Around January 9, 2026, OpenAI and SoftBank announced investments of $500 million each in SB Energy, for a combined $1 billion. The same announcements included a 1.2 GW Stargate data-center lease agreement. The whole package is positioned as the newest piece of the Stargate infrastructure buildout.
The numbers deserve a beat of pause. A billion dollars of equity into an energy developer is not a compute purchase; it is a position in the supply chain that makes compute possible. The lease converts that position into guaranteed capacity. Read together, they describe a program planning infrastructure on decade horizons.
Three Components of the Deal
Broken down, the agreement contains:
- Equity: $500 million from OpenAI and $500 million from SoftBank, $1 billion total into SB Energy
- Capacity: a 1.2 GW lease of Stargate data-center capacity
- Strategic position: an extension of the Stargate compute construction program
Why an Energy Company
The bottleneck for AI data centers is moving upstream. GPUs can be bought with capital; gigawatt-scale power requires grid capacity, land, and time. This deal upgrades “power supply” from a procurement relationship to an equity-bound partnership: a model company and a major investor hold shares in the energy producer and lock in future capacity through a lease. For a program like Stargate that must scale reliably over years, it is the most direct way to reduce uncertainty on the electricity side.
There is a sequencing logic as well. Chips can be ordered and delivered on a procurement cycle; power interconnects and generation cannot. Locking electricity years ahead is effectively locking the option to train and serve future models — and failing to lock it is how expansion plans quietly die.
The Logic of Stargate’s Expansion
A 1.2 GW lease says Stargate now plans capacity in gigawatts, not rooms or clusters. SoftBank’s role matters too: OpenAI is not acting alone but alongside a large investor, splitting both the risk and the commitment. When compute demand gets signed at this scale, the entire supply chain behind data centers and power generation has to adjust its tempo.
For competitors, the message is that Stargate is not a single campus but a program that compounds: each lease and investment makes the next phase easier to finance and harder to race against.
The Compute Race Moves Upstream
The 2026 compute race is no longer just about buying GPUs: contracting power, leasing capacity, and investing in energy infrastructure are all part of the battlefield. For the power and data-center industries that is unambiguously good news — AI giants are becoming long-horizon, large-scale, predictable demand. For observers, it is further concrete evidence that AI capital expenditure is reshaping the energy market: when model companies start holding equity in energy producers, the boundary between the two industries has effectively dissolved.
Expect more deals with this shape. When the constraint is power rather than silicon, the natural move for anyone operating at frontier scale is to stop buying capacity and start owning the inputs to it. The companies that treat energy as a core competency — not a facilities line item — will set the pace of the next buildout cycle.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.