The $ 100 billion partnership between NVIDIA and Openai, which was announced on Monday, is-at the present time-the most modern Friday that reshapes the infrastructure scene of Amnesty International. The agreement includes the non -vote shares associated with the purchases of huge chips and adequate computing power of more than 5 million American families, deepening the relationship between two of the most powerful artificial intelligence players.
Meanwhile, Google Cloud puts a completely different bet. While the largest players in this field reinforce the partnerships that are ever done, Google Cloud is hell to capture the next generation of artificial intelligence companies before they become very large in court.
Francis Desouza, Director of Operations, has witnessed the Amnesty International Revolution of multiple points. As the former CEO of Genomics Giant Illumina, an aid learning witness converts the discovery of drugs. As a co -founder to start the alignment of the two -year -old artificial intelligence, it has increasingly struggled with safety challenges in strong models. Now, after joining the C-Suite in Google Cloud in January, he organizes a huge bet in the second wave of artificial intelligence.
It is a story that Desouza loves to tell in the numbers. In a conversation with this editorIt is noted several times that nine of the top 10 Amnesty International laboratories use Google’s infrastructure. He also says that almost all artificial intelligence is on Google Cloud, and that 60 % of all emerging companies around the world around the world have chosen Google as a cloud provider, and that the company has lined up $ 58 billion in new revenue obligations over the next two years, which represents more than twice the current annual operation.
When asked about the percentage of Google Cloud revenue that comes from artificial intelligence companies, it is instead that “AI reset the cloud market, and Google Cloud drives the road, especially with startups.”
The NVIDIA-Openai deal embodies the volume of infrastructure to unify artificial intelligence. The original dollar Openai Investment has grown billion dollars to nearly 14 billion dollars. Amazon followed $ 8 billion in human investments, as she got allocations for deep devices that mainly design artificial intelligence training to work better with Amazon infrastructure. Oracle has also emerged as a sudden winner, as well as he got a $ 30 billion cloud deal with Openai, then secured a $ 300 billion commitment from five years starting in 2027.
Even Meta, despite building its infrastructure, a $ 10 billion deal with Google Cloud with planning for US infrastructure value of $ 600 billion, and adds another project from “Stargate”, which includes SoftBank, Openai and Oracle, on another layer to these intertwined partnerships.
TECHRUNCH event
San Francisco
|
27-29 October, 2025
These giant deals may seem to threaten Google, given the partnerships that it seems that companies like Openai and Nvidia seem to widen elsewhere. In fact, Google appears to be cut from some feverish deals.
But the company’s giant does not sit at his hands exactly. Instead, Google Cloud signs smaller companies such as Loveable and Windsurf – what Desouza calls “the next generation of upcoming companies” – as “initial computing partners” without significant investments provided.
The approach reflects opportunities and necessity. In the market where companies can move “from being an emerging company to a company with billions of dollars in a very short period of time,” says Desouza, the seizure of future uniforms before ripening can prove more valuable than fighting on today’s giants.
The strategy extends beyond the acquisition of simple customers. Google offers startups AI $ 350,000 of cloud credits, reaching their technical teams, and supporting going to the market through the market. Google Cloud also provides what Desouza describes as stacked by Amnesty International “no compromise” – from chips to models to applications – with “open spirit” that gives customers an option in each layer.
“Companies love the fact that they can reach the staple of artificial intelligence, and they can reach our teams to understand where our technologies are going.” “They also like to get access to the Google Class Infrastructure.”
Google’s infrastructure game has become more ambitious recently, as it reported the scenes of the scenes to expand the scope of dedicated artificial intelligence chips. According to informationGoogle has entered into deals to put TPUs in other cloud service providers for the first time, including an agreement with London -based Fluidstack, which includes up to $ 3.2 billion of financial support for the New York facility.
It competes directly with the artificial intelligence companies while providing its infrastructure at one time – let’s call it – Finese. Google Cloud provides TPU chips to Openai and hosts the Claude model through Vertex AI platform, even because their Gemini models compete directly with both. (Alphabet, Alphabet, also owns a 14 % stake in the New York Times tight documents obtained earlier this year, although it is asked directly about Google’s financial relationship with Antarbur, Desouza calls the relationship a “multi-layer partnership” and then quickly re-directs you to the Google Cloud market-noting that customers can reach forms Various basic.)
But if Google tries to be Switzerland as its own agenda advances, it has many practices. This approach has roots in open source Google contributions, from kubernetes to institutional “Attention is all you need“The paper that enabled the intention of the transformers behind most of the modern artificial intelligence. Recently, Google has published an open source protocol called the agent to the agent (A2A) to communicate between the agents in an attempt to prove his continuous commitment to openness even in the competitive areas.
“We have taken the explicit choice over the years to be open in each layer of stack, and we know that this means that companies can take our technology absolutely and use it to build a competitor in the next layer,” Desouza admits. “This has happened decades ago. This is fine.”
Google Cloud’s flirting for startups comes in particular an interesting moment. Only this month, Federal Judge Amit Mihitt issued an accurate judgment in the case of monopoly in the five -year -old government, in an attempt to reduce Google’s dominance without hindering the aspirations of artificial intelligence.
While Google avoided the most severe proposed penalties in the Ministry of Justice, including the forced disposal of its Chrome browser, the ruling emphasized organizational concerns about the company that benefits from its monopoly to search for artificial intelligence. It is understood that critics are concerned that the wide search data group of Google provides an unfair feature in developing artificial intelligence systems, and that the company can publish the same monopolistic tactics that obtained the research dominance.
In the conversation, Desouza focuses on more positive results. “I think we have an opportunity to understand some of the main diseases that we do not face a good understanding today,” Desouza says. “We want to work hard to make sure we are pioneers of technologies that will enable this work.”
Critics may not be easily calm. By placing itself as an open platform that enables instead of controlling the next generation of artificial intelligence companies, Google Cloud may show the organizers that it enhances competition rather than strangling it, all while enhancing relations with startups that may help in Google if the organizers raise pressure.
For full chat with Desouza, check this week A strict download Podcast is a new episode that comes out every Tuesday.
