Stargate’s Oracle agreement brings the physical scale of AI into view

OpenAI announced an agreement with Oracle on 22 July 2025 to develop an additional 4.5 gigawatts of Stargate data-centre capacity in the United States. Combined with its Abilene site, OpenAI said the programme would represent more than five gigawatts under development. The announcement described early workloads already running on part of the Abilene infrastructure.

This followed the January 2025 Stargate announcement, which set out an intention to invest $500 billion over four years. The capital ambition and the later capacity agreement are related programme measures, not separate sums to add together. Planned capacity should also remain distinct from fully commissioned operations.

For Bolt’s infrastructure thesis, our interpretation is that AI growth is increasingly a physical delivery story. Power supply, cooling, construction and systems integration determine when computing becomes available to customers. Software demand creates the opportunity, but practical constraints shape the timing and cost of meeting it.

That context is relevant to CUDO’s focus on designing, building and operating AI infrastructure. The Stargate announcement does not identify CUDO as a participant, supplier or beneficiary. Its significance for Bolt is the broader evidence of commitment to the enabling layer of AI, rather than a direct contribution to CUDO’s financial forecast.

Investors should follow progress from announced projects to commissioned capacity and then to economically useful workloads. An operator’s ability to secure the right resources and deliver reliably can become a differentiator in a growing market. Equally, development delays and financing costs can alter returns. Stargate makes the opportunity concrete while showing why execution, rather than the scale of a headline alone, deserves sustained attention.