A once-in-a-decade change involving one of the most fundamental components of computer chips threatens to reshuffle the rankings of chip giants in the coming years.


Intel, Samsung and TSMC are racing to make a generational leap in transistor technology. This leap must be made in order to achieve the computing demands required to achieve anything close to the idea behind the Metaverse, to produce artificial intelligence that is not a joke, to make truly self-driving cars, or even to make applications load faster.


This new generation of design is called "gate-all-around (GAA)". Using new materials, redesigned manufacturing tools, each costing tens of millions of dollars, the new gates accomplish one thing: They more tightly control the current each transistor receives. Modern chips can have more than 30 billion transistors on a device, and in some cases even tens of billions. Gartner expects chipmakers to generate about $5 billion in revenue from the new technology by 2025, up from zero last year.


Chip companies must provide more computing power each year to realize the future the tech giants have promised. Doing this requires the use of some of the most complex and expensive manufacturing equipment on the planet, as well as the development of more creative ways to improve fundamental aspects of chip architecture. This means making atomic-sized features even smaller. The process, loosely known as Moore's law, has kept the chip industry booming for half a century, but it's getting harder.


"We're definitely slowing down significantly," said Kevin Moraes, vice president of Applied Materials.


Chipmakers are improving high-volume production and performance by combining advanced tools, such as extreme ultraviolet lithography machines, and technologies that help squeeze more features on each piece of silicon. Often, manufacturers use technology and technology improvements as process nodes using smaller and smaller nanometers.


But another solution to an increasingly difficult problem is to further refine the fundamental building block of every chip: the transistor.


“The process node is just one indicator of progress,” said Jack Gold, principal analyst at J.Gold Associates. "I think it's more about the design of the transistor itself, not just the process node. To that end, Intel and others have been advancing the design of the transistor, especially the way the gate is designed."


Better gate, faster chip

The "gate" is the tiny part on each transistor that controls whether or not the transistor receives power -- a bit like stepping on a garden hose to turn the water on or off -- to represent the 0s and 1s that make up the data bits. However, gates, like hoses, can be imperfect, and even with the most advanced designs, there will be some current leakage.


"As you start making things smaller, you'll find that some of the electrical properties of the small components are not as good as they would be with the larger components," said FeibusTech analyst Mike Feibus. "There is a rule of thumb that the smaller you are, the more More current leakage means more heat."


Because chipmakers know that some of the power escapes the gate, it becomes increasingly difficult to continue to shrink functions to achieve the power consumption and performance expected by new designs. To make gates and transistors that are more efficient and produce less energy, chipmakers spend billions of dollars inventing the next and better way to develop them.


David Kanter, executive director of MLCommons, said: "A fully functional gate-based design will have better performance and efficiency than (existing) designs, potentially changing the competitive position of many high-performance products."


Surrounding four sections of the transistor with material, instead of the current three-sided design, allows the gate to better regulate current flow. At the atomic scale of a chip, better control over currents gives designers new options: most importantly, making them smaller.

Smaller transistors allow designers to squeeze more transistors onto a single chip, and adding more of these tiny features roughly means an increase in the chip's computing power.


"With better thermal and power characteristics, which means you can increase the power of the power supply, all else being equal, and you'll have higher clock rates, and you don't need to provide a high-performance design," Feibus said. Exotic cooling, so you can keep costs down."


Manufacturers who successfully apply next-generation gate technology to high-volume production will be able to create chips with leaps in computing power that are not possible with current process technologies.


the fastest

All three of the largest chipmakers are trying to figure out how to take advantage of this promising technology. The problem isn't producing a batch of chips with new capabilities; it's producing hundreds of thousands or even millions of chips at a scale large enough to meet the demands of state-of-the-art technology.


Samsung may have an advantage. Kinam Kim, one of the company's executives, co-authored an early technical paper on the new gate technology nearly 20 years ago. The company was one of the first to commit to producing the next-generation gate and discussed a path to the technology in 2017. In April, Samsung said it was on track to mass-produce chips with the new processors this year.


However, there is reason to be skeptical of its progress. Gaurav Gupta, a chip analyst at Gartner, said there had been numerous reports that Samsung had problems ensuring that its latest manufacturing process could produce a sufficient number of working chips, and had delayed plans for a new gate several times. release date. Samsung did not respond to a request for comment.


So even if Apple rolls out new designs on some of its chips, it may not be a viable option for big customers that demand reliability -- for example, a six-month delay in the launch of a new iPhone would be detrimental to Apple's business catastrophic impact.


"Samsung has been saying they're going to be the first to do it all in the 3-nanometer range," Gupta said. "They've been pushing, the next six months, six months, six months," he said. . They are now saying they hope to have it done by the end of the year."


Intel said its current plan is to roll out the technology to high-volume customers in 2024. There are also reasons to be skeptical of Intel's efforts; the company has experienced years of delays in production improvements over past generations. It was also the last company to adopt EUV lithography, which is widely regarded as a necessary technology for future improvements. An Intel spokeswoman said the company is still moving toward its roadmap goals.


But Intel is the first company to successfully mass-produce chips with the current-generation gate, suggesting it has the technical institutional knowledge to do it again. Under CEO Pat Gelsinger, the company's direction is much clearer than in the chaotic years before him. Still, Intel has a lot to prove, and part of Gelsinger's plan includes what Gupta calls a "very aggressive" strategy to regain its manufacturing leadership.


As things stand, TSMC will be the last major manufacturer to adopt the new gate technology. Executives were tight-lipped, and the company did not respond to a request for comment, but TSMC disclosed that it plans to launch the product sometime in 2025. Overall, TSMC is a conservatively run business that tends to be risk-averse if possible, and adopting a new, unproven new gate technology is a big risk, said people familiar with TSMC's business.


But the new market may have less of an impact on TSMC than its rivals. Most of its customers buy smartphone chips, which the company has shown it can make with EUV tools that Intel and Samsung have struggled to deploy. Therefore, TSMC may temporarily be able to use other technologies to make its silicon material perform better until it can roll out a new gate design.


"The point is, it's all about power and performance, and at the advanced nodes of the foundry business, more than 90 percent of customers are TSMC customers," Gupta said, adding that if TSMC uses manufacturing with existing gate technology Better than Samsung, then TSMC won't lose customers.

For chip designers, sticking with the current generation of transistors also has its advantages. The new gate requires certain aspects of the chip design to be scrapped, and industry analysts say the road to making a chip with the new gate could easily exceed two to three years. So if chipmakers like Qualcomm, AMD or Nvidia bet on a new generation of chips this year and that doesn't happen, they could be seriously hurt.