By Amara Osei

Seven years after Japan handed each of its four mobile operators 400 MHz of 28 GHz spectrum, the band still carries a sliver of the country's 5G traffic. NTT Docomo, KDDI, SoftBank and Rakuten Mobile have now signed a memorandum of understanding to study how they might squeeze more out of it together.

The Ministry of Internal Affairs and Communications (MIC) assigned the mmWave band back in 2019, through a beauty contest that also parcelled out 3.7 GHz and 4.5 GHz. Real 5G traffic, meanwhile, has flowed over low and mid-band frequencies: 700 MHz, 1.7 GHz, 3.4–3.5 GHz, 3.7 GHz, 4.0 GHz and 4.5 GHz.

What does the MoU actually cover?

On paper, plenty. In substance, the four operators promise a joint study of "technologies and approaches related to 5G millimeter-wave base stations, repeaters, and other related assets."

"While each operator actively deploys its own 5G millimeter-wave equipment, the companies will pursue concrete studies to effectively utilize deployed facilities," the announcement reads. The careful wording about deploying their own kit presumably keeps antitrust questions at bay — though the ministry may quietly welcome the whole exercise.

Nobody has spelled out what cooperation would look like. Options on the table include a neutral host model, sharing passive infrastructure, or colocating base stations.

Is coverage the problem, or demand?

Not coverage, says Phil Marshall, chief research officer at Tolaga Research. "My understanding is that Japanese operators have generally fulfilled coverage requirements for the 28 GHz band; that's not the problem," he told us. "The real issue is a lack of demand among users, which is due to several factors, including a lack of compatible devices."

The MIC reached a similar conclusion. In 2023 it fretted that 28 GHz networks carried only a tiny share of total 5G traffic, while conceding that the band's limited propagation doesn't help. It then set up the 5G Business Design Working Group, stocked with industry and academia, to stir up usage.

"I think the operators will look at pooling resources to extend their collective footprint for 28 GHz as a means to incubate demand," Marshall added.

Does shared hardware already work?

It does, in at least one lab and field trial. Docomo and KDDI are testing a shared repeater built by Japan's Kyocera. The device integrates filters and amplifier circuits from both operators into one box that previously held a single operator's equipment.

It relays 28 GHz signals from both KDDI's and Docomo's base stations simultaneously, which beats running two separate repeaters for coverage extension. Think of it as two operators splitting one window instead of each cutting their own.

The economics are the real driver. "The idea is to make 28 GHz more available at a lower cost than is possible through parallel deployments by all four operators," Marshall says. "Operators have to assign capital resources where there is a financial return. If they are unsure about that return, then the deployment must be at minimum cost."

How does Japan compare?

Japan sits in a small club: one of the few countries to assign both 26 GHz and 28 GHz. South Korea moved first on 28 GHz in 2018, the US followed, and Japan came next in 2019. The US has made the band work reasonably well. South Korea's operators dragged their feet so badly that the regulator simply withdrew their licences for missing rollout obligations.

Japan avoided that fate, and doubled down anyway. Earlier in 2026 it ran its first spectrum auction, for 26 GHz, where Docomo outbid KDDI and SoftBank for a national licence at JPY 3.93 billion ($24 million). Rakuten Mobile sat the auction out. Docomo plans to deploy those frequencies where demand concentrates — event venues and major railway stations — and regional licences were reserved for non-MNO newcomers to sharpen competition.

The MoU gives no deadline for the joint study's findings. But with a shared repeater already relaying signals for two of the four operators, the path from paper to hardware is shorter than the band's seven idle years might suggest.