They are blowing hot (and especially cold) on the stock market: what are neoclouds?

Receive a summary of important news every evening with Le Récap’

On November 3, 2026, the United States experienced a historic election night. What many analysts saw as a Democratic blue wave turned into a tsunami during the midterm elections. Both the Senate and the House of Representatives have swung to the Democratic camp, to the great dismay of the Republicans and their leader Donald Trump.

A snub for the American president, but also for the entire MAGA sphere, with Elon Musk in the lead. The billionaire has set up most of his businesses in Texas, a state historically favored by the Republican camp, which has just elected a Democrat. The new governor, Gina Hinojosa, intends in particular to freeze the construction of new data centers. A prospect that makes data centers in orbit, which Elon Musk defends, “rather attractive at the moment”.

The data centers were also put to the vote directly. On November 3, residents of dozens of American communities voted on the development of these projects, and the restriction camp won in 40 of the 41 votes.

Local votes of course, sometimes only consultative, but the signal is clear for investors. In this context, the sector's stock market values ​​suffered a brutal correction, led by those of the neocloud.

If the Nasdaq fell by 5.5% on November 5, CoreWeave, Nebius and IREN, three of the most followed listed neoclouds on Wall Street, fell by 10, 15 and 17% respectively. A decline comparable to that of chip manufacturers, but much more marked than that of traditional cloud giants, such as Microsoft (-3.6%), and which can be explained by their economic model.

Because a neocloud is a cloud provider specialized solely in the rental of computing power for artificial intelligence. While historic players in the sector, such as Amazon Web Services, Microsoft Azure or Google Cloud, offer hundreds of services, from storage to databases and software, neoclouds focus on a single activity. It involves running thousands of Nvidia graphics chips to train and operate AI models.

An activity that requires heavy investments. Neoclouds buy these processors en masse and install them in data centers designed to support their electricity consumption and heat, which pushes them into debt, like CoreWeave which had $25.1 billion in debt at the end of the first quarter.

They then rent this capacity, most often under contracts lasting several years. Their clients are AI laboratories like OpenAI, but also the hyperscalers themselves, led by Microsoft, who entrust them with part of their needs to deal with a capacity shortage.

Machine-translated from French