Meta wants to buy cloud services from Google for at least ten billion dollars over the next six years. The deal is the first formal agreement between the two competing internet giants in cloud computing and shows that Meta is primarily concerned with speed in the AI race. Meta already operates around two dozen data centers itself and is having more built, including one with a power consumption of two gigawatts and an area half the size of Manhattan.
But by the time this is finished, as Bloomberg and Reuters report, citing reliable sources, it could apparently be too late for Mark Zuckerberg. Because he has set himself the goal of being the company that provides each of its AI experts with the greatest computing power.
Service provider for the competition
Google is increasingly succeeding in acting as a service provider and provider of computing power and infrastructure for the other heavyweights of the AI world, despite having its own large language models and therefore being in direct competition. In May, Google concluded a similar agreement with OpenAI, which can thus reduce its dependence on Microsoft’s Azure infrastructure.
Nevertheless, such agreements are not unproblematic for Google for two reasons. Firstly, chatbots directly threaten the search engine business. Website operators are increasingly annoyed by a lack of traffic, for example due to Google AI summaries themselves or direct chatbot queries. Secondly, Google could end up lacking the computing power to further develop its own Gemini models.
Meta under pressure
Meta, on the other hand, needs quick successes, as the (semi-)open Llama LLMs are no longer necessarily the first choice for code development and reasoning. This is why, until recently, the company used aggressive methods and extreme salary offers to poach AI experts from its competitors. OpenAI boss Sam Altman complained that Meta was offering transfer bonuses of up to 100 million dollars. Now, however, the ripcord has been pulled and new recruitment has been halted to get the teams working productively on the development of a “superintelligence”.
(ulw)
Don’t miss any news – follow us on
Facebook,
LinkedIn or
Mastodon.
This article was originally published in
It was translated with technical assistance and editorially reviewed before publication.
Dieser Link ist leider nicht mehr gültig.
Links zu verschenkten Artikeln werden ungültig,
wenn diese älter als 7 Tage sind oder zu oft aufgerufen wurden.
Sie benötigen ein heise+ Paket, um diesen Artikel zu lesen. Jetzt eine Woche unverbindlich testen – ohne Verpflichtung!