This week, the United Arab Emirates (UAE), and France announced plans for an AI datacenter campus of one gigawatt dedicated to advancing artificial intelligence.
On Thursday evening, the project to build a datacenter in France was announced during a meeting with UAE President Mohamed bin Zayed Al Nahyan and French President Emmanuel Macron.
With a capacity of one gigawatt, the facility is enormous and dwarfs more traditional cloud and Hyperscale campuses that typically top out at tens or hundreds of megawatts. Despite this, facilities of this size are not unheard of given the insatiable appetite for AI workloads to consume power-hungry graphics cards and high-end networking kit. Meta, a social media giant in the US, has recently begun construction on a 2.3 gigawatt facility in Richland Parish. Louisiana.
The funding for the Franco-Emirati Project will come from a variety of sources, including a reported EUR30-EUR50billion ($31-$51billion) — just a fraction of $500billion associated with the United States Stargate project . MGX, a $100-billion Abu Dhabi investment fund is reportedly involved in both the French datacenter and ambitious Stargate initiatives. The nations announced in a joint press release that the first phase of UAE investment will be announced at the Choose France 2025 Summit later this year.
In addition to supporting the expansion and infrastructure of datacenters, the partnership will see the UAE and France work together to acquire semiconductors and talent of the highest quality, as well as establish virtual data embassies in order to develop sovereign AI and cloud service in both countries. The tie-up was not unexpected. The UAE firms, like G42, have courted several high-profile AI infrastructure providers such as Microsoft or Cerebras. This is because the UAE has sought to protect itself from export controls that are becoming increasingly restrictive on American-designed AI accelerators. If implemented by the Trump Administration, the latest export restrictions set out by the Biden administration will severely limit how many accelerators are available to the UAE and the rest of the world.
France is home to Mistral.AI, one of Europe’s leading model developers. The startup has produced large language models, the kind that power AI chatbots such as ChatGPT or Gemini, that are competitive with its larger and better funded rivals in US.
- SoftBank’s $40B woos OpenAI, making Microsoft’s modest $13B look small
- Datacenters’ energy consumption will more than double by 2030 due to AI’s insatiable appetite[196590010]Hardware problems and server supply-chain kinks slow Amazon’s $100 billion AI project
- DeepSeek’s impact on AI hardware is not a concern for AMD CEO Lisa Su
Speaking of Not AMD CEO Lisa Su.
Speaking about Mistral, Cerebras – backed by UAE AI darling G42 – revealed that it is providing its wafer scale compute platform to French model builder. Cerebra’s “Flash Answers” platform for Mistral’s Le Chat platform promises 1,100 tokens per second generation speed. However, Cerebra has inflated the numbers a little by using speculative coding .
These are announcements made on the eve before the Artificial Intelligence Action Summit, which begins next week in Paris. Discussions will likely revolve around the recent developments surrounding Chinese model maker DeepSeek and the newly elected US President Donald Trump. (r)
