APAC Data Centre Update: H2 2024

GREATER TOKYO ASIA PACIFIC PRIMARY MARKET

Colo Hyperscale Cloud Telco

SAITAMA

FUCHU

CENTRAL TOKYO

KEY INDICATORS*

INZAI

1,136MW In Operation

30 / 118 Operators / Data Centres

223MW / 1,568MW UC / Planned

9% Colo Vacancy

SAGAMIHARA

KAWASAKI

ECOSYSTEM DEVELOPMENTS •

MARKET OVERVIEW Tokyo remained one of the most active data centre markets in the Asia Pacific region. Despite challenges related to land and power availability, operational capacity grew by 11% over the past year, reaching 1.14GW. Meanwhile, the development pipeline comprising under-construction and planned projects expanded by 29% to 1.8GW. Driven by sustained leasing activity, the vacancy rate remained low at 9%. While operators actively pursued their expansions, three key trends emerged in H2 2024: Artificial Intelligence (AI), Sustainability, and urban small-scale data centres. In December 2024, the Japanese government released the draft of its Seventh Strategic Energy Plan, reinforcing its commitment to decarbonization. The plan targets a renewable energy share of 40%-50%, nuclear power at 20%, and a reduction in thermal power reliance to 30%-40% by 2040. This initiative will support data centre operators in achieving carbon neutrality. Several operators, including At Tokyo and NTT amongst others, have already intensified their sustainability efforts. During H2 2024, Amazon signed a 9.5MW solar power purchase agreement (PPA). Artificial Intelligence (AI) also emerged as a major focus area in Japan, with several operators and service providers— SoftBank, Microsoft, Kyndryl, Rutilea, and Sakura Internet—announcing plans for AI-focused data centres and services. Sakura Internet further revealed its intention to acquire over 10,000 GPUs annually to meet the surging demand for AI processing. Meanwhile, Japan’s National Institute of Advanced Industrial Science and Technology (AIST) appointed Hewlett Packard Enterprise (HPE) to develop the AI Bridging Cloud Infrastructure (ABCI) 3.0 supercomputer, powered by Nvidia H200 Tensor Core GPUs, to be hosted at AIST’s Tokyo facility. The growing risk of evolving location preferences owing to technology-led disruptions and the government restriction created a need for end-clients to seek flexible urban locations. In response to such growing demand from end-clients, Obayashi Corporation launched MiTASUN to develop multiple small-scale data centres targeting the fringe of Tokyo CBD. Each facility is proposed to have a capacity of 2–10 MW, 300–800 racks, and cost JPY 5–15 billion.

Ares Management has entered into an agreement with GLP Capital Partners (GCP) to acquire its international business including its existing data centres in Tokyo and Osaka. The acquisition excludes GCPs’ assets and operations in Greater China. • Japan’s Cabinet Office has announced the "Watt-Bit Collaboration" plan to integrate power systems with communication infrastructure through public-private investments. The strategy aims to locate data centres near decarbonized power sources while utilizing cost-effective optical fiber for data transmission, reducing reliance on costly power grid expansions given optical fiber installation costs only about 1% of high-voltage transmission lines. • Google plans to commence the construction of its Proa subsea system cable in July 2025. The first segment of the cable connects Shima, Japan, to a branching unit in the Philippine Sea. The second and third will connect Tanguisson in Guam and Tinian in the Marianas. • Kyndryl launched an artificial intelligence (AI) private cloud in collaboration with Dell’s AI Factory, equipped with Nvidia Blackwell GPUs. The AI cloud is called “Kyndryl Vital AI Lab” and is aimed at enabling BFSI companies, manufacturers, retailers, and academics to deploy AI-powered solutions. • Microsoft announced its plans to invest US$2.9B to expand its cloud and AI infrastructure in Japan over a period of two years. A part of the investment will be used to skill three million people in AI and to set up a Research Asia lab in Tokyo. • SoftBank Corp added 4,000 Nvidia H100 GPUs for its Japan-based AI computing platform and plans to further expand to 10,000 GPUs by the end of 2025. While the chips are initially dedicated for its subsidiary—SB Intuitions. Softbank also intends to offer the platform as an "Infrastructure-as-a-Service" offering to companies and research institutions.

CUSHMAN & WAKEFIELD | ASIA PACIFIC DATA CENTRE MARKET OVERVIEW