Google is making an ambitious plan for a big slice of the most important market of the 21st century: the chips that power artificial intelligence. Its financial strength and years of technological development have put it in a prime position to succeed.
It faced one hurdle: tech companies’ fear of crossing Nvidia’s uber-regional chief executive Jensen Huang.
On the southern shore of Lake Ontario, just a short distance from Niagara Falls, Google is demonstrating how it can use Nvidia’s own playbook to win customers.
The site in western New York hosts an AI data-center cluster known as Lake Mariner. Alphabet-owned Google has provided a $3.2 billion financial guarantee for the project, whose developers will rent computing power from thousands of its microprocessors to AI giant Anthropic, according to people familiar with the matter.
It’s the same strategy Nvidia has used repeatedly to increase already growing demand for its own artificial-intelligence chips.
Until recently, Nvidia had that market, its graphics processing units, or GPUs, coveted by tech companies for their power to train and run AI models. But as the AI race has turned into a competition for computing resources over the past year, challengers have begun to move forward – none of them more formidable than Google.
“You have all these very well-capitalized companies that believe there is going to be tremendous value in this market around computing,” said Nazar Khan, co-founder and chief technology officer of AI infrastructure company Terawolf. “They want to stay in the game, they don’t want to be left behind.”
In attracting customers for its chips – known as tensor processing units, or TPUs – Google has copied Nvidia’s practice of using financial guarantees to help data centers raise cheap loans and providing so-called circular financing, where some of the money it invested flows back in the form of chip purchases.
People familiar with the matter say the change in leadership of its cloud unit has increased the level of urgency. Nvidia has a close partnership with OpenAI and is a major investor in it; Google has a similar relationship with Anthropic as well as its own Frontier model, Gemini.
Privately and publicly, Huang has undermined Google’s ability to compete meaningfully with his company.
In April, appearing on podcaster Dwarkesh Patel’s show, Huang said Nvidia has a wide lead over Google and other makers of custom chips, known as ASICs, and argued that Anthropic is Google’s only significant outside customer for TPUs.
“Our market reach is probably much greater than any TPU or ASIC,” Huang said. “I would love to hear them demonstrate the cost benefits of TPU. It doesn’t make sense in my mind.”
Google recently issued its most direct challenge yet made a deal worth 5 billion dollars With Blackstone to set up a new cloud-services company that will compete with CoreWave and Nebius, two Nvidia-backed cloud providers that exclusively use the chip giant’s hardware stack.
“Compared to a few years ago, they’re clearly being more opportunistic and more aggressive about monetizing what they have,” said Stacey Rusgon, a technology analyst at Bernstein. “But a few years ago, the opportunity was not there. Today, all we hear is that no one has adequate calculations.”
keep its chips
Google saw the computing crisis coming a long way, starting with what one of its top scientists called a “thought experiment.”
In 2013, Jeff Dean was working with other artificial-intelligence researchers on speech recognition, using the neural-network techniques that underpin today’s large language models.
“I said, ‘Okay, if we want to scale this speech model to 100 million users, and they use it for a few minutes a day, it would require doubling the number of computers Google has,'” said Dean, now chief scientist at Google’s DeepMind AI lab, in an interview. His conclusion: “We need to build specialized hardware.”
At first, the company kept that hardware for itself. It used the chips to develop AI models and features for its search engine and other products.
As demand for the chips increased, the company began making them available to other companies through its cloud platform. This step has led to rapid growth of that unit.
“Is this the end of Nvidia’s dominance?”. Semianalysis, an influential tech research firm, asked in November in a post linked to the release of Google’s seventh-generation TPU, which Anthropic has used to train its models.
going straight
The tension is increasing. Google stepped up its pace in May by announcing plans to sell its chips directly to customers. The company also unveiled its first TPU optimized for estimation, The type of AI computing involved in posing the questions. Its product will likely face Nvidia’s new Grok 3 LPU.
Mark Lohmeyer, vice president of AI and computing infrastructure for Google Cloud, said the Inference-specific chip, combined with improvements the company has made in making its chips work across multiple systems, has sparked new interest in using TPUs.
“We’re seeing a group of customers who might not have considered it before,” he said.
They include Citadel Securities, a longtime Google Cloud customer, which recently began using TPUs for some of its research software workloads. Josh Woods, the firm’s chief technology officer, said the company can run major workloads with TPU at 30% lower cost and up to four times faster.
There is astronomical demand for AI computing encouraged many challengersWhich includes experienced rivals like Advanced Micro Devices and Broadcom. as well as new entrants Like Cerebras Systems, to take on Nvidia.
Success requires great customer loyalty and breaking through Nvidia’s defensive trenches. Its plug-and-play connectivity hardware and easy-to-use programming library, known as CUDA, are powerful lures for AI labs and large enterprise computing partners. Huang is protective of his company’s market share in AI chips, estimated north of 90%, and sensitive to intrusion by rivals, according to people familiar with the matter.
Adam Fisher, partner at Bessemer Venture Partners, said some neo-clouds are worried they can’t stray from buying the entire stack of Nvidia hardware for fear of being put in “Jensen jail,” meaning they could lose their allotment of Nvidia chips.
“Not all Nvidia neo-cloud will say it that way – some will say Nvidia gives them what they need – but there are some who are dying for something else, but they can’t get it from another supplier,” Fischer said.
Huang has emphasized in public comments that Nvidia welcomes customers buying a la carte.
“Nothing makes me happier than when you buy everything from Nvidia,” the CEO said at a 2025 conference. “But if you buy something from Nvidia it makes me very happy.”
balance sheet weight
Among Nvidia’s challengers, Google stands alone in terms of the financial firepower to attract customers. The company said this month it plans to raise $85 billion in equity to fund its AI infrastructure needs.
Industry insiders pointed to Google’s deal with Blackstone, which has close ties to both Nvidia and CoreWave, as a sign of the changing dynamics resulting from computing shortages. As recently as a year ago, he said, such a deal would have been unthinkable, because companies were afraid of angering Nvidia’s Huang.
“Anybody not named Nvidia is probably going to have to spend more than their balance sheet to get ahead,” said Terawolf’s Khan.
Google is blocking another Anthropic deal, a $7 billion project known as River Bend near Baton Rouge, LA. And in Colorado City, Texas, Google is providing an additional $1.4 billion in financial guarantees for AI computing leases.
Much of the change in approach took place under the leadership of Amin Vahdat, who in December was promoted to chief technologist in charge of Google’s AI infrastructure build-out.. The promotion expanded Vahdat’s portfolio to include chip design, supply and deployment. He now reports to both Google Cloud chief Thomas Kurian and Alphabet chief executive Sundar Pichai.
People who have worked with Vahdat say that as a boss, he demands excellence and has a quiet competitive streak. In 2021, Google raided Intel’s top talent in Israel, appointing 25-year Intel veteran Uri Frank to lead Google’s silicon efforts.
Current and former employees say Google turned its attention to the commercial potential of its TPUs about two years ago, particularly by investing in their inference capabilities.
Since Vahdat’s promotion, Google’s AI infrastructure team has been working more diligently, some people said. A current employee said Vahdat focuses intensely on improving chip performance, often challenging engineers to consistently improve various functions by 10% – a difficult margin to achieve.
Vahdat said in an interview that he is not focused on competing with Nvidia or any other rival in particular, and said the chip giant is a major partner as well as a competitor because Google uses Nvidia GPUs in its data centers. His focus, he said, is simply to create better products for Google and its customers.
“For me and for us, this is not zero-sum,” Vahdat said. “There’s a lot of demand.”
Write to Robbie Whalen robbie.whelan@wsj.com And on Katherine Blunt katherine.blunt@wsj.com
