According to Bloomberg, Oracle is engaged in high-level negotiations with the media giant Meta Platforms on a more than $20 billion annual cloud computing agreement. This potential trade will provide Meta with a big computing resource to train and deploy its next generation of artificial intelligence models to further strengthen its infrastructure layout in the AI competition.
According to informed sources, the negotiations are at an advanced stage, with the agreement likely to be of three to five years ‘ duration, with a total value of approximately $20 billion, and mainly covering the supply of GPU and CPU resources in Oracle Cloud Industries (OCI). Meta plans to expand its data centre capacity to support the training of Llama 3.5 models – These models require tens of thousands of NVIDIA H100/H200 GPUs, well above the current capacity of Meta’s own data centre. Oracle ‘ s OCI platform, known for its high performance storage and low delay, has become the hotbed of AI loads, especially in mixed cloud scenarios. Meta will receive an exclusive private virtual cloud (VPC) subdivision, giving priority to Oracle ‘ s fourth generation AMD EPYC processor and the NVIDIA GPU cluster, with a projected peak of millions of FLOPS (floating point/sec). The agreement included a fee-to-pay model, combined with Meta ‘ s advance fee discount, with an average unit price lower than 20 per cent of the AI instance rate for AWS or Azure. Oracle will provide customized API interfaces to ensure a seamless interface with the PyTorch framework in Meta, accelerating the process from training to deployment.
In 2025, the company projected AI-related expenditures at over 40 per cent of the total Capex for the global deployment of the Llama Open Source Model and the upgrade of Meta AI ‘ s assistant. Earlier, Meta had entered into multi-billion-dollar agreements with AWS and Microsoft Azure, but the GPU shortfall for these partners (the global H100 supply meets only 30 per cent of the demand) led to a shift to Oracle. As an important complement to Meta’s 2020 capital expenditure budget ($66 billion to $72 billion), this underscores the centrality of cloud services in the AI era — from OpenAI’s Microsoft Dependence to today’s Oracle collaboration, where Meta is building a “twilight” ecology to spread risk.
The Oracle stock price boom of 4 per cent, with an increase of over 80 per cent during the year, was seen by the market as another milestone in its AI cloud operations. This negotiation, if concluded, would be followed by a multi-billion-dollar agreement with OpenAI, followed by a heavy-pound AI order to drive the global cloud market towards a new pattern of “calculating for kings”.
