Equinix partners with Nvidia to carve a niche in the AI ​​data center boom

Long before the era of hyperscalers and neoclouds, and decades before the AI ​​data center boom. Equinix.

Dubbed a colocation facility, Equinix provides data center space to more than 10,500 customers, providing space, power, cooling and security to businesses that need a place to house servers, routers and storage systems.

With its experience dating back to the 1998 dotcom success, Equinix has a significant and growing role in the artificial intelligence data center market, with hyperscalers expected to spend more than $5 trillion by 2030, according to Goldman Sachs Research. On Wednesday, Equinix signed a deal with. Nvidia This gives customers a flexible way to run their AI models on the open source cloud platform Together AI.

Equinix’s stock price has risen 33% this year, outpacing all the mega-tech companies and giving the company a market capitalization of $100 billion. It is by far the most valuable data center real estate investment trust (REIT), surpassing Digital Realty, which has a market capitalization of $68 billion.

like a hyperscaler Amazon and googleand like Neocloud coreweave Collectively, companies spend hundreds of billions of dollars a year on infrastructure, a huge amount of which goes to major AI labs like OpenAI and Anthropic. In contrast, Equinix rents out space in its facilities to customers of all shapes and sizes who can run servers on a variety of computing platforms. Nvidia to AMD.

On Wednesday, Equinix CEO Jensen Huang spoke via video with Equinix CEO Adair Fox Martin at an event related to the NVIDIA deal in San Francisco, California. Huang said the location of Equinix’s facility allows it to be “close to where the action is happening, where all the sensors are located.” And because the architecture is distributed, “you can be close and far at the same time,” he says.

Equinix is ​​one of the oldest companies in the data center business. The company’s 281 legacy colocation facilities span 77 metropolitan areas on six continents, said Mariam Zand, the company’s vice president of AI ecosystem strategy.

“All the companies you already use are running on us,” she says.

Financing details of Wednesday’s deal were not disclosed. Zand said Together AI, which provides access to 200 open source models, will be the seller of record and charge end customers for using the new program. The service, called Equinix Inference Exchange, is expected to be available in the first quarter of 2027.

While AI training teaches models to learn from patterns in large amounts of data, inference refers to decisions made based on new information. As AI evolves from simple call-and-response chatbots to more complex agent apps, inference becomes more important than training. This means that workloads must run on a variety of chips, including central processing units (CPUs), rather than relying solely on general-purpose flagship graphics processing units (GPUs).

Equinix has largely stuck to its traditional strategy of building small data centers near city centers and positioning them as network interconnection hubs. digital realtyby contrast, has begun to focus more on large-scale facilities for hyperscale customers.

In the case of xScale, the hyperscaler-serving side of Equinix’s business, most facilities are small. Although some AI data centers are measured in gigawatts, they are less than 100 megawatts.

Vlad Galabov, a longtime data center analyst, said Equinix is ​​”too slow” to meet gigawatt-scale demand.

“There were some hungrier newcomers who came in to start building data centers for some of these burst projects and quickly took off,” said Galabov, host of the podcast AIDC Discussions. “Equinix, Digital Realty and all colocation companies now recognize the need to plan more strategically.”

Equinix’s announcement on Wednesday also includes Equinix Fabric One, a new connectivity service aimed at simplifying networks operated using multiple clouds and AI models.

Zand described the announced program by Nvidia and Together AI as an “inference-as-a-service platform.” This will allow customers to connect to different clouds and providers, run inference on open source models, and “optimize tokenomics”, thus reducing costs.

He told CNBC that Equinix’s data centers are optimized for Nvidia’s B300 Blackwell Ultra GPUs, but also have some water cooling facilities where customers can use the new Vera Rubin chips.

Equinix has a particular advantage in urban areas, which are important for inference and the high-speed communications required between servers and end users. According to McKinsey, by 2030, inference will account for half of all AI computing and 30-40% of total data center demand.

“In today’s environment, throughout the AI ​​ecosystem, change is very rapid,” Zand said. “Customers need to be able to move with the market.”

Not all investors are excited about the story.

Short seller Jim Chanos told CNBC’s “Closing Bell” in May that he was betting on the company’s stock as well as Digital Realty, saying “they’re not a great business” and “they’re not making a lot of money.”

“These are businesses with very low return on capital, very capital-intensive businesses, and they don’t grow very quickly,” he said.

He added: “Traditional data center companies are clearly different than data centers that are being built for AI.”

In its most recent quarter, Equinix reported revenue of $2.63 billion, up 16% year-over-year. coreweaveone of the leading neoclouds, more than doubled its revenue to $2.58 billion. However, Equinix reported net income of $477 million in the same period, while CoreWeave posted a loss of $626 million.

Galabov said Equinix is ​​”very diversified” with a huge amount of general-purpose computing and services across a very broad customer base.

“They are not at risk of an AI bubble,” Galabov said. “It means you’re missing out on some of the boom. That’s inevitable. High risk, high return.”

WATCH: Jim Chanos says Equinix and digital real estate are ‘not a great business’