On September 2, 2026, Jensen Huang, the CEO of Nvidia, participated in a discussion with Equinix CEO Adaire Fox-Martin during an event in San Francisco, California. Equinix, long before the rise of hyperscale cloud providers and the current AI data center surge, has established itself as a prominent colocation facility. The company has provided data center space since 1998, accommodating over 10,500 customers with power, cooling, security, and space for their servers and storage needs. As per Goldman Sachs Research, Equinix is poised to play a significant role in the AI data center market, with hyperscalers expected to invest over $5 trillion by 2030.
Recently, Equinix announced a partnership with Nvidia, introducing a flexible framework for customers to utilize their AI models through the open-source cloud platform Together AI. Its stock has surged by 33% this year, leading the megacap tech sector and elevating its market capitalization to $100 billion, making it the foremost data center real estate investment trust (REIT) ahead of Digital Realty, which stands at $68 billion.
As major tech firms like Amazon and Google, along with newer companies like CoreWeave, collectively spend enormous sums on infrastructure—specifically to support significant AI laboratories like OpenAI and Anthropic—Equinix continues to rent out space to a diverse clientele. This enables the operation of servers on multiple computing platforms, including Nvidia and AMD systems. During the event in San Francisco, Huang emphasized that Equinix's geographic advantages position clients "close to where the action is," allowing for a distributed architecture that facilitates vital connections.
With 281 legacy colocation facilities across 77 metropolitan areas and six continents, Equinix is one of the industry’s long-standing players. Maryam Zand, a vice president at Equinix overseeing AI ecosystem strategy, noted, "The companies you already use are all running on us." Though financial specifics of the Nvidia deal remain undisclosed, Zand indicated that Through AI, which grants access to 200 open-source models, will be the official seller and bill customers utilizing the new service named Equinix Inference Exchange, expected to launch in the first quarter of 2027.
Focusing on inference—where AI systems make decisions based on new data—this component is becoming increasingly vital as the technology matures from simple responses to sophisticated applications. Inference now requires more diverse chip utilization beyond traditional graphics processing units (GPUs) and calls for central processing units (CPUs) as well. While Equinix continues its strategy of establishing smaller, city-centered data centers positioned as hubs for network interconnections, Digital Realty has shifted towards deploying extensive facilities aimed at service hyperscale clients.
According to Vlad Galabov, a data center analyst, Equinix has been slower to adapt to the demand for larger facilities as more aggressive competitors have constructed data centers for rapidly expanding projects. The trends indicate that both Equinix and other colocation companies are now more conscious of the need for strategic planning.
Recent announcements from Equinix also included their new service, Equinix Fabric One, designed to streamline the connectivity across various cloud services and AI models. Zand characterized the newly launched program with Nvidia and Together AI as an "inference platform as a service," enabling clients to connect to different cloud providers efficiently, employ inference on open-source models, and optimize costs. Equinix facilities are tailored for Nvidia’s B300 Blackwell Ultra GPUs and feature liquid-cooled environments for advanced Vera Rubin chips. Their urban location offers significant advantages for inference, especially for high-speed server communications with end users.
As the AI landscape evolves, McKinsey predicts that by 2030, inference will account for half of all AI computing and 30% to 40% of total data center demand. Zand remarked on the rapid transformations within the AI ecosystem, emphasizing that customers need the flexibility to adapt to market changes. However, not all investors share this optimism; short seller Jim Chanos has publicly expressed skepticism regarding the profitability of mature data center operators, labeling them as low-return and capital-intensive endeavors. While Equinix reported a 16% revenue increase to $2.63 billion last quarter, CoreWeave, another participant in the neocloud sector, saw its revenue more than double to $2.58 billion, albeit with a significant net loss. Galabov further underscored Equinix's broad diversification, shielding it from the risks associated with the AI sector bubble, suggesting that while it may miss out on rapid gains, its stability is advantageous in the long run.
