of development. Yet every time that apocalyptic scenario approached, a new technology would appear to prolong the maxim a little while longer.
That seems to be what's happening now for TSMC and the ecosystem it inhabits. Qualcomm to ASML and Nvidia to Merck Group all rely on this continued pace of development and upgrades to push their latest chips, equipment and chemicals. While the Stinson Rule applies -- new is always better -- the more important truth is that new is also more expensive. That helps increase revenue, widen margins and boost profits.
But no matter how awesome the technology, you can't increase sales and profitability indefinitely if the major demand driver (handsets) continues to slow, and there's nothing yet to replace it.
For TSMC, 28-nanometer circuitry (the technology behind many of Qualcomm's Snapdragon processors) was the standout winner of the past decade, accounting for NT$914 billion ($29 billion) in company revenue since it was introduced five years ago, making it the second-most popular in the company's history. Interestingly, though its most successful technology, 0.15- and 0.18-micron (that is, 150- and 180-nanometer) was introduced 15 years ago and still garners almost 10 percent of sales. That's like the Model T accounting for one in 10 cars sold five decades after Ford released its Mustang.
Recently, I wrote that raw processing power is no longer the driver and differentiator that it once was in smartphones (batteries are the new frontier). And with handset sales slowing, the industry can no longer take for granted that a newer, smaller semiconductor circuitry will receive the warm reception it once did.
Slipped in among the numbers Thursday was news that TSMC sees capital expenditure at the bottom end of its earlier $9.5 billion to $10.5 billion guidance. The company's explanation is that it's ramping up the next technology quicker than expected, yet more telling is its pledge to maintain capital spending as a proportion of sales -- called capital intensity -- over the next few years.
It's all well and good that TSMC is beating the competition in the rush to the latest technology, but spending only helps when there's someone to buy your product. With the waning need for something smaller and faster, the risk isn't that the company won't keep up with Moore's Law, but that it will.
This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
Moore wrote a paper in 1965 pointing to a doubling every year. In 1975 he revised that prediction to every two years.
To contact the author of this story:
Tim Culpan in Taipei at email@example.com
To contact the editor responsible for this story:
Matthew Brooker at firstname.lastname@example.org