OpenAI and Oracle reportedly agreed to a cloud-computing arrangement worth about $300 billion over roughly five years, beginning around 2027. The agreement is tied to approximately 4.5 gigawatts of additional Oracle-backed data-center capacity for OpenAI as part of the broader Stargate infrastructure effort.
That headline number requires an important qualification: the $300 billion figure was reported by The Wall Street Journal and cited by other outlets. It was not itemized in a joint OpenAI–Oracle announcement with a public contract, pricing schedule, or payment timetable.
As an Amazon Associate I earn from qualifying purchases.
The short answer
The reported deal is best understood as a massive, long-term commitment to purchase AI-computing capacity from Oracle—not as $300 billion of immediately delivered cloud services.
Recommended Free Tools
OpenAI publicly confirmed in July 2025 that it was working with Oracle to develop up to 4.5 gigawatts of additional Stargate data-center capacity in the United States. OpenAI said that, combined with its existing Abilene, Texas, site, the arrangement would represent more than 5 gigawatts and support more than 2 million chips.
#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
The reported commercial value and the infrastructure commitment are related, but they are not identical. The dollar amount came from reporting, while the 4.5-GW capacity figure came directly from OpenAI.
What happened, and when?
| Date | Development |
|---|---|
| January 21, 2025 | OpenAI, Oracle, SoftBank and MGX introduced Stargate, describing an initiative intended to invest up to $500 billion over four years in U.S. AI infrastructure. |
| July 22, 2025 | OpenAI announced an agreement with Oracle to develop 4.5 GW of additional Stargate data-center capacity. |
| July 2025 | Oracle disclosed an unnamed cloud customer contract expected to generate approximately $30 billion in annual revenue beginning in fiscal 2028. Reporting later identified OpenAI as the customer. |
| September 10, 2025 | Reporting attributed a roughly $300 billion, five-year computing contract to OpenAI and Oracle. |
| September 23–24, 2025 | OpenAI, Oracle and SoftBank announced five additional Stargate sites, with the announced sites and nearby expansion described as capable of delivering more than 5.5 GW. |
| March 2026 | Reports said OpenAI and Oracle abandoned plans to expand the flagship Abilene campus. Microsoft subsequently moved into the adjacent expansion opportunity. |
| June 1, 2026 | Oracle, OpenAI and partners celebrated construction progress at the Stargate campus in Saline Township, Michigan. |
Relevant announcements include OpenAI’s Oracle partnership announcement, OpenAI’s five-site update and SoftBank’s announcement.
Is the $300 billion figure official?
Not in the sense of a publicly disclosed joint contract. The appropriate wording is that The Wall Street Journal reported the approximate value and term. No public document in the cited material provides the full contract language, minimum purchase obligations, pricing formula, or payment schedule.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Oracle separately disclosed an unnamed customer agreement expected to produce about $30 billion in annual revenue beginning in fiscal 2028. Reporting linked that disclosure to OpenAI, but Oracle’s cited disclosure did not publicly name the customer.
The figures should not be treated as automatically interchangeable. Thirty billion dollars per year for five years would equal about $150 billion, not $300 billion. The difference could reflect different timing, purchase commitments, capacity reservations, accounting treatment or other contractual provisions, but the available sources do not establish the exact reconciliation.
Rank #2
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
What does 4.5 gigawatts mean?
A gigawatt is a measure of power capacity. In this context, 4.5 GW describes the scale of the data centers and their supporting electrical infrastructure—not a standardized amount of computing performance.
That capacity encompasses facilities, power delivery, cooling, networking, storage and accelerator deployment. It does not mean that all of the capacity, or all of the more than 2 million chips referenced by OpenAI, was installed and operating when the partnership was announced.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsChip type, utilization, cooling overhead, networking design and facility efficiency all affect how much useful computing a given power budget can provide. It would therefore be misleading to convert 4.5 GW into an exact GPU count or performance figure.
How the Oracle agreement fits into Stargate
There are three separate layers:
- Stargate: a broad infrastructure initiative announced in January 2025 with an intended investment of up to $500 billion over four years.
- The Oracle–OpenAI relationship: a commercial and infrastructure arrangement involving Oracle Cloud Infrastructure and Oracle-backed data-center capacity for OpenAI.
- Individual projects: campuses and expansions in locations including Abilene, Texas; other Texas sites; New Mexico; and Michigan.
These terms should not be used as synonyms. Stargate’s $500 billion figure was an announced ambition, not proof that $500 billion had already been spent. Similarly, a change to one campus does not by itself prove that the entire Oracle relationship ended.
OpenAI later announced additional Stargate sites through its September 2025 update. The projects were part of a wider buildout rather than a single facility or a single payment.
Rank #3
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
What OpenAI is buying
This is not simply a conventional self-serve cloud account. The reported arrangement concerns large-scale capacity for AI workloads, potentially including:
- High-density accelerator clusters for training frontier models.
- Inference capacity to serve ChatGPT and other products.
- Specialized data-center buildings and electrical systems.
- High-speed networking and storage.
- Cooling, operations and related infrastructure services.
OpenAI needed more capacity as both model training and inference demand grew. A large, dedicated arrangement can provide greater visibility into future compute availability than buying small amounts of on-demand capacity, although it also creates significant long-term obligations.
Why Oracle matters to OpenAI
The deal helps OpenAI pursue a more diversified infrastructure strategy. Frontier-model training requires enormous accelerator clusters, while product usage creates a continuing need for inference capacity. Relying on multiple providers can improve access to different regions, hardware configurations and power supplies.
It can also reduce dependence on a single infrastructure partner and give OpenAI more negotiating leverage. But diversification is not the same as replacement. OpenAI’s July announcement said Microsoft would continue providing cloud services, including through Stargate.
Later reporting also described OpenAI relationships involving other providers, including Google Cloud and AWS. The broader picture is an increasingly multi-cloud infrastructure portfolio, not a clean switch from Microsoft Azure to Oracle.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #4
- 48GB AI graphics accelerator
What Oracle contributes
Oracle’s role extends beyond reselling generic virtual machines. Oracle Cloud Infrastructure provides the hosting environment, while Oracle and development partners help arrange or build large AI data centers designed for high-density accelerator deployments.
Oracle does not manufacture the GPUs used in these systems. Nvidia supplies the relevant accelerators; Oracle provides cloud and infrastructure services and works with construction, financing and hardware partners. The June 2026 Oracle announcement about Saline Township illustrated the physical, multi-party nature of the buildout.
Did the deal reduce Microsoft’s role?
There is no evidence that the Oracle relationship initially removed Microsoft from OpenAI’s infrastructure picture. Microsoft remained a cloud provider, while Oracle supplied additional capacity. OpenAI’s use of other providers reinforces the interpretation that it was expanding capacity options as AI workloads grew.
Microsoft’s later involvement in a specific Texas expansion should also be separated from the overall relationship. The Associated Press reported that Microsoft took over an adjacent expansion opportunity after OpenAI backed away from that plan. That does not mean Microsoft took over the entire Oracle agreement.
What happened at Abilene?
In March 2026, reports said plans to expand the flagship Abilene, Texas, campus had been dropped. The reporting pointed to negotiations involving financing and OpenAI’s changing infrastructure needs.
Best Value
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
This is an important distinction: a campus expansion can be cancelled, delayed, reassigned or redesigned while a broader capacity agreement continues through other sites. Oracle said the broader 4.5-GW OpenAI arrangement remained on track, and construction activity linked to Stargate continued elsewhere, including Michigan.
As of August 18, 2026, the available evidence supports describing the Oracle–OpenAI relationship as active, while acknowledging that not every originally discussed site or capacity tranche is guaranteed to be delivered exactly as first described. See the Bloomberg report on Abilene and Oracle’s response.
Why the arrangement carries risks
For OpenAI
- Large long-term financial commitments may come before demand and revenue are certain.
- Permitting, power interconnection, construction and financing can delay capacity.
- AI hardware can become economically outdated before a facility’s useful life ends.
- Forecasting the balance between training and inference demand several years ahead is difficult.
- Changing model architectures or product demand could leave OpenAI with inflexible obligations.
For Oracle
- Construction and financing costs may arrive before customer revenue.
- A single very large customer creates concentration risk.
- Specialized facilities may be difficult to redeploy if OpenAI’s requirements change.
- Announced capacity is not the same as operational capacity, as the Abilene change demonstrates.
- Oracle’s cloud-growth story depends partly on whether major commitments convert into recognized revenue.
What remains uncertain
Readers should not assume the public reporting answers the following questions:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- The exact contract language and termination provisions.
- Minimum purchase obligations and pricing.
- The payment schedule and accounting treatment.
- Which facilities count toward the 4.5-GW commitment.
- How much capacity is operational at any given time.
- Whether the full reported $300 billion value will ultimately be realized.
- Whether future site changes will alter the delivery schedule.
Those unknowns matter to investors and business customers. A headline contract value can indicate strategic importance, but it is not the same as booked revenue, cash already paid, completed construction or guaranteed profit.
What this means for ordinary cloud buyers
The OpenAI arrangement should not be used as a retail price benchmark. Enterprise AI capacity at this scale is negotiated and depends on hardware generation, region, reservation term, networking, power availability and support requirements.
Organizations evaluating providers should compare:
- Actual GPU availability and quota.
- Reserved versus on-demand capacity.
- Hardware and software compatibility.
- Managed model services.
- Data residency and compliance.
- Networking, storage and egress costs.
- Contract minimums and service-level commitments.
- Total cost of ownership, including staffing and migration.
Potential options include Oracle Cloud Infrastructure, Microsoft Azure AI services, AWS AI services, Google Cloud AI infrastructure, specialized providers such as CoreWeave, and Nvidia DGX Cloud. None should be presumed to offer the same terms as OpenAI’s reported arrangement.
Bottom line
OpenAI’s reported Oracle deal is a roughly $300 billion, five-year cloud-capacity commitment associated with about 4.5 GW of additional AI data-center infrastructure. Its importance lies in the scale of the compute buildout and in OpenAI’s effort to diversify beyond a primarily Microsoft-centered infrastructure relationship.
But the figure remains a reported estimate rather than a fully public contract disclosure. Stargate’s up-to-$500 billion ambition is broader than the Oracle agreement, and the reported cancellation of an Abilene expansion does not establish that the entire relationship collapsed. The clearest current description is a large, evolving infrastructure partnership whose final cost, timetable and delivered capacity remain partly uncertain.
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.
