You are here

Dutch Company Aims to Make Chips Do More

VELDHOVEN, the Netherlands— ASML Holding NV, a little-known company based next to corn fields here, may hold the answer to a question hanging over the global semiconductor industry: how to make chips do more while keeping them the same, compact size.

The industry’s past prowess has been codified into what's been called Moore's Law,  named after an observation Intel Corp. co-founder Gordon Moore first made in 1965. He postulated that chip makers could double the number of transistors in—and boost the performance of—a typical microprocessor every two years.

Last year, though, Intel Chief Executive Brian Krzanich warned that after decades of incredible leaps, that timeline was slipping closer to every 2.5 years. Some in the industry feared the eventual death of Moore’s Law, a rule of thumb underpinning modern computing.

ASML believes its breakthrough technology can postpone the demise. “I’m not concerned yet about the next 10-plus years,” said Hans Meiling, who oversees ASML’s effort trying to solve this problem.

Many in the industry, including big backers like Intel itself and Samsung Electronics Co., are hoping ASML can quicken the pace of innovation once again. With around 15,000 employees and €6.3 billion ($7.05 billion) in revenue last year, the company manufactures equipment that makes chips—specializing in a field called photolithography. Specifically, ASML uses light rays to essentially lay out billions of transistors—the brain cells of a chip—in a microprocessor.


Employees in protective clothing at an ASML factory in Veldhoven, the Netherlands, in 2013. PHOTO: MICHAEL KOOREN/REUTERS

Optical experts such as Canon Inc. and Nikon Corp. are also active in this market, but ASML dominates it. For the past decade, ASML has been focused on making more sophisticated light rays, which it uses to create smaller transistors that can be packed more tightly onto silicon wafers.

“It is hugely important for the [semiconductor industry’s] ability to continue Moore’s Law,” said Greg McIntyre of Imec, a Belgium-based chip-research institute. ASML’s latest chip-making machines could be a big commercial leap forward in this pursuit, he said.

To get an idea of the scale of the challenge: A strand of human hair is about 75,000 nanometers in width. The current industry-standard chip-making machine produces a light ray that draws 38-nanometer-wide lines. ASML’s technology, called extreme ultraviolet lithography, or EUV, draws lines that have a width of 16 nanometers.

Here’s how the process works: Manufacturers first melt silicon, an abundant material found in beach sand and prized for its semi-conductive properties, allowing for a controlled flow of electricity. The melted silicon is cooled into a sausage-like structure, which is then sliced into thin wafers that look like an LP record and given a photosensitive coating. Then chip makers, using thin light rays, draw a grid of lines.

Imagine those grids of lines as streets in a city. Chemicals etch away the “streets,” and the remaining “buildings” essentially become transistors. Compared with conventional machines, ASML’s EUV tools make those streets narrower, and thus the buildings smaller. That means more transistors, and more computing power.

By packing more transistors on chips, manufacturers could increase their computing speed or memory-storage capacity, among other things. ASML’s new technology may keep Moore’s Law in play.

Four years ago, ASML’s EUV technology was considered so promising that Intel, Samsung and Taiwanese Semiconductor Manufacturing Co. pledged to invest a combined $6 billion in ASML. Samsung in September sold half of its ASML shares to raise cash, but said its EUV partnership with the company wouldn’t change.

ASML started in the early 1980s as a joint venture co-founded by electronics giant Royal Philips NV. In the 1990s, it spun off from Philips and made rudimentary versions of the chip-making equipment it sells today. Now, all major chip makers use its machines.

ASML, Canon and Nikon all produce conventional machines, which sell for between about $55 million and $60 million apiece. They can make the smaller grids as the EUV machines do, but only after a number of time-consuming workarounds.

The newest machines from ASML, which cost about €95 million each, aren’t yet perfect. The company has sold 12 and is currently working with those buyers to optimize the machines at the chip makers’ sites. ASML said it expects companies to start using the machines for high-volume production by 2018 or 2019.

“It’s a combination of physics, chemistry and mechanics that is pretty complex,” said Mr. Meiling, who has a doctorate in physics and oversees ASML’s EUV program. “It has taken us a long time to get where we are.”

No rival has the research-and-development budget to attempt to compete with ASML’s EUV business, but the Dutch company still faces a challenge in getting new orders, said Bernstein Research analyst Pierre Ferragu. Chip makers “will all hesitate until the last minute about the best approach to introduce EUV in their production lines,” said Mr. Ferragu, especially given the “uncertainties” with the “extremely complex technology.”

Intel isn’t yet convinced. The company has been testing ASML’s EUV systems and published positive results, but is also prepared to continue using the time-consuming trick.

One of ASML’s many challenges is keeping the machine running 24 hours a day—the sort of efficiency chip-making customers prize. The machines use specks of liquid tin, which when properly stimulated emits the EUV light. Sometimes, though, that tin contaminates the mirrors used to project the light rays, requiring maintenance.

ASML acknowledges it still has some way to go. It says its EUV machines can make about 1,500 wafers a day and can be relied on to be operational more than 80% of the time. Its conventional machines produce more than 5,000 wafers a day—but with thicker lines, at least before the tedious workarounds—and are available 95% of the time. Customers demand that the EUV machines run at least 90% of the time.

“We’re not there yet,” said ASML’s Mr. Meiling.