(Tokyo) — SoftBank Group is expanding into the power business as the rapid growth of artificial intelligence creates an unprecedented demand for electricity. Under Chairman and CEO Masayoshi Son, the Japanese conglomerate is building positions across the AI value chain, from semiconductors and computing infrastructure to data centers and the power systems required to keep them running.
The strategic shift is becoming increasingly visible in the United States. Through SB Energy and other entities within the SoftBank ecosystem, the group is pursuing gigawatt-scale power generation and AI data center projects designed to support the next phase of computing growth.
One of its largest projects is located in Pike County, Ohio, at the PORTS-Pike Technology Campus. SoftBank, Energy Global and AEP Ohio, together with U.S. government agencies, announced a public-private partnership in March 2026 to develop approximately 10 GW of power generation capacity and 10 GW of AI data center capacity at the site.
The project illustrates a fundamental change in how SoftBank views data centers. They are no longer simply buildings that consume electricity from an existing utility network. Instead, SoftBank is pursuing an integrated model in which power generation and computing infrastructure are developed together.
The scale is significant. SoftBank says the Ohio project includes a planned 9.2 GW-scale power generation development, while the broader project is designed around 10 GW-scale power and AI data center capacity. The power component is part of a wider U.S.-Japan strategic investment framework and is being developed with American Electric Power’s AEP Ohio.
The strategy reflects a new problem created by the AI boom. Generative AI, AI agents and high-performance computing require increasingly large data centers, while the electricity needed to operate those facilities must be available continuously and at massive scale.
SoftBank is attempting to solve both problems at the same time. Rather than simply purchasing electricity from utilities, the company is becoming involved in developing the energy and infrastructure required to support large-scale AI computing.
The Ohio development also sits within a broader technology ecosystem. SoftBank’s AI infrastructure strategy is closely connected with companies such as OpenAI and NVIDIA, linking capital, data centers, computing technology and power infrastructure into a single industrial chain.
That creates a business model extending well beyond SoftBank’s traditional identity as a technology investor. The company is increasingly positioning itself as an infrastructure developer capable of connecting the physical resources required by AI with the companies developing and operating AI systems.
Texas is the next major piece of the strategy.
In January 2026, SoftBank Group’s Energy Global entered into a strategic partnership with OpenAI to advance the development of a 1.2 GW-scale AI data center project in Milam County, Texas. The facility is being developed with powered infrastructure designed to accelerate construction and provide the computing capacity required for large-scale AI operations.
The Texas project is important because it demonstrates how SoftBank is approaching energy and computing as a combined infrastructure business. Instead of developing a data center first and solving its power requirements later, the project incorporates power and network infrastructure into the development process from the beginning.
SoftBank’s expansion is not limited to the United States. In May 2026, the group announced a commitment to develop and operate 5 GW of AI data center capacity in France, representing an investment of up to €75 billion. The first phase involves an initial investment of €45 billion to develop 3.1 GW of AI data center capacity in the Hauts-de-France region.
The French project marks SoftBank’s largest AI infrastructure commitment in Europe. The company plans to develop sites in Dunkirk, Bosquel and Bouchain, while working with Energy Global and other strategic partners to build additional capacity across France.
At Bosquel, SoftBank Group and French data center company Sesterce have also formed a majority SoftBank-owned joint venture to develop and operate a 1 GW AI data center campus. The project forms part of SoftBank’s broader 5 GW French commitment.
The French expansion shows that SoftBank’s infrastructure strategy is becoming global. The company is betting that the demand for computing capacity and reliable electricity will extend far beyond the United States as governments and businesses race to build domestic AI capabilities.
For Masayoshi Son, the logic is straightforward. Advanced AI requires three critical layers: semiconductors, computing infrastructure and electricity. Controlling or participating in all three gives SoftBank a broader position in the economic system that will support the next generation of artificial intelligence.
SoftBank already has a major semiconductor asset through Arm, the British chip-design company it acquired in 2016. The group has also made OpenAI a central part of its AI strategy while expanding its exposure to data centers, cloud infrastructure and robotics.
Electricity is now becoming the next strategic layer.
The shift is transforming SoftBank itself. For years, the company was primarily known as a technology investor that made large bets on companies with high growth potential. Its emerging strategy resembles a combination of investment company, technology platform and infrastructure developer, with the aim of capturing value from the physical systems that allow AI to scale.
That transition also introduces a different set of risks. Power plants, transmission infrastructure and data centers require enormous amounts of capital, complex permitting, construction expertise and long development cycles that are fundamentally different from investing in publicly traded technology companies.
Financing will therefore become one of the most important parts of SoftBank’s strategy. The company has acknowledged that its AI infrastructure projects require substantial funding and said it intends to use external financing to limit the burden on its own capital while continuing to develop the infrastructure needed for the expansion of AI.
SoftBank is already building mechanisms around this infrastructure. In July 2026, SoftBank Group and SoftBank Corp. established SB Neo in the United States to operate a neocloud business using the group’s 10-gigawatt-scale energy and AI infrastructure. The company plans to launch services for U.S. enterprises in the fiscal year ending March 2028 and eventually expand AI infrastructure deployment to a 10 GW scale.
This creates a potentially powerful chain of businesses: power generation, data centers, computing, cloud services and AI applications. Instead of capturing value from only one part of the technology ecosystem, SoftBank is attempting to build exposure across multiple layers.
The implications extend beyond SoftBank. The emergence of AI is creating a new relationship between the technology and energy industries, as hyperscale data centers increasingly require power capacity comparable to that of major industrial facilities or even cities.
Ohio provides the clearest example of this transformation. A single development combines power generation, transmission infrastructure, AI data centers and high-performance computing, creating an industrial ecosystem in which energy is no longer merely an operating expense but a strategic component of the technology platform.
If SoftBank succeeds, the group will occupy a very different position from that of a conventional technology investor. It could potentially capture value not only from the growth of AI companies and semiconductor assets, but also from the physical infrastructure required to operate AI systems at global scale.
The next SoftBank battle, therefore, is not simply about which company develops the most powerful AI model. It is increasingly about who can secure the chips, land, data centers, electricity and computing capacity required to run the next generation of artificial intelligence.
