It’s More Than Just Your Imagination: Google Cloud is Dominating the Market

The newly unveiled $100 billion partnership between Nvidia and OpenAI signifies a pivotal shift in the AI infrastructure domain. This deal encompasses non-voting shares tied to major chip acquisitions and offers enough computing power to support over 5 million American households, further cementing the collaboration between two heavyweight players in AI.

In contrast, Google Cloud is adopting a different approach. While major industry players are forging tighter alliances, Google Cloud is intent on capturing emerging AI startups before they become too large to work with.

Francis deSouza, the COO, brings a unique perspective on the AI revolution. As the former CEO of Illumina, a leading genomics company, he observed how machine learning transformed drug discovery. He also co-founded Synth Labs, an AI alignment startup, where he tackled the safety challenges posed by increasingly powerful AI models. Now, as part of the C-suite at Google Cloud since January, he is spearheading a significant investment in the next phase of AI.

DeSouza often reinforces this narrative with impressive statistics. In a discussion earlier this week, he pointed out that nine of the top ten AI labs utilize Google’s infrastructure. Additionally, he noted that virtually all generative AI unicorns are hosted on Google Cloud, with 60% of global generative AI startups selecting Google as their cloud provider. Furthermore, the company has secured $58 billion in new revenue commitments over the next two years, double its current annual run rate.

When questioned about the share of Google Cloud’s revenue from AI companies, he underscored, “AI is transforming the cloud market, and Google Cloud is leading, especially with startups.”

The Nvidia-OpenAI alliance illustrates the ongoing consolidation trend in AI infrastructure. Microsoft’s initial $1 billion investment in OpenAI has skyrocketed to nearly $14 billion. Amazon has countered with an $8 billion investment in Anthropic, allowing for significant hardware adaptations that enhance AI training within its infrastructure. Notably, Oracle has emerged as a critical player, securing a $30 billion cloud deal with OpenAI and an impressive $300 billion commitment slated to begin in 2027.

Even Meta, while developing its own infrastructure, has entered a $10 billion agreement with Google Cloud and plans to invest a staggering $600 billion in U.S. infrastructure by 2028. The “Stargate” initiative, involving SoftBank, OpenAI, and Oracle, further complicates these interconnected collaborations.

These enormous deals may pose challenges for Google, given the strengthening partnerships that firms like OpenAI and Nvidia appear to be cultivating elsewhere. It seems that Google may be on the sidelines during some of these high-stakes negotiations.

Nonetheless, the tech behemoth is not standing idle. Google Cloud is forming partnerships with smaller companies like Loveable and Windsurf—entities deSouza describes as the “next generation of companies”—serving as “primary computing partners” without requiring significant upfront investments.

This approach signifies both opportunity and necessity. In an environment where startups can quickly evolve into multi-billion-dollar companies, engaging emerging unicorns before they mature could offer more advantages than competing for established giants.

This strategy transcends basic customer acquisition. Google provides AI startups with $350,000 in cloud credits, technical team access, and go-to-market support through its marketplace. Google Cloud also features what deSouza terms a “no compromise” AI stack—consisting of chips, models, and applications—with an “open ethos” that empowers customers at every level.

“Companies value our AI stack access and appreciate engaging with our teams to grasp the trajectory of our technologies,” deSouza remarked during our conversation. “They also prioritize receiving enterprise-grade infrastructure from Google.”

This infrastructural advantage was recently highlighted when reports surfaced about Google’s strategic moves to expand its custom AI chip business. According to The Information, Google has made arrangements to deploy its tensor processing units (TPUs) in the data centers of other cloud providers for the first time, including a partnership with London-based Fluidstack involving up to $3.2 billion in funding for a New York facility.

Balancing the roles of competing with AI companies and simultaneously offering them infrastructure demands skill. Google Cloud provides TPU chips to OpenAI and supports Anthropic’s Claude model via its Vertex AI platform, even as its Gemini models compete directly with both. (Google Cloud’s parent organization, Alphabet, holds a 14% stake in Anthropic, as noted in court documents from the New York Times earlier this year; however, when specifically asked about Google’s financial connections with Anthropic, deSouza referred to the relationship as a “multi-layered partnership” before quickly redirecting to Google Cloud’s “model garden,” emphasizing that customers can access various foundational models.)

However, if Google aims to maintain a neutral position while advancing its agenda, it has ample experience in this regard. This strategy is rooted in Google’s contributions to open source, ranging from Kubernetes to the pivotal “Attention is All You Need” paper that laid the groundwork for modern AI transformer architecture. Most recently, Google introduced an open-source protocol called Agent-to-Agent (A2A) to enhance inter-agent communication, reinforcing its commitment to openness amidst competitive pressures.

“We have intentionally chosen to remain open at every layer of the stack, and we recognize that this allows companies to utilize our technology to create competitors at the next tier,” deSouza admits. “This has been happening for decades, and we accept it.”

Google Cloud’s focus on startups comes at a crucial juncture. Just this month, federal judge Amit Mehta delivered a nuanced ruling in the government’s search monopoly case, trying to curb Google’s dominance without impeding its AI ambitions.

While Google evaded the Justice Department’s most severe penalties, including the forced divestment of its Chrome browser, the ruling signaled regulatory concerns regarding the firm’s use of its search monopoly to dominate AI. Critics express legitimate concerns that Google’s extensive search data provides it an unfair advantage in developing AI systems and worry that the company might apply similar monopolistic tactics that facilitated its search supremacy.

In discussions, deSouza remains hopeful about future outcomes. “We have a chance to fundamentally deepen our understanding of major diseases that remain elusive,” he says, envisioning a future where Google Cloud contributes to research in Alzheimer’s, Parkinson’s, and climate technologies. “We are dedicated to pioneering the technologies essential for this important work.”

Skeptics may still be unconvinced. By positioning itself as an open platform that empowers rather than restricts the next generation of AI companies, Google Cloud could be signaling to regulators that it fosters competition rather than stifling it, all while building relationships with startups that may bolster Google’s stance if regulatory scrutiny intensifies.

For the complete conversation with deSouza, be sure to listen to this week’s StrictlyVC Download podcast; new episodes drop every Tuesday.