Although fiber-optic cables today are fast, converting their photons to electric signals at the internet server level still uses a lot of electricity.

The Japanese telecom firm NTT and the Tokyo-based electronics giant Toshiba are working on new ways around this problem. In November, the duo demonstrated high-speed factory production via an optical and wireless network that was controlled from a data center 300 kilometers away.

They described the demo as an industry first—of the kind that NTT has lately been promoting to convince the tech world that photonics will form a “next-generation information and communications infrastructure.”

Is the Internet’s Bottleneck Inside the Server Rack?

Optical fiber revolutionized data transmission decades ago. However, it still requires components such as electronic routers and transceivers to convert data back and forth between electrical and optical signals. In traditional fiber networks, information is carried at the data center electronically, which in a high-speed setting can lead to packet loss as well as substantial speed and energy limitations for the network. Photonic systems encode information directly into light—using photon number, polarization, phase, and amplitude to encode and transmit the signals through optical fibers.

NTT says its Innovative Optical and Wireless Network (IOWN) photonics platform can reduce the power consumption of telecom networks to one-hundredth of what they are now, increase data capacity 125 times and cut network latency to a fraction of a percent of its current levels. Meanwhile, the power footprint of data centers in the AI era is expanding rapidly and is expected to more than double. In fact, according to Fatih Birol, executive director of the International Energy Agency, data centers’ electricity-consumption footprint worldwide is expected to rival that of Japan by 2030.

“We need to think differently to overcome this,” says C. Sean Lawrence, cohead of NTT’s IOWN Development Office. “The core idea is to move from electrical wiring to optical, inside data centers, between circuit boards in servers, between silicon packages on circuit boards, and eventually between the silicon die inside a package. We think we can revolutionize high-performance data transmission and computing by making this shift.”

Putting the Photonic Chip to the Test

NTT faces challenges of miniaturization of optical components and the high cost of getting them into chips. It began offering elements of IOWN to data centers in 2023, the same year it established NTT Innovative Devices to develop and manufacture what they’re calling photonic-electronic convergence (PEC) devices. PECs are similar to pluggable optical transceivers that convert digital optical and electrical signals. NTT says putting optics and electronics into a single package yields lower power and heat compared with that of conventional electronics for networking and computing.

The company has been selling its vision via demos that include long-distance data center transmission. Collaborating with Chunghwa Telecom, it organized colorful “Cho-Kabuki” performances in which stages in Osaka and Taipei, some 1,700 km apart, were linked through photonics, video, and a large onscreen stage, allowing actors at either end to interact. The time lag, barely noticeable, was 17 milliseconds.

NTT later showed off PEC hardware at its Tokyo research center. Among other IOWN demos, the center showcased a mock TV studio. NTT says the board-to-board prototype used in the Kabuki show has a capacity of 51.2 terabits per second and relies on second-generation PEC switches. NTT says it also developed resource-control tech to optimize the use of hardware resources, and by combining that with PEC switches, it was able to lower power consumption compared with that of conventional optical computing.

The company is partnering with U.S. chipmaker Broadcom and others to commercialize the second-generation PEC in 2026. The hardware is a step in NTT’s envisaged road map that calls for optical communication between boards as the second phase of IOWN, followed by interchip links from 2028 and intrachip connections from 2032.

“Package-to-package connections are under development,” says Yosuke Aragane, the other leader of the IOWN Development Office. “We are developing production technologies with diverse ecosystem partners and a government funding program. The die-to-die connection is under consideration. However, reviewing the history, I believe the connection could be an essential technology in the early 2030s.”

Can NTT Convince the World to Switch?

NTT knows it can’t pull off this transformation by itself, so in 2020 it joined with Sony and Intel to found a photonics ecosystem called the IOWN Global Forum. It now has more than 160 members, including chip and server makers as well as internet companies like Google and Microsoft.

IOWN joins a two-decade-old initiative in Europe called Photonics21, a public-private partnership aimed at boosting the continent’s photonics industry.

NTT, however, has a mixed record when it comes to popularizing new technologies. In 1999, when it was one of the world’s most valuable companies, it failed to sell its groundbreaking “i-mode” cellular internet overseas. Today, it has far less global clout.

“Telcos have a history of missing out on opportunities like the cloud and AI, but their one strength is edge-network connectivity, so this is their last chance to claim some territory,” says Roy Rubenstein, an analyst at the research firm LightCounting, based in Eugene, Ore. “What’s unusual here is we are seeing a telco-led initiative and support for it. I think NTT’s road map is realistic and matches that of industry in general, but it can’t do it alone, and even with all these companies it’s not enough.”

“With the advent of AI,” Rubenstein adds, “computing has returned to the center of everything. If the AI boom slows, then the urgency will disappear. But if AI continues as it has done, in five years it will be much closer to that vision.”

Takasumi Tanabe, a professor of electronics and electrical engineering at Tokyo’s Keio University, says IOWN is contributing to important R&D in silicon photonics and optical packaging.

“At the device level, some aspects are more challenging,” Tanabe says. “A completely ‘all-optical’ system, in which electronics are removed entirely, may not be feasible with the current state of device physics. Electronics will still be necessary for control, modulation, and signal processing. Even so, I expect photonic devices to play an increasingly important role in the most critical parts of future systems, where low-power consumption, high bandwidth, and low latency are required.”

“While some elements are ambitious,” Tanabe adds, “the essential ideas behind IOWN are realistic, and the initiative has stimulated valuable advancements in photonic technologies.”

​Although fiber-optic cables today are fast, converting their photons to electric signals at the internet server level still uses a lot of electricity.The Japanese telecom firm NTT and the Tokyo-based electronics giant Toshiba are working on new ways around this problem. In November, the duo demonstrated high-speed factory production via an optical and wireless network that was controlled from a data center 300 kilometers away. They described the demo as an industry first—of the kind that NTT has lately been promoting to convince the tech world that photonics will form a “next-generation information and communications infrastructure.”Is the Internet’s Bottleneck Inside the Server Rack?Optical fiber revolutionized data transmission decades ago. However, it still requires components such as electronic routers and transceivers to convert data back and forth between electrical and optical signals. In traditional fiber networks, information is carried at the data center electronically, which in a high-speed setting can lead to packet loss as well as substantial speed and energy limitations for the network. Photonic systems encode information directly into light—using photon number, polarization, phase, and amplitude to encode and transmit the signals through optical fibers. NTT says its Innovative Optical and Wireless Network (IOWN) photonics platform can reduce the power consumption of telecom networks to one-hundredth of what they are now, increase data capacity 125 times and cut network latency to a fraction of a percent of its current levels. Meanwhile, the power footprint of data centers in the AI era is expanding rapidly and is expected to more than double. In fact, according to Fatih Birol, executive director of the International Energy Agency, data centers’ electricity-consumption footprint worldwide is expected to rival that of Japan by 2030. “We need to think differently to overcome this,” says C. Sean Lawrence, cohead of NTT’s IOWN Development Office. “The core idea is to move from electrical wiring to optical, inside data centers, between circuit boards in servers, between silicon packages on circuit boards, and eventually between the silicon die inside a package. We think we can revolutionize high-performance data transmission and computing by making this shift.” Putting the Photonic Chip to the TestNTT faces challenges of miniaturization of optical components and the high cost of getting them into chips. It began offering elements of IOWN to data centers in 2023, the same year it established NTT Innovative Devices to develop and manufacture what they’re calling photonic-electronic convergence (PEC) devices. PECs are similar to pluggable optical transceivers that convert digital optical and electrical signals. NTT says putting optics and electronics into a single package yields lower power and heat compared with that of conventional electronics for networking and computing. The company has been selling its vision via demos that include long-distance data center transmission. Collaborating with Chunghwa Telecom, it organized colorful “Cho-Kabuki” performances in which stages in Osaka and Taipei, some 1,700 km apart, were linked through photonics, video, and a large onscreen stage, allowing actors at either end to interact. The time lag, barely noticeable, was 17 milliseconds. NTT later showed off PEC hardware at its Tokyo research center. Among other IOWN demos, the center showcased a mock TV studio. NTT says the board-to-board prototype used in the Kabuki show has a capacity of 51.2 terabits per second and relies on second-generation PEC switches. NTT says it also developed resource-control tech to optimize the use of hardware resources, and by combining that with PEC switches, it was able to lower power consumption compared with that of conventional optical computing. The company is partnering with U.S. chipmaker Broadcom and others to commercialize the second-generation PEC in 2026. The hardware is a step in NTT’s envisaged road map that calls for optical communication between boards as the second phase of IOWN, followed by interchip links from 2028 and intrachip connections from 2032. “Package-to-package connections are under development,” says Yosuke Aragane, the other leader of the IOWN Development Office. “We are developing production technologies with diverse ecosystem partners and a government funding program. The die-to-die connection is under consideration. However, reviewing the history, I believe the connection could be an essential technology in the early 2030s.”Can NTT Convince the World to Switch?NTT knows it can’t pull off this transformation by itself, so in 2020 it joined with Sony and Intel to found a photonics ecosystem called the IOWN Global Forum. It now has more than 160 members, including chip and server makers as well as internet companies like Google and Microsoft. IOWN joins a two-decade-old initiative in Europe called Photonics21, a public-private partnership aimed at boosting the continent’s photonics industry.NTT, however, has a mixed record when it comes to popularizing new technologies. In 1999, when it was one of the world’s most valuable companies, it failed to sell its groundbreaking “i-mode” cellular internet overseas. Today, it has far less global clout.“Telcos have a history of missing out on opportunities like the cloud and AI, but their one strength is edge-network connectivity, so this is their last chance to claim some territory,” says Roy Rubenstein, an analyst at the research firm LightCounting, based in Eugene, Ore. “What’s unusual here is we are seeing a telco-led initiative and support for it. I think NTT’s road map is realistic and matches that of industry in general, but it can’t do it alone, and even with all these companies it’s not enough.”“With the advent of AI,” Rubenstein adds, “computing has returned to the center of everything. If the AI boom slows, then the urgency will disappear. But if AI continues as it has done, in five years it will be much closer to that vision.”Takasumi Tanabe, a professor of electronics and electrical engineering at Tokyo’s Keio University, says IOWN is contributing to important R&D in silicon photonics and optical packaging.“At the device level, some aspects are more challenging,” Tanabe says. “A completely ‘all-optical’ system, in which electronics are removed entirely, may not be feasible with the current state of device physics. Electronics will still be necessary for control, modulation, and signal processing. Even so, I expect photonic devices to play an increasingly important role in the most critical parts of future systems, where low-power consumption, high bandwidth, and low latency are required.”“While some elements are ambitious,” Tanabe adds, “the essential ideas behind IOWN are realistic, and the initiative has stimulated valuable advancements in photonic technologies.”  IEEE Spectrum