[ad_1]
TSMC Headquarters. Credit: TSMC
For more than 30 years, Intel has dominated chip manufacturing by producing the most important component of most computers in the world. This race is now threatened by a company that many Americans have never heard of.
Taiwan Semiconductor Manufacturing Co. was established in 1987 to manufacture chips for businesses that do not have the money to build their own facilities. This approach was rejected at the time by the founder of Advanced Micro Devices, Jerry Sanders. "Real men have tricks," he joked at a conference, using industry jargon for factories.
These days, the ridicule has given way to envy as TSMC 's factories have stood up to challenge Intel at the pinnacle of the $ 400 billion industry. AMD recently chose TSMC to manufacture its most advanced processors, having created its own troubled factories years ago.
The threat that TSMC represents for Intel reflects a radical change in chip making: one company after another hires TSMC to manufacture the chips that they design. TSMC, based in Hsinchu, has many customers, including giants Apple and Qualcomm technology, second-tier players such as AMD and minnows such as Ampere Computing. The explosion of the components thus built has allowed TSMC to acquire the technical know-how needed to manufacture the smallest, most efficient and most powerful chips in the highest volumes.
"This is a unique situation in the last 50 years," said Renee James, former Intel number 2, at the head of the Ampere startup. Her company is less than two years old and yet she is interested in Intel's mainstream server chip market. What Ampere thinks he can compete with is a stumbling-block on Intel's part.
It's been a decade since Intel faced major competition and its 90% turnover in IT processing will again produce record results this year. But some Wall Streeters are worried and their rivals are emboldened because TSMC has a real chance to replace Intel as the best chip maker on the market. Last year, the Taiwanese company has for the first time amassed a market value superior to that of its American rival.
Production rights of semiconductors on production are judged by the width of the space between the lines on the tiny circuits that ensure the function of the chips. Reducing this gap (measured in nanometers or billions of meters) gives designers the ability to make chips faster, less power hungry, store more data, or simply cost less. In high-end processors, where Intel makes the most of its money, space is paramount. A Xeon server processor stuffs billions of transistors into an area the size of a postage stamp.
Intel was the first to use 14 nanometer technology on a scale in 2013, according to Goldman Sachs. The 10-nanometer technology will not be ready for prime time until the end of 2019, by far the longest wait in its history. TSMC has gone from 20 nanometers to 7 nanometers at the same time.
The Intel blocking revolves around the yield, the number of good chips that emerge from each production cycle. In factories that cost about $ 7 billion and operate 24 hours a day, producing millions of chips every month, the slightest failure can be financially disastrous. Intel has not yet sufficiently eliminated these manufacturing issues, and the company will not switch to a 10-nanometer output until everything will work properly.
Sanders' current successor at AMD, Lisa Su, president and CEO, does not have to worry about this situation as the company sold its factories and let TSMC handle the complex production.
"It's one of the best decisions we've made," says Su. "It allows us to manage risks and focus on the elements that make the product great."
With the help of TSMC, Su pursues a goal that Sanders has never achieved: a credible and enduring challenge for Intel's blocking in computing. AMD is now telling investors and customers that its new chip models will surpass those of Intel. TSMC makes this competition possible, even though AMD has about a tenth of Intel's R & D staff and budget.
TSMC, however, did not catch Intel alone. The real break in society took place a decade ago, when smartphones began to fill the pockets of consumers. Intel has used mobile phone chips, but the US company has never engaged its best production and design systems in the region, preferring to prioritize its existing PC and cash cow server operations.
When smartphone sales took off, phone manufacturers resorted to other processors from companies like Qualcomm. Or they have designed their own ARM technology, like Apple. And the TSMC factories rejected these components.
The volume of smartphones is now almost six times larger than that of the PC industry. This gives TSMC the benefit of a large volume manufacturing experience that previously belonged to Intel.
With billions of transistors on chips, a problem with a small number of these tiny switches can make the entire component useless. Production cycles can take up to six months and involve hundreds of steps requiring careful attention. Whenever there is a mistake, the plant operator has a chance to make adjustments and try a new approach. If the change works, this information is kept for the next challenge. The more production there is, the better. And TSMC has the most nowadays.
"TSMC is just delivering the latest chips on time, without any mistakes," said Mark Li, an analyst at Sanford C. Bernstein. He believes that Intel's leadership in server and server chips, as well as its pricing power, is compromised by the fall of its smartphone and the hard-won consistency of TSMC.
Yet, this is far from the first challenge that Intel faces. The company is working on its production issues and, in the meantime, will deliver new chips built with existing production technology that, she says, will keep the competition at bay.
Navin Shenoy, head of Intel's servers division, says that nanometer-based production metrics have never been the only factor of success (although the company has enjoyed talking more about it in the past) . Intel's short-term solution is to design better chips using the old production technology.
"I am confident that we will offer customers what matters to them – the performance of the system," he said.
Historically, the company has crushed its rivals using a research budget that has wiped out everything that is happening in the industry. But TSMC's approach undermines this advantage.
While Intel still exceeds TSMC in investments in new installations and equipment, the dashboard is reversed when you combine TSMC's customer search budgets such as Qualcomm, Apple, Nvidia Corp. and Huawei Technologies Co.
According to Goldman Sachs, combined budgets of TSMC customers are not only larger than those of Intel, but the gap is widening. By 2020, they will spend nearly $ 20 billion, according to his estimate, at least 4 billion more than Intel.
"It's now a self-fulfilling prophecy," says Debora Shoquist, executive vice president of operations at Nvidia. "They are the best and the best are the best."
-Bloomberg News
[ad_2]
Source link
