Skip to Content

A Record-Breaking Optical Chip

Intel researchers have built a superfast silicon chip for optical networking.
June 25, 2008

The road to a faster Internet, data center, and personal computer is paved with silicon. Or so believe researchers at Intel who have unveiled a test chip–made entirely from silicon–that can encode 200 gigabits of data per second on a beam of light. In contrast, the most advanced chips used in today’s fastest optical networks operate at speeds of 100 gigabits per second. And these 100-gigabit chips, which are made from nonsilicon materials, have limitations that Intel’s chip doesn’t: they can’t scale to faster speeds as inexpensively as can those made from silicon.

Light it up: The silver-colored rectangular chip in the middle of the copper-colored holder is Intel’s latest advance in silicon photonics. The chip contains eight modulators that encode data onto light that enters and exits from the side via optical fibers (not pictured). This chip can process 200 gigabits of data per second and is used to test designs that could ultimately process a terabit of data per second.

While silicon is the material of choice in the electronics industry, it has been overlooked in the photonics industry because its optical properties are inferior to those of other semiconductors. Silicon doesn’t produce, detect, and manipulate photons as well as materials such as indium phosphide and gallium arsenide. But within the past few years, optical engineers have been giving silicon a second look and cleverly engineering around some of its natural limitations.

The new Intel test chip splits an incoming beam of light into eight channels. Within each channel is a modulator, a device that encodes data onto light. After the beams are encoded with data, they are recombined. In the tests, each modulator ran at a rate of 25 gigabits per second, and each performed nearly identically, says Mario Paniccia, director of the company’s silicon-photonics lab. He notes that only one modulator was tested at a time but says that in a future paper his team will publish the results from running multiple channels simultaneously. The multiple channels could produce cross talk, electrical or optical activity that could hinder performance. However, preliminary results, Paniccia says, show that due to the design, cross talk is limited.

In 2004, Intel researchers, led by Paniccia, proved that silicon could be used to build a one-gigabit-per-second modulator; in 2005, the team boosted the speed to 10 gigabits per second. Also in 2005, the researchers built a remarkably good all-silicon laser, and in 2006, they introduced a hybrid laser that combines indium phosphide with silicon, allowing a practical telecom laser to be fabricated on a silicon wafer. Most recently, they have sped up the modulator to 40 gigabits and built a silicon detector.

Other companies are now also exploring the capabilities of silicon for photonics. IBM and Sun Microsystems have active research groups, and a startup called Luxtera has already made advances in silicon-based optical interconnects for data centers. At Intel, however, the pieces are coming together to make a single chip that could process a terabit of data in the space of a thumbnail. This chip and its accompanying electronics could replace racks full of expensive hardware that currently occupy rooms at Internet switching stations. And if all goes well, optical devices made of silicon could allow engineers to replace copper wiring in computers with beams of data-encoded light.

“Intel has pioneered a lot of high-speed silicon-photonics devices, and it’s certainly one of the premier research groups,” says Jack Cunningham, co-principle investigator of Sun Microsystems’ proximity interconnect project, which focuses on low-power interchip communication for high-performance computers. Cunningham says that the Intel test chip is another important step in the evolution of silicon photonics. “It’s the right direction in the sense that high-bandwidth optical signaling on silicon chips is very important,” he says.

Paniccia notes that there is still a lot of work to do before Intel’s optical chips find their way to market. Instead of having only 8 modulators, the goal is to have 25 on a chip. In addition, the modulators will run faster–at 40 gigabits per second. And it’s still unclear how light will be piped into the modulators in the future. Currently, it enters via an optical fiber on one end of the device, but future versions of the chip may include hybrid lasers fabricated on the chip. Paniccia hopes that in three to five years, Intel’s silicon-photonics chips will be ready for market.

Keep Reading

Most Popular

The inside story of how ChatGPT was built from the people who made it

Exclusive conversations that take us behind the scenes of a cultural phenomenon.

Sam Altman invested $180 million into a company trying to delay death

Can anti-aging breakthroughs add 10 healthy years to the human life span? The CEO of OpenAI is paying to find out.

ChatGPT is about to revolutionize the economy. We need to decide what that looks like.

New large language models will transform many jobs. Whether they will lead to widespread prosperity or not is up to us.

GPT-4 is bigger and better than ChatGPT—but OpenAI won’t say why

We got a first look at the much-anticipated big new language model from OpenAI. But this time how it works is even more deeply under wraps.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.