Bottom line: OpenAI’s 2025 agreement with Oracle added 4.5 gigawatts of planned U.S. data-center capacity to the Stargate infrastructure program. The widely cited $30 billion figure was reported as approximately $30 billion in annual Oracle revenue—not a publicly disclosed total contract value. Microsoft was not cut out: OpenAI said Microsoft would continue providing cloud services.
What OpenAI and Oracle actually announced
On July 22, 2025, OpenAI and Oracle announced plans to develop 4.5 gigawatts of additional U.S. Stargate data-center capacity. Including Stargate’s existing Abilene, Texas campus, OpenAI said the program would have more than 5 GW of capacity under development and could support more than 2 million chips.
As an Amazon Associate I earn from qualifying purchases.
The announcement concerned planned infrastructure—not 4.5 GW of immediately available computing power. Data-center capacity must still be financed, permitted, constructed, connected to the grid, equipped with accelerators, and brought into operation. OpenAI’s announcement did not establish that every announced facility was operational.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesOpenAI described the broader Stargate ambition as $500 billion of U.S. AI-infrastructure investment and 10 GW of capacity over four years. That is a program-wide target, not money already spent or a guaranteed return for Oracle.
#1 Best Overall
What the reported $30 billion means
The headline number needs careful interpretation. Reporting around the arrangement described approximately $30 billion in annual revenue for Oracle, associated with a multiyear cloud-capacity agreement and a ramp beginning in Oracle’s fiscal 2028.
| Figure | What it represents | Status |
|---|---|---|
| $30 billion | Reported annual Oracle revenue | Secondary reporting; not fully disclosed by OpenAI |
| 4.5 GW | Additional planned data-center capacity | Publicly announced by OpenAI |
| More than 5 GW | Stargate capacity under development, including Abilene | Publicly stated by OpenAI |
| $500 billion | Broader Stargate investment ambition | Program target, not demonstrated spending |
It would therefore be inaccurate to say simply that OpenAI signed a disclosed $30 billion cloud contract. The public announcement did not specify a total contract value, guaranteed minimum payments, termination rights, or the exact accounting treatment. The $30 billion figure should be described as a reported annual revenue expectation, not as a confirmed upfront payment or total commitment. Contemporaneous reporting is the basis for the annual-revenue figure.
Why the 4.5 GW matters
For AI infrastructure, power is a foundational constraint. Large model training and inference require dense clusters of accelerators, high-throughput networking, cooling systems, storage, and reliable electricity. More power and more physical sites can give OpenAI room to expand workloads as demand grows.
But a gigawatt figure is not the same as compute performance. It does not reveal the exact accelerator mix, utilization rate, network topology, training throughput, cost per token, or availability date. A useful distinction is:
Rank #2
- Announced: publicly described as planned.
- Under construction: physical development is underway.
- Energized: connected to usable power.
- Operational: equipped and available for production workloads.
The Oracle announcement primarily established the first category, with development expected to progress over time.
Oracle’s role in Stargate
Oracle is contributing infrastructure and cloud capacity rather than a proprietary frontier AI model. Its role includes data-center development, Oracle Cloud Infrastructure capacity, high-density computing environments, networking, power, cooling, and operational support.
The arrangement gives Oracle a major potential anchor customer for OCI and supports its effort to compete for hyperscale AI infrastructure. It may also improve Oracle’s credibility with other model developers and enterprises evaluating large GPU deployments.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →For OpenAI, Oracle adds another route to dedicated capacity. That can reduce dependence on a single provider, improve geographic and operational diversification, and give OpenAI more leverage when negotiating capacity and infrastructure terms.
Rank #3
Microsoft was not replaced
The deal was widely framed as a Microsoft cloud shake-up, but OpenAI’s own announcement explicitly said Microsoft would continue providing cloud services, including through Stargate.
The better description is supplier diversification. Oracle can add capacity while Microsoft remains important for existing workloads, enterprise distribution, software integration, and contractual relationships. A multicloud strategy does not require OpenAI to abandon Azure.
There is no evidence in the cited announcement that Oracle became OpenAI’s exclusive provider or that Azure was eliminated. The agreement suggests a broader infrastructure strategy; it does not prove a clean Microsoft–OpenAI split.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchHow this fits into Stargate
Stargate is a broader AI-infrastructure platform, not one data center. OpenAI has described a network involving Oracle, SoftBank, CoreWeave, the Abilene campus, and additional U.S. projects.
OpenAI later announced five additional Stargate sites involving Oracle and SoftBank after reviewing more than 300 proposals from more than 30 states. That expansion indicated that the Oracle relationship was part of an ongoing site-development program rather than a one-time capacity reservation. OpenAI’s five-site announcement provides the company’s account of that expansion.
Timeline
- January 2025: Stargate’s broader U.S. infrastructure ambition was announced.
- July 22, 2025: OpenAI and Oracle announced 4.5 GW of additional planned capacity.
- Later in 2025: OpenAI, Oracle, and SoftBank announced five additional sites.
- 2026: Construction, financing, power availability, and site plans continued to require site-by-site verification. Announced capacity should not automatically be treated as live capacity.
Who benefits?
OpenAI
- More potential capacity for training and inference.
- Less reliance on a single infrastructure provider.
- Greater flexibility in placing workloads across sites and clouds.
Oracle
- A large prospective anchor customer for OCI.
- Long-term demand supporting data-center investment.
- Greater visibility in the AI-cloud market.
Microsoft
- Continued participation in OpenAI’s cloud ecosystem.
- Ongoing access to OpenAI-related enterprise opportunities.
- A stronger competitive reason to maintain capacity and integration.
SoftBank and other Stargate participants
- Exposure to new data-center and infrastructure projects.
- Potential construction and economic-development benefits.
- Financial exposure if projects are delayed, resized, or underutilized.
The main risks
Execution and grid risk
Building 4.5 GW requires years of construction, transmission, interconnection, equipment delivery, and commissioning. Power availability may be more restrictive than land or announced demand.
Financing and utilization risk
AI campuses require substantial upfront capital. Their economics depend on sustained demand for training and inference. If model architectures, customer demand, or pricing change, facilities may take longer to fill than expected.
Recommended Free Tools
Technology risk
Accelerator generations can advance before a facility is fully equipped. A site designed around one deployment plan may need costly changes to remain competitive.
Best Value
Customer-concentration risk
A major AI customer can support rapid cloud expansion, but Oracle’s economics could become unusually dependent on a small number of large customers. Reported revenue expectations are not the same as realized revenue.
Multicloud complexity
Using several providers can improve resilience and bargaining power, but it can also increase networking, security, software, data-movement, and operations complexity.
Local and environmental impacts
Large campuses can affect electricity demand, water use, land use, transmission planning, construction activity, and nearby communities. Those effects vary by site and cannot be inferred from the headline gigawatt figure alone.
What investors and enterprise buyers should watch
- Operational capacity: Which sites are energized, equipped, and serving production workloads?
- Deployment timing: Are facilities arriving quickly enough to support OpenAI’s model roadmap?
- Oracle disclosures: Do filings or earnings reports clarify backlog, customer commitments, revenue timing, or concentration?
- Contract certainty: Are minimum purchases, termination rights, and revenue-recognition terms disclosed?
- GPU availability: How many usable accelerators are installed, and what generation are they?
- Infrastructure economics: What are the costs of power, networking, storage, cooling, and utilization?
- Microsoft’s role: Does Oracle supplement Azure or materially replace some workloads?
- Local milestones: Do permitting, transmission, water, and construction schedules support the announced plans?
Common mistakes in interpreting the deal
- Calling $30 billion the total contract value without evidence.
- Treating annual reported Oracle revenue as a guaranteed payment from OpenAI.
- Equating 4.5 GW of planned power capacity with 4.5 GW of computing performance.
- Saying OpenAI abandoned Microsoft.
- Assuming every announced site is already operational.
- Confusing Stargate’s $500 billion ambition with money already invested.
- Assuming more infrastructure guarantees better models, lower prices, or cheaper ChatGPT access.
What it means for ordinary cloud customers
This is primarily a hyperscale infrastructure story. Most small businesses and individual developers do not need Stargate-scale capacity. They are more likely to use managed AI APIs or ordinary cloud GPU instances.
Organizations evaluating large AI deployments should compare guaranteed GPU availability, accelerator generation, cluster scaling time, networking and egress costs, storage throughput, security, compliance, orchestration, contract minimums, and the ability to operate across more than one provider. The reported $30 billion figure should not be converted into a retail GPU or per-token price.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
