Openai, the company behind Chatgpt, has made a deal with Google to use its cloud computing services. It is a surprising development that both are direct contestants in Artificial Intelligence Space. According to a report, the move was finalized in May after final discussion in May. RootsThe objective is to help face Openi with its growing demand for computing power – known as “calculations” in the industry – because its AI model’s popularity increases.

This step marks a significant change of how AI rival AI is navigating the resource-loving landscape of development. Supported by Microsoft and its Azure cloud platform, now bringing diversity in its infrastructure partnership. The company has recently announced collaboration with Oracle and SoftBank on the $ 500 billion Stargate project and signed billions of deals with Korvives for additional calculations.

Recently, Openai and many American tech companies participated in an attempt to create a large -scale data center complex in Abu Dhabi, G42, G42, an Emiti AI company. This feature, which will be part of a major initiative called Stargate UAE (under Project Stargate), is expected to become one of the world’s largest AI computing hubs. These hubs have servers powered by advanced chips from companies such as Nvidia and operated by Oracle, Cisco and SoftBank.

Allegedly, Google Cloud will now support the training and deployment of Openai’s model, including the Chatgpt. Google’s Cloud Unit, which earned $ 43 billion in sales in 2024, would benefit from this unexpected partnership, and strengthens its growing position as a neutral cloud provider for AI firms – even in direct competition with Google.

Despite being a terrible competitive – Chatgpt is also widely seen as a danger for Google’s dominance in search – both companies have decided to put rivalry aside, at least for now, to solve a more immediate problem: ability.

Openai’s annual revenue run rate was $ 10 billion in June, according to the company, thanks to an increase in users. However, that development has brought its problems. The demand for AI-related material is increasing. In April 2025, Openai launched a new image feature, which allowed users to create scenes in the style of studio Ghibli. It overwhelmed the infrastructure of Openai. The image tool became so popular that CEO Sam Altman made fun of the company’s “JPU melting”.

Additionally, on Tuesday, June 10, Openai suffered a global outage, which affected Chatgpt in web and mobile platforms. While services like Sora and playground have been cured to a great extent, Openai’s status report continued to show a high error rate for chat even after a day. This was the third major outage that was faced by Chatpt this year.

Get away from microsoft

Openai’s decision to work with Google is part of a comprehensive strategy to reduce its dependence on Microsoft, which was a special right to provide data center support via Azure by January. Companies are now negotiating on new investment conditions, including Microsoft’s future stake in OpenEA.

Meanwhile, Google has opened access to its in-house chips known as tenser processing units (TPU), which is first used only for internal operations. This has allowed Google Cloud to attract major customers such as Apple, as well as AI Startups – both, including anthropic and safe superintendent, were established by pre -Openai employees.

There is an increased appetite for calculating AI

As the competition is hot, tech companies are putting money in AI Infrastructure. Google’s original company, alphabet, expects to spend $ 75 billion this year on capital investment related to AI. While it is fuel innovation, it also pressures Google to prove that its AI Prasad can generate strong financial returns. Useful for Google’s cloud ambitions, selling compute power to Openai, means dividing resources between its enterprise customers and their own in-house products, including the Deepmind AI unit.

Google Cfo Anat Ashkenazi admitted in April that the company was already struggling to meet the existing cloud demands. In addition to Openai, resource allocation can be made even more complicated, especially in the form of OpenEE.

Openai’s own chips

To further manage the needs of its calculation, Openai is allegedly in the final stages of designing its own AI chips. The chip, which is to be manufactured by TSMC, is expected to go into large -scale production by 2026. The goal is to reduce dependence on Nvidia, whose hardware currently strengthens the majority of the AI ​​system globally.

By creating a custom chip, Openai will allow its hardware and software to be better integrated, eventually improvement in performance. But this is an expensive and time -consuming process, which is an estimated cost of $ 500 million for the allegedly single chip version, overall the software and infrastructure are likely to double the infrastructure.