Faced with the explosion of artificial intelligence, the European cloud arms race

Rare, expensive and essential graphics cards, the cornerstones of artificial intelligence (AI) development, are the subject of intense competition among European cloud operators, who rely on significant computing power to rent to their customers, in a context of tension offer.

Eight days apart at the end of January, two French companies, OVHcloud and Scaleway (a subsidiary of the Iliad group, owned by entrepreneur Xavier Niel), announced new graphics processors (GPUs in English). computing units much faster and more powerful than a classic microprocessor) from the American brand Nvidia.

You know the DATAROOM of FRENCHWEB.FR our database of French start-ups and innovative companies: basic information, raised funds, turnover, organization chart, growth areas. Access the information we have collected for more than 1000 companies

Once acquired by these companies, these GPUs can be rented remotely by customers who wish to develop AI models, for hourly or even small fees.

The principle is widely used in Europe: the German giant SAP, as well as the small British start-up Ori, are trying to exist against the American heavyweights Google, Amazon and Microsoft.

Faced with competition from these behemoths that, according to the French Competition Authority, accounted for “80% of the increase in spending on infrastructure and public cloud service applications in France” in 2021, the Europeans are playing the dominance card.

If the French start-up Mistral AI, which wants to establish itself as a European alternative to the American leader OpenAI, has mainly chosen to train its Mixtral model on Scaleway, “it's also a matter of image”, notes Hanan Ouazan, partner. at Artefact, a data and artificial intelligence consultancy. “The sovereignty argument would collapse if all the data ended up on US servers.”

Appearing in 1999 and first used in video games, graphics cards are now getting a second wind. “There is no alternative to producing artificial intelligence other than using GPUs,” explains Hanan Ouazan.

But their cost is high: if the price of the star H100 model, marketed by Nvidia, is not disclosed, it would sell for around $40,000, according to the most common estimate in the market. However, a single GPU is far from sufficient for the development of sophisticated artificial intelligence, and their maintenance is also a delicate matter.

– Opacity –

Therefore, the cloud becomes an interesting solution for AI developers, including the biggest ones like Mistral AI.

However, cloud platforms are not immune to recurring supply tensions in the market for semiconductor materials, on which GPUs are based. Therefore, every operator tries to do their part to take advantage of the deliveries of the latest graphics cards.

In fact, Nvidia accounted for 82% of GPUs shipped worldwide at the end of 2022, according to California-based Jon Peddie Research. Contacted by AFP, the company would not confirm, but it does list several hundred “partners” in Europe on its website, a situation that seems far from guaranteeing a privileged connection.

“What makes the difference is the technical know-how: dealing with GPUs that are extremely rare, the worst thing that can happen to Nvidia is someone who buys GPUs and doesn't know how to get them to people who need them,” argues Damien Lucas . , managing director of Scaleway.

David Chassan, director of strategy at Outscale, a cloud subsidiary of Dassault, praises a “very close relationship with Nvidia,” which allows the company to forecast demand and availability for a year.

The platforms also maintain a certain opacity, refusing to reveal the number of GPUs and H100 models they own. “In this environment, to exist, you have to count in the thousands,” admits the general manager of Scaleway, however, who clarifies that his company has made an investment of 100 million euros in artificial intelligence in 2023, mainly for the acquisition of GPUs.

Despite this aggregation and flexibility that clouds promise, the scarcity of the resource is still felt by developers. “We have customers who invest directly in GPUs because they can't have committed usage rates unless they commit to a cloud provider to use GPUs for a fairly high amount,” observes Hanan Ouazan.

The latest articles of THE EDITORIAL OF FRENCHWEB.FR (see all)

Leave a Reply

Your email address will not be published. Required fields are marked *