Key Takeaways
- OpenAI is significantly expanding its strategic partnership with Oracle, choosing to run key artificial intelligence workloads on Oracle Cloud Infrastructure (OCI). This move signals a deliberate diversification of its cloud providers beyond its primary relationship with Microsoft.
- The immense resource requirements of frontier AI models are driving unprecedented infrastructure projects. For context, Microsoft’s separate ‘Stargate’ initiative with OpenAI is projected to cost upwards of $100 billion and may eventually consume multiple gigawatts of power.
- For Oracle, securing a larger portion of OpenAI’s workload provides a substantial validation of its cloud strategy and bolsters its competitive standing against larger rivals like Amazon Web Services and Google Cloud in the high-stakes AI market.
- The partnership highlights an accelerating trend among leading AI firms towards multi-cloud architectures. This strategy is essential for securing sufficient computational capacity, enhancing resilience, and gaining commercial leverage with providers.
Oracle and OpenAI have deepened their collaboration, with OpenAI selecting Oracle Cloud Infrastructure (OCI) to run a significant portion of its AI model training and inference workloads. This development is more than a simple vendor agreement; it is a telling indicator of the tectonic shifts occurring in the world of AI infrastructure, where computational demand is beginning to outstrip the readily available supply of power and specialised hardware.
While OpenAI’s foundational partnership with Microsoft remains firmly in place, this expansion with Oracle underscores a critical reality: no single cloud provider may be sufficient to quench the thirst of a leading AI laboratory. The move points to a maturing, multi-cloud strategy at OpenAI, designed to secure vast resources and mitigate concentration risk as it pushes the boundaries of model scale and capability.
The Staggering Cost of Intelligence
To appreciate the context of this partnership, one must grasp the sheer scale of the resources now required. The industry is witnessing a capital expenditure cycle of historic proportions. Consider Microsoft’s own AI supercomputer project, reportedly codenamed ‘Stargate’, which it is developing for OpenAI. Leaked plans suggest a multi-phase build-out that could ultimately cost over $100 billion and require an astonishing five gigawatts of power by 2028. [1] A single gigawatt is roughly the output of a large nuclear power plant and can power hundreds of thousands of homes.
These figures illustrate that the primary constraints on AI development are no longer just algorithms or data, but increasingly raw power and physical infrastructure. Building data centres on this scale involves immense logistical, engineering, and political challenges, from securing land and permits to ensuring the stability of local and national energy grids. The partnership with Oracle provides OpenAI with another vital avenue to secure the necessary capacity to fuel its ambitions.
Oracle’s Strategic Coup
For Oracle, this is a landmark achievement. Long considered a legacy database giant, the company has invested heavily to refashion itself as a credible hyperscale cloud provider. Its success has been particularly notable in high-performance computing, where its RDMA Cluster Networking is engineered to connect tens of thousands of GPUs into a single, cohesive supercomputer, a configuration highly sought after for training large models efficiently.
Securing a greater share of the workload from the world’s most prominent AI company lends significant credibility to Oracle’s strategy and technical claims. It allows Oracle to directly challenge the narrative that the AI cloud market is an unassailable duopoly of Microsoft Azure and Amazon Web Services. The financial impact is already visible in the company’s performance metrics, with Remaining Performance Obligations (RPOs), a measure of future contracted revenue, showing robust growth driven by large cloud deals.
| Fiscal Period | Total RPO | Year-over-Year Growth | Commentary |
|---|---|---|---|
| Q4 FY2024 | $98 billion | 44% | Driven by record-level sales contracts for Oracle Cloud Infrastructure. [2] |
| Q3 FY2024 | $80 billion | 29% | Reflected strong demand for AI workload capacity. [3] |
A Multi-Cloud Imperative
The notion of an AI company being “all in” on a single cloud is rapidly becoming antiquated. The fierce competition for GPU supply and data centre space means that diversification is no longer an option but a necessity. By leveraging both Microsoft and Oracle, OpenAI can play providers off one another, access different hardware configurations, and ensure it is not left wanting for capacity during critical training runs. This multi-cloud approach provides resilience against outages, supply chain disruptions, and the negotiating leverage of any single partner.
This trend is likely to cascade across the industry, with other well-funded AI startups and research labs seeking to establish relationships with multiple cloud providers to secure their computational futures.
The Coming Energy Reckoning
The Oracle and OpenAI partnership is a clear signal of the current infrastructure landscape. However, it also offers a glimpse into the future. As AI models continue to scale, the direct availability of massive amounts of power will become the most significant differentiating factor for cloud providers. The competition for gigawatts will be as fierce as the competition for customers.
This leads to a speculative but logical hypothesis: the next evolution in AI infrastructure will involve the direct integration of power generation with data centres. We may see cloud providers or AI firms entering into joint ventures to build their own dedicated power sources, such as small modular reactors (SMRs) or next-generation geothermal plants, co-located with their supercomputers. In such a future, the most valuable partner for an AI company might not be the one with the best cloud software, but the one that can guarantee five gigawatts of uninterrupted, sustainable power. The balance of power, in every sense of the word, is shifting.
References
[1] Dastin, J., & Hu, K. (2024, March 29). Microsoft and OpenAI plot $100 billion ‘Stargate’ AI supercomputer, The Information reports. Reuters. Retrieved from https://www.reuters.com/technology/microsoft-openai-plot-100-billion-stargate-ai-supercomputer-information-reports-2024-03-29/
[2] Oracle. (2024, June 11). Oracle Announces Fiscal 2024 Fourth Quarter and Full-Year Financial Results. Oracle Press Releases. Retrieved from https://www.oracle.com/news/announcement/q4-2024-oracle-earnings-announcement-2024-06-11/
[3] Oracle. (2024, March 11). Oracle Announces Fiscal 2024 Third Quarter Financial Results. Oracle Press Releases. Retrieved from https://www.oracle.com/news/announcement/q3-2024-oracle-earnings-announcement-2024-03-11/
@StockSavvyShay. (2024, August 2). [$ORCL & OPENAI STRIKE DEAL TO EXPAND STARGATE AI PROJECTS ACROSS THE U.S.]. Retrieved from https://x.com/StockSavvyShay/status/1881834822093750703
