Oracle and OpenAI expand Stargate deal with 4.5 GW of US data center capacity

Oracle Corporation (NYSE: ORCL) and OpenAI have expanded their landmark Stargate infrastructure initiative, securing an estimated 4.5 gigawatts of new U.S.-based data center capacity in a deal that could generate approximately US$30 billion in annual revenue starting fiscal 2028. The expanded agreement significantly strengthens Oracle’s position in the generative AI infrastructure race and reinforces OpenAI’s commitment to diversifying its high-performance compute partnerships beyond Microsoft Azure.

The new capacity buildout spans multiple U.S. states, including a major upgrade to Oracle’s existing data center in Abilene, Texas, which is being expanded from 1.2 GW to nearly 2 GW. Additional hyperscale campuses are being considered in Michigan, Wisconsin, Wyoming, New Mexico, Georgia, Ohio, and Pennsylvania. Oracle expects to deploy US$25 billion in capital expenditures over the next fiscal year, in addition to its US$7 billion equity commitment to Stargate.

Launched in January 2025 with backing from SoftBank Group, OpenAI, MGX, and Oracle, the Stargate initiative is projected to channel as much as US$500 billion into U.S.-based artificial intelligence infrastructure by the end of the decade.

Why is Oracle committing 4.5 gigawatts of compute capacity to OpenAI and how does it impact hyperscale AI infrastructure?

The scale of this expansion reflects OpenAI’s growing need for low-latency, high-throughput compute infrastructure to support next-generation generative AI models, including the systems behind ChatGPT and its enterprise platforms. The 4.5 GW figure represents enough capacity to power approximately 2.3 million high-end GPUs, based on Oracle and industry estimates, and is equivalent to the output of roughly four nuclear power plants.

Oracle is aggressively positioning itself as a hyperscale cloud infrastructure provider specializing in AI workloads. In fiscal year 2025, Oracle Cloud Infrastructure (OCI) generated US$10.3 billion in revenue. With the new OpenAI contract expected to begin revenue realization by FY28, analysts suggest this single relationship could nearly triple Oracle’s current cloud topline over the following years.

Institutional investors have responded positively, driving Oracle’s stock up by over 36 percent year-to-date, with shares hitting record highs in the weeks following the announcement. Analysts have highlighted the deal as a validation of Oracle’s strategic pivot toward AI-native cloud infrastructure, particularly in an environment where compute supply is the gating factor for AI innovation.

How does this expansion compare to the original Stargate plan and Oracle’s broader cloud transformation?

The original Stargate framework envisioned a four-year, multibillion-dollar investment program to build sovereign U.S. AI infrastructure. As announced by President Donald Trump at the White House in January 2025, the program includes capital commitments of US$19 billion each from OpenAI and SoftBank, with Oracle and MGX pledging US$7 billion apiece. The initiative is intended to ensure U.S. competitiveness in artificial intelligence and to stimulate economic development across high-tech industrial zones.

The Abilene, Texas data center campus remains the centerpiece of Stargate, with plans to deploy up to 400,000 Nvidia GB200 GPUs. Oracle is collaborating with Crusoe Energy Systems for infrastructure integration and on-site energy optimization. With an estimated US$40 billion GPU procurement pipeline already underway, the Abilene site alone could support some of the most compute-intensive model training workloads globally.

With the addition of 4.5 GW in leased capacity, Oracle is now actively pursuing a buildout that rivals Amazon Web Services and Microsoft Azure in scale. TD Cowen analysts estimate that Oracle’s total AI-focused data center ambitions could reach 5 GW by the end of 2026, with cumulative hardware and facility spending exceeding US$150 billion over time.

What are institutional investors and analysts saying about the deal’s revenue potential and execution risk?

While institutional sentiment is largely optimistic, some caution is emerging around the deal’s long-term revenue realization timeline and associated capital risks. Several market observers have noted that while the contract is expected to deliver US$30 billion annually by 2028, Oracle will need to front-load capital expenditures through 2026–27, putting potential pressure on free cash flow and short-term margins.

Analysts view Oracle’s unique positioning as a competitive edge, particularly its ability to bundle compute capacity with enterprise-class software and database services for AI customers. However, there is growing scrutiny around execution, especially given the geographic and regulatory complexity of launching multiple hyperscale sites within a compressed time frame.

Investors are closely watching Oracle’s reported backlog of performance obligations, which reached new highs in recent quarters. Analysts expect continued double-digit growth in Oracle Cloud Infrastructure, with some projecting annual cloud revenue run rates in the US$30–35 billion range by FY2029 if execution proceeds on schedule.

What risks are associated with Oracle’s aggressive buildout strategy and reliance on OpenAI?

Despite strong investor enthusiasm, there are meaningful risks associated with the Stargate expansion. Credit rating agencies have flagged Oracle’s front-loaded capex cycle as a potential drag on cash generation, even as the company transitions from an asset-light to infrastructure-heavy model. Execution risks include construction delays, equipment procurement bottlenecks, GPU availability, and rising input costs—particularly for power and real estate.

There are also demand-side risks. Oracle’s financial exposure to OpenAI is significant. Reports suggest that even if OpenAI reduces its compute requirements, Oracle could still be obligated to pay up to US$1 billion in pre-committed payments to hardware vendors and energy partners. Environmental approvals and permitting delays at new sites could also impact deployment schedules and operational costs.

State governments have submitted over 250 site proposals to host Stargate facilities, but concerns are growing around the energy consumption, water usage, and emissions footprint of large AI campuses. Environmental groups and state regulators are beginning to ask tougher questions, particularly around energy equity and sustainability metrics.

How does this reshape competition among U.S. cloud infrastructure providers and AI compute vendors?

The Oracle–OpenAI alliance is altering the landscape of U.S. AI infrastructure, pushing Oracle into direct competition with Microsoft, Amazon, and Google—not just as a software partner but as a foundational compute supplier. By securing 4.5 GW in capacity commitments from OpenAI, Oracle is establishing itself as a key enabler of next-gen AI ecosystems.

The deal also offers OpenAI greater supplier diversification, reducing dependency on Microsoft Azure and allowing it to stage model development across multiple compute partners. As OpenAI prepares future rollouts of GPT-5-class systems and enterprise copilots, access to scalable and dedicated infrastructure becomes an existential requirement.

Looking forward, Oracle could see further upside by signing similar compute-as-a-service contracts with other AI-native firms and large enterprises. Analysts believe that once infrastructure is in place, Oracle will gain pricing power and operational leverage, particularly if demand outpaces capacity over the next five years.

What is the forward outlook for Oracle and OpenAI following this infrastructure milestone?

The next key milestones will include Oracle’s site selection and groundbreaking activities in at least four new U.S. states, likely beginning in late 2025. The company aims to bring the first tranche of its new 4.5 GW capacity online between late 2026 and 2027, ahead of OpenAI’s commercial usage ramp in FY2028.

OpenAI is expected to consume the majority of the new compute capacity for model training, inference, and deployment across its enterprise and consumer product stack. Analysts are watching closely for signs of broader enterprise GPU demand and the potential emergence of tier-two compute customers who could lease capacity through Stargate.

For Oracle, delivering on time, within budget, and at full reliability will be critical to maintaining institutional confidence. If successful, this deal could reposition the enterprise software giant as one of the largest AI infrastructure providers globally, unlocking long-term cloud revenues, pricing control, and strategic leverage in the generative AI ecosystem.