OpenAI’s Stargate initiative is taking a major step forward in Texas, with a newly disclosed 1.2 gigawatt (GW) data center lease tied to an expanded partnership with SoftBank Group’s SB Energy. The deal signals how quickly AI infrastructure is scaling, from “large data centers” to power-plant-scale campuses designed specifically for frontier-model training and inference.
Beyond the line capacity, the expansion brings together capital, operations, and an energy strategy meant to support rapid growth without shifting costs to local ratepayers. With initial facilities already under construction and service expected to begin in 2026, Stargate’s “starting in Texas” roadmap is becoming more concrete, both in square footage and in megawatts.
What the 1.2 GW Texas lease means for Stargate
OpenAI has partnered with SB Energy and signed a “1.2 GW data center lease” for the initial Stargate buildout in Milam County, Texas. In practical terms, 1.2 GW is not a normal enterprise footprint, it’s a utility-scale load, implying a multi-building campus with long-term planning around transmission, interconnection, and phased commissioning.
SB Energy is set to build and operate OpenAI’s 1.2 GW Stargate data center in Milam County. That operator role matters: the company isn’t just providing electrons, it is positioned to deliver the full package of site development, power procurement, and ongoing facility operations that determine how quickly compute can be brought online.
The Milam County move also reinforces Stargate’s Texas-first posture. OpenAI’s official Stargate overview describes a new company targeting up to $500B of investment over four years to build AI infrastructure for OpenAI in the US, explicitly “starting in Texas.” The 1.2 GW lease provides an early anchor for that ambition.
Partners and backers: how Stargate is being assembled
Stargate is framed as a coordinated effort across capital providers and technology partners. OpenAI has named initial equity funders including SoftBank, OpenAI, Oracle, and MGX, establishing that the program is not a single-vendor build but a consortium-style infrastructure push.
On the technology side, OpenAI lists partners including Arm, Microsoft, NVIDIA, Oracle, and OpenAI itself. That lineup reflects the reality of modern AI infrastructure: chips and systems (NVIDIA), CPUs and ecosystem (Arm), cloud and software layers (Microsoft/Oracle), and model/platform demand (OpenAI) must align to make gigawatt-scale deployments usable.
The SB Energy partnership sits inside that broader network as an “infrastructure execution” lane, turning investment intent into land, substations, cooling systems, and operational discipline. OpenAI President Greg Brockman underscored that focus in a statement beginning, “Partnering with SB Energy…,” emphasizing the goal of scaling compute through optimized AI data centers.
Capital stack: $1B invested plus $800M in preferred equity
Scaling to 1.2 GW isn’t just a technical challenge; it is a financing problem. As part of the Stargate buildout, OpenAI and SoftBank Group are investing $1B total into SB Energy, $500M each, helping fund the development capacity needed for the Milam County campus.
SB Energy also secured $800M of Redeemable Preferred Equity from Ares, described as supporting growth. That additional layer of financing matters because megaproject timelines depend on procurement schedules, grid interconnection milestones, and construction sequencing, each requiring capital well before revenue ramps.
Secondary market summaries have highlighted the combined scale of these commitments (for example, line recaps referencing roughly $1.8B when combining the $1B investment with the $800M preferred equity). While such roundups simplify the details, they capture the core point: Stargate’s Texas expansion is being backed with substantial upfront funding.
Construction and timeline: why 2026 is a pivotal year
SB Energy says initial facilities are under construction and are expected to enter service starting in 2026. This aligns with the broader cadence being discussed for Texas-based Stargate capacity, bringing early phases online while later phases scale toward the full gigawatt target.
In parallel, reporting around Stargate’s Texas “flagship” buildout in Abilene describes a first phase of two buildings totaling about 980,000 square feet, supporting 200 MW+ of IT load. Later phases are described as expanding to roughly 1.2 GW total, indicating that multiple Texas sites and phases may ultimately converge on similar end-state power levels.
Timelines for Abilene also point to mid-2026 as a key energization window for a second phase. Taken together with SB Energy’s 2026 service-start guidance, the year looks like a hinge point when Stargate transitions from planning and construction into real, measurable compute availability.
Compute at scale: chips, leases, and the Abilene precedent
Stargate’s Texas narrative is also being shaped by the economics of GPUs and long-term leasing. Coverage tied to the Abilene site has described Oracle as planning a roughly $40B purchase of NVIDIA GB200-class chips intended to power OpenAI’s data center, alongside a 15-year lease framing for the buildout.
Event coverage has gone further, suggesting OpenAI and Oracle could deploy more than 450,000 NVIDIA GB200 GPUs at the Abilene Stargate data center. While such figures can evolve with procurement realities, they illustrate why power numbers like 1.2 GW are showing up: frontier AI at this scale is inseparable from massive electrical demand.
Oracle Chairman Larry Ellison was also quoted describing Stargate Abilene’s 1.2 GW as enough power for “one million four-bedroom homes.” Even allowing for the roughness of household-equivalent comparisons, the quote helps contextualize what a 1.2 GW lease represents: a new class of industrial facility whose primary output is compute.
Energy strategy: grid upgrades, renewables, and “pay our own way”
OpenAI has said it will “pay our own way” so Stargate data centers don’t increase others’ electricity bills. The company has pointed to funding grid upgrades and using flexible loads, an approach meant to address public concerns that AI growth could strain local infrastructure or shift costs to residents.
Trade coverage of the Milam County 1.2 GW site has described it as being supported by new solar generation and battery storage. That aligns with SB Energy’s identity as an energy developer/operator and suggests the campus will be designed with a power supply strategy that is more sophisticated than simply buying from the spot market.
If executed well, this model could become a template: pair large, phased AI campuses with new generation, storage, and grid investments so that incremental demand is matched with incremental capacity. The alternative, racing a of interconnection and leaving communities to absorb volatility, would slow future permitting and weaken the social license to operate.
SB Energy as an OpenAI customer: APIs and internal deployment
The partnership isn’t only about land and power. SB Energy will become a major customer of OpenAI, including using OpenAI APIs and deploying ChatGPT internally. That creates a two-way relationship: SB Energy helps build the compute, and OpenAI helps provide the software layer that can improve SB Energy’s own operations.
For OpenAI, this kind of linkage can tighten feedback loops between infrastructure and product needs. Data center operators sit at the intersection of energy markets, maintenance schedules, and reliability engineering, areas where AI tools can drive better forecasting, faster incident response, and smarter procurement.
For SB Energy, the internal deployment angle hints at an operational playbook where AI supports everything from project development to grid analytics. In that sense, Stargate’s expansion is also a demonstration of how AI infrastructure builders intend to “eat their own cooking” by using advanced models to run complex physical systems.
OpenAI’s 1.2 GW Texas lease in Milam County, executed with SB Energy and backed by significant capital commitments, is a clear marker that Stargate is moving from concept to concrete. With construction underway and initial service expected starting in 2026, the expansion signals that the next wave of AI progress will be shaped as much by infrastructure delivery as by algorithmic breakthroughs.
Just as importantly, the deal highlights the emerging rules of the road: deep partnerships across finance, chips, cloud, and energy; long-term planning for gigawatt-scale power; and an explicit commitment to fund grid upgrades so communities are not left paying for AI’s growth. If Stargate’s Texas buildouts hit their timelines, 2026 may be remembered as the year industrial-scale AI compute truly arrived.
