World's 2 Largest Data Center Firms Say DeepSeek Will Spark 'Avalanche' Of Demand

The unveiling of DeepSeek’s artificial intelligence model last month raised questions about the data center industry’s long-term growth. But data center REITs Equinix and Digital Realty say the Chinese firm’s innovations will drive a massive flow of AI demand to their portfolios. 

When DeepSeek triggered a Wall Street sell-off of data center, tech and energy stocks last month, investors were concerned that the more efficient AI training the firm pioneered could reduce the demand for data center capacity. While Big Tech firms have since indicated they have no plans to slow their data center spending anytime soon, the idea that fewer megawatts of computing will be needed to train and develop AI models has made some investors wary of the long-term data center growth predictions.

Yet leaders of the world’s two largest data center providers say that, far from being a threat, DeepSeek has kickstarted a fundamental shift in how AI drives demand that will ultimately be a boon for their businesses. 

Following the release of fourth-quarter earnings results this week, top executives at Equinix and Digital Realty said that cheaper AI training will accelerate demand from AI inference and corporate AI deployments. They said these types of workloads are tailor-made for their portfolios centered around multitenant colocation and interconnection in major metro areas.

“We're going to continue to see AI being democratized,” Digital Realty Chief Technology Officer Chris Sharp said on the REIT's earnings call Thursday. “This shift will drive higher and higher AI utilization to more and more customers, ultimately creating more and more demand for our facilities.”

While Digital Realty and Equinix operate the most extensive global data center portfolios, with more than 550 facilities between them, the two REITs have thus far not been the primary beneficiaries of the AI-boosted data center boom that has transformed the industry over the past five years. 

Much of the industry’s rapid growth has come in the form of ever-larger campuses with access to hundreds of megawatts of power as cloud providers and social media giants scramble to secure the infrastructure to support the training of generative AI models. Most such campuses have been self-developed by the world’s largest tech companies or developed by private firms backed by institutional capital — companies like Stack Infrastructure, PowerHouse Data Centers and Compass Datacenters.

The two REITs do build large-scale campuses and single-tenant facilities: hyperscale accounts for a significant percentage of Digital Realty’s business, while Equinix has ramped up its hyperscale joint venture development program over the past 18 months. But their business models are largely geared toward providing multitenant colocation and connectivity solutions close to large population centers.

Additionally, as public companies, the REITs face constraints on debt and capital deployment that has put them at a disadvantage to well-capitalized private firms when it comes to pursuing large-scale campus projects. 

But both Digital Realty and Equinix have long insisted that their biggest AI opportunity is not in this initial wave of demand for AI training. Rather, they anticipate that the adoption of AI technologies by corporations and consumers will spark a surge in demand for their colocation and connectivity business lines. 

As adoption increases, they say, the need for computing capacity shifts away from AI training and toward inference, the computing through which end users interact with an AI model. Inference workloads are often smaller and need to be located close to large concentrations of end users. Similarly, large corporations are expected to increasingly deploy their own AI computing in “hybrid” AI infrastructure models with major cloud providers. 

Executives at the two firms expressed similar beliefs that their distinct data center portfolios — with colocation capacity and strong interconnection to cloud providers in most major population hubs — leave them uniquely positioned to capture a key chunk of the emerging AI marketplace. 

This second wave of demand has been slow to emerge. It is only in recent months that AI has given significant lift to the REITs' demand picture beyond their hyperscale arms. Equinix reported this week that half of its 25 largest deals last year were related to AI, while 38% of the megawatts Digital Realty signed during the fourth quarter were AI-related. 

“We’re certainly starting to see our fair share of AI coming to the core markets, coming to enterprises, certainly coming to inference,” Digital Realty CEO Andy Power said. “But I believe we're still at the tip of an iceberg here.”

Still, there has been widespread anticipation across the data center sector that enterprise AI and inference demand will pick up significantly throughout 2025.

Now, Digital Realty and Equinix executives believe the emergence of DeepSeek is likely to accelerate this transition, with Power citing “a potential avalanche of AI inference demand we anticipate around the world.”

Power joins a chorus of prominent voices, including Meta’s Mark Zuckerberg and Microsoft CEO Satya Nadella, suggesting that if AI training becomes significantly cheaper, it will subsequently drive down the cost of adoption by enterprises and consumers and produce a more vibrant AI economic ecosystem. This would mean more AI startups, more companies incorporating AI into their existing businesses and more people using AI integrated into consumer products. 

As this phenomenon, known as “Jevons paradox,” emerges, it will both ramp up enterprise colocation demand and dramatically accelerate the shift of demand away from AI training and toward inference, according to Equinix CEO Adaire Fox-Martin.

“The drop in inferencing costs that's implied by the work released into the open-source market by DeepSeek will enable the economics of AI transformation to become a little bit more feasible for a broader set of organizations,” Fox-Martin said on the REIT's earnings call Wednesday. “You can see the absolute relevance of Equinix to this market opportunity.”

Executives at both firms said they are pursuing development strategies designed to capture this shift in demand as it accelerates. 

Overall, Equinix now has 62 major projects underway in 36 metros across 25 countries, with 16 of those being hyperscale campus developments. Digital Realty has 644 megawatts under construction, and it says it grew its pipeline by 75% last year. 

“The demand from inference, it will still have more and more proximity to the end consumer,” Digital Realty's Sharp said. “I think that's the important piece that we always look at, and where we apply our capital is that long-term durability of where that inference matures, because that's where the actual consumption or monetization of AI will happen, and that's why we're very excited about how that could be maturing over time.”

Rather than chasing available power into new geographies, both Equinix and Digital Realty point to a strategic focus on bringing contiguous blocks of new capacity into the most constrained markets, allowing them to provide colocation and connectivity as well as hyperscale capacity for customers across different key digital infrastructure hubs. 

Equinix has been particularly deliberate in pursuing this one-stop-shop approach.

Rather than pursue dozens of projects targeting various use cases scattered across global markets, Equinix announced a new development strategy last year that consolidates new data center construction on large-scale campuses near the largest global data center hubs. The campuses are planned to be scaled up over time and include an array of different kinds of data centers serving Equinix’s various business lines, with single-tenant hyperscale data centers, multitenant retail colocation facilities and other use cases all located at the same site.

Fox-Martin said this development framework positions Equinix to be the beneficiary as the looming surge in inference computing pushes demand back into core markets where capacity is harder and harder to come by. 

“We absolutely are pursuing that opportunity with a vigor … we're leaning into those capacity constraints that we see in the market,” she said.  “We are very firmly in the center of what is a very hot demand market at the moment.”