Initiatives toward Social Implementation in the IOWN 2.0 Era

3.1 Application to the broadcasting industry

The need for more efficient video production and digital transformation (DX) has been growing because competition has been intensifying due to increasing demand for digital media services. To support live broadcasts with a high-realistic video from remote locations such as sports and concert venue, broadcasting stations face issues to possess expensive broadcasting vehicles and a large on-site staff with long working hours. Connecting between a variety of shooting locations and production sites with the large-capacity and low-latency APN enables a video production DX that can produce high-quality content in real time without on-site vehicles and a large staff.

At NTT R&D Forum held in November 2025, remotely distributed graphics processing units (GPUs) were connected via the IOWN APN to demonstrate a virtual production technique as one example of a video production DX. NTT showed high-quality video production that could be achieved with low latency without a permanent installation of high-performance equipment at production sites [6] (Fig. 1).

Fig. 1. Video production DX by using GPUs via the IOWN APN.

3.2 Application to the construction and manufacturing industries

In the construction and manufacturing industries, an urgent need has arisen for DX and other measures in the face of work-style reforms (e.g., overtime cap) and labor shortages. Studies are therefore being conducted to achieve radical improvements in work efficiency through operation of machinery and centralized management from remote locations.

At NTT R&D Forum held in November 2023, NTT demonstrated remote operation of a hydraulic excavator installed in the field in Chiba Prefecture and a tower crane installed in the field in Osaka from a cockpit in NTT Musashino R&D Center in Tokyo [7]. We achieved equivalent operability comparable to on-site operation of those heavy machineries with only several-millisecond latency by using the APN between Tokyo and Osaka, a distance of over 500 km. The IOWN Global Forum organized the documentation for use cases of remote and automated construction management in mountain tunnels and discusses the use of the APN for remote monitoring, remote analysis, remote inspection, and remote maintenance [8].

At NTT R&D Forum held in November 2025, we then held a demonstration on AI-based visual inspection via the APN between a factory and datacenter separated by 300 km and showed that this technology could be used to establish unified quality standards and improve productivity. We also confirmed that remote control of production facilities could be executed within a control period of 20 milliseconds by a cloud-shifted controller. This result is expected to lead to reducing on-site working hours and improving productivity across multiple factories.

3.3 Application to datacenter connection

The securing of land for datacenters in urban areas has increasingly become difficult; therefore, there is an urgent need to secure suburban land and electric power for datacenters. Against this background, movements to maximize power consumption efficiency through flexible use of renewable energy and the expansion of the ICT infrastructure are accelerating on the basis of the concept of watt-bit collaboration, which aims to encourage the effective collaboration between electricity and telecommunications entities.

Efficient computational processing that uses regional renewable energy can be achieved by connecting datacenters spanning different areas, as in connections between datacenters in urban and suburban areas. This will enable advanced distributed processing using high-speed, large-capacity, and low-latency connections with the APN.

4.1 Customer expectations for the APN

In the broadcasting industry, when companies proceed video production DX for live sports broadcast, they need flexible connection of the APN to a variety of stadiums based on the type of sports, such as baseball and soccer. They pursue reducing facility-usage costs with the minimum time needed for pre- and post-event preparation for broadcasting; therefore, the APN also must achieve on-demand connectivity for video production DX. When the video-production studio and equipment are being shared by several companies, the efficient allocation of computing resources, such as GPUs, is also necessary based on the usage conditions of each company.

In the construction industry, the need has been expressed for executing heavy machinery work remotely at different construction sites in the morning and afternoon. Since the location at which the customer would like to use the APN will change in a short period, there is a need for on-demand use of the APN. On-site offices, such as prefabricated houses in many cases, are frequently used as locations to install the network equipment. Because of the limitation of the available space and power capacity, on-site workers expect to use such resources effectively with simple configurations (Fig. 2). With the expansion of remote and automatic construction, AI for danger predictions or the other functions will be available in on-site work by using the connection to computing resources for AI processing.

Fig. 2. Customer expectations for the IOWN APN in the construction industry.

4.2 APN technology development

As we expand and diversify APN-application use cases, there will be a need to develop additional functions that contribute to enhanced customer convenience.

The open converged transponder (OCT) is being studied as a technology for making the devices that use the APN more compact and fewer in number and for lowering power consumption. It is expected that equipment cost, power consumption, and installation space can be reduced though equipment integration by developing optical transceivers with a pluggable-module format and mounting some of the Layer 1 transmission functions in Layer 2/3 equipment such as switches and routers.

Development of an APN controller (APN-C) is also progressing for achieving control, operation, and intelligent functions for APN provision. It will execute multi-vendor device control including the above OCT. The APN-C features functions that enable the sharing of limited wavelength resources and efficient design of routes and wavelengths. The aim is to provide a service that can flexibly change APN connection points by switching optical paths in an on-demand manner of tens of minutes.

We are also researching and developing a wide range of technologies including wavelength-conversion technology to support a variety of connections using wavelengths, automatic optical-path provisioning technology that can contribute to shortening delivery times from the viewpoint of accelerating global expansion, and 1.6-Tbit/s optical transmission technology for achieving further increases in capacity.

7.1 Composable servers

Traditional servers package CPU, GPU, memory, and storage together in a single chassis. Composable servers instead use servers that host CPUs, boxes that host multiple GPUs or storage devices, and boxes with compute express link (CXL) memory modules; these boxes are interconnected via Peripheral Component Interconnect Express (PCIe) or CXL fabric switches. This architecture allows for flexible combinations of CPU, GPU, storage, and CXL memory to match workload requirements. For example, in a conventional server with one CPU and four GPUs, two GPUs might be unused if a user needs only two GPUs; composable servers reduce such waste by enabling more flexible allocation of GPUs and memory to CPUs. For more detail, see the article “Initiatives toward Multiple Vendor-sourced Composable Servers in DCI Technology Development” [9] in this issue.

7.2 PEC-2

A high-performance data network that efficiently connects many servers is important for DCI. AI datacenters require a large inter-server bandwidth as multiple GPU servers collaborate and exchange massive amounts of data, often leading to servers being equipped with multiple network interface cards and a rapid increase in network port counts. The cost and power consumption of networking equipment have thus become significant, increasing demand for higher-capacity and lower-power network switches. A PEC device enables large-capacity, low-power switches. A PEC device can improve the capacity and power efficiency of optical transceivers that execute the electrical-to-optical and optical-to-electrical conversions in network switches. In the Expo demonstration, we used a PEC-2 device compliant with OIF (Optical Internetworking Forum) standards. It delivered a 3.2-Tbit/s per module, eight times the throughput of a 400-Gbit/s transceiver, demonstrating a major leap in per module capacity. This PEC-2 device reduced power per bandwidth to roughly half that of conventional transceivers, making it a key technology for large, low-power network switches required by AI datacenters. While the PEC-2 device used at the Expo 2025 operated at 3.2 Tbit/s, an improved PEC-2 device with 6.4-Tbit/s capacity is currently under development. For more detail, see the article “A High-capacity, Energy-efficient Photonics-electronics Converged Switch with PEC-2” [10] in this issue.

7.3 DCI controller

A key element of IOWN optical computing is the DCI controller, which integrates and controls the components described above. The DCI controller can allocate compute resources while accounting for resource and power conditions across multiple datacenters. It consists of application frameworks and a dynamic hardware resource controller (DHRC). The application frameworks provide well-tuned software templates so applications can efficiently use DCI resources. When an application runs on DCI, the DHRC allocates optimal resources on the basis of current resource utilization and workload characteristics. These components increase resource-utilization efficiency and reduce cost and power consumption.