Oracle’s recent announcement reveals their intention to harness small modular reactors (SMRs) for an AI-driven data centre.
- During Oracle’s Q1 2024 earnings call, plans for three SMRs were unexpectedly disclosed in response to a market query.
- Chairman Lawrence Ellison highlighted the enormous power demands of AI training and inference, necessitating gigawatt-scale facilities.
- The SMRs, although significant, will only partially meet the massive energy requirements of future data centres.
- Oracle’s move underscores the broader trend of AI’s growing role in civil engineering and other sectors.
Oracle has made public its ambitious strategy to incorporate three small modular reactors (SMRs) into its upcoming AI-focused data centre. This revelation came to light during the company’s Q1 2024 earnings call. In response to a question posed by JPMorgan analyst Mark Murphy regarding the transition from AI training to the inferencing phase, Oracle chairman Lawrence Joseph Ellison elaborated on the massive electricity needs that AI technologies demand.
Ellison stated that AI training is set to be a continual endeavour. He anticipates that a few influential companies and potentially a nation-state will each invest approximately $100 billion in AI development over the next decade. He remarked on the ever-expanding nature of this business, earmarking it as a phase of unceasing growth, with no deceleration on the horizon.
Describing the scope of the project, Ellison mentioned the design of a data centre requiring more than a gigawatt of power. He revealed that Oracle has secured building permits for three SMRs, which he termed as “small modular nuclear reactors to power the data centre.” Although these reactors are a significant step forward, they alone will not suffice to power such a large-scale operation, given that new data centres, like the 77MW Bidder Street Data Center in London, still fall short of the gigawatt requirement.
Ellison did not disclose the exact locations for the SMR building permits. However, he noted that Oracle, being US-based, has extensive operations globally. The current designs of SMRs produce around 300MW each, indicating that three reactors would collectively provide 900MW. Thus, additional generation or supplementary grid capacity will be essential for Oracle’s gigawatt-scale data centre vision.
The utilisation of AI is rapidly becoming an indispensable tool across various civil engineering applications. AI is already being employed in landslip detection, reducing sewer spills, and in the construction and maintenance of tunnels. This strategic move by Oracle highlights the increasing intersection of advanced technology and traditional industries.
Oracle’s pioneering energy solution reflects both the monumental scale of modern AI aspirations and the innovative approaches necessary to sustain them.
