For years, one company has owned the chips that power artificial intelligence, and everyone else has paid up. That company is Nvidia. Now the most serious challenge to its grip is coming from a rival using Nvidia’s own moves against it: Google.
The clearest sign came in mid-May, when Blackstone, the investment giant and the world’s largest owner of data centers, said it would put $5 billion into a new cloud company built around Google’s in-house AI chips. Google Cloud chief executive Thomas Kurian said the venture would give companies more ways to rent computing power. With borrowed money added in, the project could eventually command about $25 billion in spending power — and it is aimed squarely at the business Nvidia has dominated.
To understand why this matters, start with the chips.
Nvidia sells graphics processing units, or GPUs, that train and run AI models. Demand has been so high for so long that Nvidia became one of the most valuable companies in the world. Google builds its own AI chips instead, called TPUs, short for tensor processing units. For years they mostly powered Google’s own products. Now Google is selling access to them to outside customers and directly targeting Nvidia’s core market.
Here is the clever part. Nvidia did not just sell chips. It helped customers pay for them.
Using its enormous balance sheet, Nvidia helped support financing for data-center projects, making it easier and cheaper for operators to raise money, build facilities and buy more Nvidia hardware. Google is now running a remarkably similar strategy.
One example sits on the southern shore of Lake Ontario near Niagara Falls. A data-center campus known as Lake Mariner is being developed by TeraWulf, a former bitcoin miner, together with cloud provider Fluidstack. Google has provided roughly $3.2 billion in financial guarantees backing the project. In return, it received warrants that increased its stake in TeraWulf to approximately 14%.
The computing power generated by the site will be rented to Anthropic, the AI company behind the Claude chatbot, and powered by thousands of Google chips.
The strategy does not stop there.
Google has also backed an Anthropic project near Baton Rouge, Louisiana, and guaranteed roughly $1.4 billion in leases tied to a facility in Colorado City, Texas. The formula remains the same: help finance large AI infrastructure projects and then fill those facilities with Google’s hardware.
For local economies, the projects bring substantial investment. The broader Anthropic-Fluidstack infrastructure expansion is expected to create thousands of construction jobs and hundreds of permanent positions across multiple states, adding a major economic development angle to the AI boom.
The effort extends all the way to the top of the AI industry.
Google has agreed to invest up to $40 billion in Anthropic and reserve massive amounts of computing capacity for the company. The arrangement is part of a broader battle among technology giants to secure long-term AI customers. Anthropic has lined up computing resources from Google, Amazon and others as demand for AI processing power continues to surge.
The message from Google is increasingly clear. The company no longer wants to be viewed simply as a search engine and software provider. It wants to become one of the primary suppliers of the infrastructure powering the AI economy.
Nvidia, at least publicly, is not concerned.
Co-founder and chief executive Jensen Huang has repeatedly downplayed the threat posed by custom AI chips. During a widely followed technology podcast appearance in April, Huang argued that Nvidia’s ecosystem is far broader than any individual custom-chip effort can match. He also suggested that Anthropic remains Google’s only major outside TPU customer and questioned whether Google’s chips are actually cheaper when all costs are considered.
Analysts see meaningful change underway nonetheless.
Stacy Rasgon, a semiconductor analyst at Bernstein, said Google is being far more aggressive about monetizing its AI infrastructure than it was in previous years. The reason is straightforward: demand now exists on a scale that simply did not exist before.
Across the technology sector, one complaint dominates conversations among AI developers, cloud providers and investors: there is not enough computing power.
That shortage is shaping the entire industry.
As artificial intelligence evolves from a race over software models into a race over computing capacity, the companies supplying the chips gain enormous leverage. The ability to provide hardware, cloud services and financing has become just as important as the technology itself.
Google recognized that reality years ago when its engineers began designing custom processors for machine-learning workloads long before today’s AI explosion. What started as an internal project has now become the foundation of a major challenge to Nvidia’s dominance.
In the short term, the battle is a corporate showdown between two technology giants. In the longer term, it will help determine who controls the computing infrastructure that powers artificial intelligence.
For the first time in years, Nvidia faces a competitor with the capital, customer relationships, chip technology and patience needed to challenge its position. And rather than inventing a completely new strategy, Google is borrowing directly from the playbook that helped make Nvidia one of the world’s most powerful companies.
JBizNews Desk | Silicon Valley
© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.
