How Oracle & SoftBank are Expanding AI Data Centre Capacity

In the race to scale data centre capacity to power surging global AI demand, OpenAI, Oracle and SoftBank have unveiled five new US facilities under the Stargate AI infrastructure initiative.
Stretching from Texas to Norway, Stargate is accelerating worldwide deployment, pushing cumulative investment past US$400bn and driving planned capacity toward 7GW.
This rapid expansion keeps the consortium on track ā and in fact ahead of schedule ā to hit its goal of a US$500bn, 10GW buildout by the end of 2025.
First revealed at the White House in January, Stargate has quickly emerged as one of the technology sectorās most ambitious long-term infrastructure plays.
The five newly announced locations represent the programmeās opening phase in the US, with further domestic sites expected as the rollout continues.
Sam Altman, CEO of OpenAI, says that āAI can only fulfil its promise if we build the compute to power it".
He adds: āThat compute is the key to ensuring everyone can benefit from AI and to unlocking future breakthroughs.ā
Oracleās role: Delivering multi-state expansion
Oracle, the database and cloud services giant, is set to build three of the five new facilities as part of a July agreement that committed up to 4.5GW of added capacity.
- Shackelford County, Texas
- DoƱa Ana County, New Mexico
- An undisclosed site in the Midwest (to be announced)
- Lordstown, Ohio
- Milam County, Texas
The partnership accounts for more than US$300bn in joint investment over five years.
Oracleās sites include Shackelford County, Texas; DoƱa Ana County, New Mexico and a third, undisclosed Midwest location.
When coupled with a proposed 600MW expansion near the flagship Abilene campus in Texas, the buildout will supply more than 5.5GW of capacity.
The Abilene facility is already live on Oracle Cloud Infrastructure, having deployed its first Nvidia GB200 server racks in June.
These state-of-the-art systems integrate GPUs with processors tailored for machine learning workloads, delivering the parallel compute power essential for advanced AI model training.
Clay Magouyrk, CEO of Oracle, says that āOracleās reliable, scalable and secure AI infrastructure is helping OpenAI rapidly scale its businessā.
He adds that āto meet this enormous demand, we continue to expand OCIās footprint at an unrivalled pace to deliver the most performant and cost-effective AI training and inferencingā.
The Oracle facilities are projected to generate more than 25,000 on-site jobs, alongside tens of thousands of additional roles across the wider economy.
OpenAI has already tapped into the Abilene capacity for early training and inference tasks, advancing its next-generation AI research efforts.
SoftBankās role: Brings innovative data centre design
The remaining two sites will be developed through its partnership with OpenAI, with plans to scale capacity to 1.5GW over the next 18 months.
The Japanese investment group adds deep expertise in telecommunications infrastructure and energy systems to the collaboration.
In Lordstown, Ohio, one facility is already under construction using SoftBankās advanced data centre design, with operations slated to begin next year.
A second site in Milam County, Texas, will be built in partnership with SB Energy ā SoftBankās renewable energy arm ā which will supply powered infrastructure to enable rapid deployment.
Together, these locations are intended to showcase faster build timelines and greater cost efficiency in hyperscale development, while demanding significant electrical infrastructure ā each gigawatt of capacity drawing power on par with roughly 700,000 homes.
Masayoshi Son, Chairman and CEO of SoftBank Group, says that “Stargate is harnessing SoftBank’s innovative data centre design and energy expertise to deliver the scalable compute that powers AI’s future.
“Together with OpenAI, Arm and our Stargate partners, we are paving the way for a new era where AI advances humanity.”
The partnership leverages SoftBank’s ownership of Arm, the semiconductor design company whose processors are increasingly used alongside traditional graphics processing units in AI applications.
“We’re already making historic progress toward that goal through Stargate and moving quickly not just to meet its initial commitment, but to lay the foundation for what comes next,” Sam says.

