Free tools Windows power users keep installed
One-click scans. No signup required.
Yes, the deal is real—but the headline needs an important qualification. Oracle disclosed a cloud-services agreement expected to generate approximately $30 billion in annual revenue, and later disclosures identified OpenAI as the customer. The arrangement covers roughly 4.5 gigawatts of planned AI data-center capacity and is described by OpenAI as exceeding $300 billion over about five years.
That does not necessarily mean OpenAI immediately began sending Oracle a fixed $30 billion cash payment every year. The figure is best understood as expected annual revenue once the infrastructure is built, equipped, activated, and fully ramped.
As an Amazon Associate I earn from qualifying purchases.
The short version
- Customer: OpenAI.
- Infrastructure provider: Oracle Cloud Infrastructure, or OCI.
- Headline figure: Approximately $30 billion in expected annual Oracle revenue.
- Physical scale: Up to 4.5 gigawatts of additional capacity.
- Broader commitment: OpenAI described the partnership as exceeding $300 billion over five years.
- Strategic context: The agreement is a major part of Stargate, but it is not the entirety of Stargate.
- Timing: Revenue was expected to begin around 2027 or Oracle fiscal 2028, depending on the accounting reference and deployment schedule.
OpenAI confirmed the Oracle relationship in July 2025 after Oracle had first disclosed the customer anonymously. TechCrunch reconstructed the announcement and confirmation timeline, while SoftBank’s investor materials referred directly to a $30 billion annual Oracle–OpenAI contract.
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1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesHow the deal became public
On June 30, 2025, Oracle disclosed an unnamed cloud-services agreement expected to add about $30 billion in annual revenue. The filing did not initially name OpenAI as the customer.
OpenAI later confirmed its relationship with Oracle, and SoftBank’s investor presentation referred specifically to the Oracle–OpenAI contract. The evidence therefore comes from a sequence of disclosures rather than from Oracle’s original announcement naming OpenAI outright. Oracle’s filings are available through the SEC filings portal.
What OpenAI is buying
This is not simply a purchase of empty data-center space. The arrangement concerns large-scale cloud infrastructure for artificial-intelligence workloads, including:
- NVIDIA GPU systems for model training and inference;
- high-speed networking and storage;
- power and cooling systems;
- AI-optimized data-center capacity;
- Oracle-operated cloud services; and
- the deployment environment needed to run OpenAI models.
OpenAI said its Abilene, Texas, Stargate facility was operating on OCI. It also said Oracle had begun delivering NVIDIA GB200 racks and that early training and inference workloads had started at the site. The physical facility and the cloud infrastructure do not necessarily have the same owner: The Register reported that Crusoe built and owned the Abilene facility while Oracle furnished and operated the cloud infrastructure.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What 4.5 gigawatts means
The 4.5-gigawatt figure refers to power capacity associated with the planned data-center infrastructure. It is not a GPU count, a measure of model performance, or proof that all of the capacity was online when the agreement was announced.
For scale, TechCrunch cited comparisons to roughly two Hoover Dams and enough electricity for approximately four million homes. Those comparisons are illustrations, not exact engineering equivalents. Power capacity alone does not reveal:
- how many GPUs will ultimately be installed;
- which GPU generations will be used;
- how much capacity is reserved versus consumed;
- the utilization rate;
- the price per GPU-hour; or
- how much electricity the facilities will draw at any particular time.
How the Oracle agreement fits into Stargate
Stargate is OpenAI’s broader infrastructure initiative, initially framed as a $500 billion, 10-gigawatt U.S. investment commitment. The Oracle arrangement is a major component of that initiative, involving up to 4.5 gigawatts of additional capacity.
It should not be treated as synonymous with Stargate. Different sites can involve different developers, owners, financiers, cloud operators, and hardware suppliers. OpenAI later said that five newly announced sites, together with Abilene and CoreWeave-related projects, represented nearly 7 gigawatts of planned capacity and more than $400 billion in investment over three years. Those figures describe a broader portfolio than the Oracle contract alone.
SoftBank has described its role as central to the wider Stargate framework, but that does not mean SoftBank is the payer or contracting party for the entire Oracle–OpenAI commercial arrangement.
What “$30 billion a year” actually means
The most important distinction is between expected annual revenue and an immediate, fixed annual payment.
Oracle’s disclosure concerns revenue it expects to generate from the agreement. That revenue would generally arise as capacity is delivered, equipped, made available, and used—not necessarily on the day the contract is announced.
Several different figures are therefore being discussed:
| Term | Meaning |
|---|---|
| Contract value | The total economic value described for the multiyear arrangement. |
| Annualized revenue | A run-rate or expected yearly contribution once the agreement reaches scale. |
| Recognized revenue | Revenue recorded as contractual services and capacity are delivered under accounting rules. |
| Cash payments | The actual payment schedule, which has not been publicly detailed. |
| Capacity commitment | Infrastructure that may be planned or reserved, rather than fully operational and consumed. |
The arithmetic also shows why the public figures cannot be interpreted too literally. A $30 billion annual run-rate multiplied by five years equals $150 billion—not $300 billion. That does not prove either figure is wrong. It indicates that the public sources do not disclose enough detail about ramp timing, additional commitments, renewals, pricing, or the precise contract term to reverse-engineer the payment schedule.
For that reason, “approximately $30 billion a year in expected revenue at scale” is more accurate than “OpenAI will pay Oracle $30 billion in cash every year starting immediately.”
When will the revenue begin?
The milestones should be kept separate:
- Contract announcement.
- Site selection and construction.
- Power and network interconnection.
- GPU delivery.
- System commissioning.
- Workload activation.
- Revenue recognition.
Oracle’s contribution was reported as beginning around 2027, while other coverage referred to Oracle fiscal 2028. Those statements are not necessarily contradictory: Oracle’s fiscal year does not align exactly with the calendar year. The precise timing also depends on when facilities and services become operational.
OpenAI later said that the Abilene site had begun early workloads, but early operations are not the same as full deployment of the planned 4.5 gigawatts or recognition of the eventual annual revenue run-rate. Oracle subsequently said the broader $300 billion arrangement remained on schedule; that is Oracle’s position and should not be confused with independent proof that every planned milestone had been completed.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Why OpenAI needs this much infrastructure
Frontier-model training requires enormous clusters of specialized accelerators. Inference—the process of generating answers for users and businesses—can require even more sustained capacity as adoption grows.
OpenAI has historically depended heavily on Microsoft Azure but has expanded its infrastructure relationships through Oracle, CoreWeave, and other providers. A multicloud approach can:
- reduce dependence on a single provider;
- increase the amount of available compute;
- give OpenAI more control over hardware configurations and deployment schedules; and
- allow capacity to be built in multiple locations.
The Oracle agreement therefore signals expanding compute requirements, not necessarily a complete break with Microsoft. Microsoft remains relevant to OpenAI’s commercial and technology relationship, while Oracle adds another major source of infrastructure.
Why the deal matters to Oracle
The eventual $30 billion annual figure would be extraordinary relative to Oracle’s previous cloud business. TechCrunch compared it with Oracle’s fiscal 2025 cloud-services revenue of approximately $24.5 billion across all customers. The comparison is not perfectly like-for-like, but it demonstrates the potential scale: one customer’s eventual contract could exceed Oracle’s prior annual cloud-services revenue base.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →For Oracle, the potential benefits include:
- rapid expansion of OCI;
- a stronger position in the AI-cloud market;
- higher utilization of data-center, networking, and GPU infrastructure;
- greater visibility with other AI companies; and
- large recurring revenue if the capacity is delivered and used.
But the contract does not guarantee equivalent profit. Oracle must pay for or arrange facilities, GPUs, electricity, networking, operations, leases, financing, and depreciation. The revenue may be large while the capital requirements are also enormous.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The main execution and financial risks
Construction and power
Announcing capacity is not the same as completing a data center. Land, permitting, buildings, cooling systems, transmission connections, and network links can all delay the delivery of usable compute. Grid connections and new generation capacity may take years.
Rank #4
GPU supply and technology changes
GPU delivery schedules affect when a facility can begin productive work. Hardware planned today may also be economically different from newer systems available before a site is fully commissioned.
Financing
Oracle may need to spend or borrow substantially before receiving the full expected revenue. The Register highlighted the company’s debt and capital-expenditure requirements in the context of pursuing the arrangement. Those figures are time-specific, but the underlying issue is enduring: infrastructure must be financed before the revenue run-rate arrives.
Customer concentration
A contract of this size would make OpenAI a highly important customer for Oracle. That creates exposure if OpenAI changes its infrastructure plans, negotiates different terms, delays workloads, or cannot support its own compute obligations.
OpenAI’s demand and cash flow
The agreement demonstrates infrastructure ambition, not profitability or financial sustainability. OpenAI must generate enough revenue from subscriptions, APIs, enterprise contracts, and other products to support its long-term compute commitments.
Contract uncertainty
The complete agreement has not been publicly released. Important unknowns include minimum purchase commitments, cancellation rights, pricing escalators, delivery milestones, utilization requirements, and remedies if construction or deployment falls behind.
Utilization
A technically complete data center can still be underutilized. The commercial outcome depends not only on building capacity but also on installing hardware, connecting systems, running workloads, and maintaining sufficient demand.
Recommended Free Tools
What the deal does—and does not—prove
The agreement supports several conclusions:
- OpenAI is pursuing infrastructure on a scale far beyond a conventional software company’s server needs.
- AI-cloud demand is creating very large commitments for data-center operators and GPU suppliers.
- OpenAI is building a multicloud infrastructure strategy rather than relying exclusively on one provider.
- Oracle sees AI infrastructure as a major growth opportunity for OCI.
It does not prove that OpenAI is already paying $30 billion per year, that all 4.5 gigawatts are operational, that Oracle will earn $30 billion in profit, or that OpenAI is abandoning Microsoft.
What to watch next
The meaningful indicators are physical and financial milestones rather than headline contract totals:
- completion of announced data-center sites;
- availability of power and network interconnections;
- delivery and installation of NVIDIA systems;
- the amount of capacity actually running OpenAI workloads;
- Oracle’s reported cloud revenue and capital spending;
- changes to site plans, capacity targets, or delivery schedules; and
- evidence about the contract’s minimum commitments and payment structure.
Bottom line
OpenAI really is the customer behind Oracle’s enormous cloud-infrastructure agreement. But “$30 billion a year” is a simplified headline for Oracle’s expected annual revenue at scale, not proof of an immediate $30 billion yearly cash transfer. The arrangement is tied to up to 4.5 gigawatts of planned capacity and a broader commitment described as exceeding $300 billion over five years. Its ultimate value depends on construction, power, GPUs, financing, workload demand, and the terms of a contract that remain only partly public.
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.
