It isn’t your imagination: Google Cloud is flooding the zone

The partnership of $ 100 billion between Nvidia and OpenAi, announced on Monday, represents the last mega-deal that reforms the AI ​​infrastructure landscape. The Agreement includes non-voting shares that are linked to massive chip purchases and sufficient computing power for more than 5 million American households, so that the relationship between two of AI’s most powerful players is deepened.

In the meantime, Google Cloud completely places another bet. While the largest players in the industry is constantly cementing decreasing partnerships, Google Cloud Hellbent is on catching the next generation AI companies before they get too big in court.

Francis Desouza, his COO, has seen the AI ​​revolution from several vantage points. As a former CEO of Genomics Giant Illumina he saw machine learning drug discovery transform. As a co-founder of a two-year AI Calms Startup, Synth Labs, he has struggles with the safety challenges of increasingly powerful models. Now that he had joined the C-suite in Google Cloud in January, he orchestrates an enormous commitment on the Second Gulf of AI.

It is a story that Desouza likes to tell in numbers. In a conversation with this editorHe notes several times that nine from the top 10 AI laboratories use the infrastructure of Google. He also says that almost all generative AI -a horns on Google Cloud are performed, that 60% of all Gen AI startups have chosen Google worldwide as their cloud provider, and that the company has drawn up $ 58 billion in new sales obligations in the coming two years, which has more than double the current Run rate.

Asked what percentage of the income from Google Cloud comes from AI companies, offers instead that “AI the cloud market reset, and Google Cloud is paramount, especially with startups.”

The NVIDIA-OPENAI deal is an example of the scale of consolidation that sweeps AI infrastructure. Microsoft’s original $ 1 billion openi -investment has grown to almost $ 14 billion. Amazon followed with $ 8 billion in anthropic investments, in which deep hardware adjustments were obtained that in essence AI training adjust to work better with the infrastructure of Amazon. Oracle has also emerged as a surprise winner, who has launched a cloud agreement of $ 30 billion with OpenAI and then a stunning $ 300 billion five -year commitment secured from 2027.

Even Meta, despite building his own infrastructure, signed a $ 10 billion deal with Google Cloud while he has $ 600 billion plant in American infrastructure expenditure up to and including 2028. The “Stargate” project of the Trump administration, the $ 500 billion project, with Softbank, OpenAIs and Oracle.

These gigantic deals may seem threatening to Google, given the partnerships that companies such as OpenAi and Nvidia seem to cement elsewhere. In fact, it looks a lot like Google is cut out of some crazy dealing.

But the company games are not exactly in your hands. Instead, Google Cloud signs smaller companies such as Lovelable and Windsurf – what Desouza calls the “Next generation of companies that” calls income ” – as” primary computer partners “without major investments.

The approach reflects both opportunities and necessity. In a market where companies can go “from starting to a multi-billion dollar company in a very short period”, as Desouza says, capturing future unicorns before they can be more mature more valuable than fight today’s giants.

The strategy extends beyond simple customer acquisition. Google offers AI-startups $ 350,000 to cloud credits, access to its technical teams and go-to-market support via its marketplace. Google Cloud also offers some Desouza as a “no compromise” AI pile – from chips to models to applications – with an “open Ethos” that gives customers with every low choice.

“Companies love the fact that they can access our AI stack, they can access our teams to understand where our technologies are going,” says Desouza during our interview. “They also like to have access to the infrastructure of the Google Class Enterprise class.”

Google’s infrastructure game has recently become even more ambitious, with reporting that revealed the maneuver of the company behind the scenes to expand its adapted AI chip activities. According to the informationGoogle has concluded deals to place its tensor processing units (TPUs) for the first time in the data centers of other cloud providers, including an agreement with the London-based Fluidstack with a maximum of $ 3.2 billion in financial support for a facility in New York.

Comparing directly with AI companies, while at the same time providing an infrastructure, Finesse requires. Google Cloud offers TPU chips for OpenAI and organizes Anthropic Claude model via its Vertex AI platform, even while his own Gemini models compete with both of the head. (The parent company of Google Cloud, Alphabet, also has an interest of 14% in Anthropic, according to the court documents of the New York Times that were obtained earlier this year, although when you are immediately asked about the financial relationship of Google with anthropic, Desouza, the relationship calls a ‘Meerlagige Partnership’ and quickly revised to Google Clouds.

But if Google tries to be Switzerland while promoting his own agenda, it has had a lot of practice. The approach has roots in the open-source contributions from Google, from Kubernetes to the basics “Attention is all you need“Paper with which the transformerarchitecture was based on the most modern AI. More recently, Google published an open-source protocol called Agent-too-Agent (A2A) for communication between agent in an attempt to demonstrate his continuous dedication to openness, even in competful areas.

“Over the years we have made the explicit choice to be open at every layer of the stack, and we know that this means that companies can absolutely take our technology and use it to build a competitor at the next layer,” Desouza acknowledges. “That has been happening for decades. That is something we are good with.”

Google Cloud’s dating of startups comes at a particularly interesting moment. This month, Federal Judge Amit Mehta has made a nuanced statement in the five-year search monopolistic matter of the government, in an attempt to curb the dominance of Google without hindering the AI ​​ambitions.

While Google avoided the most seriously proposed fines of the Ministry of Justice, including the forced disinvestment of his Chrome -Browser, the prevailing concerns about regulatory concerns about the company that uses his search monopoly to dominate AI. Critics are understandably ensure that the huge search from Google offers an unfair advantage in the development of AI systems, and that the company could use the same monopolistic tactics that have protected its search dominance.

In a conversation, Desouza is aimed at much more positive results. “I think we have the opportunity to fundamentally understand some of the most important diseases that we do not have good insight into today,” says Desouza, for example, where a vision is set out in which Google Cloud helps to support research into Alzheimer’s, Parkinson’s and climate technologies. “We want to work very hard to ensure that we pioneer in the technologies that make that work possible.”

Critics cannot be easily claimed. By positioning themselves as an open platform that enables the next generation AI companies instead of checking, Google Cloud shows supervisors that it promotes competition instead of suppressing it, while they forge relationships with startups that can help the Google matter as regulators increase pressure.

View this week for our full conversation with Desouza Download strictlyVC podcast; A new episode comes out every Tuesday.