In an exclusive conversation with Africa Hyperscalers, Raul Martynek, CEO of DataBank, the largest data center company in North America, discussed the surging demand for data centers propelled by the increasing need for Artificial Intelligence applications and the accompanying surge in the construction of new data center facilities.
According to Martynek, power density is a critical consideration in our designs. Over the years, as equipment has become more powerful and dense, power to the rack has increased from 5KW to approximately 15KW per rack. traditional data center designs coped well, but AI introduces a unique challenge, especially with power-hungry GPUs like Nvidia’s. “For instance, an H100 Tensor Core, which is about 7 to 8u high, will by itself consume 10KW per rack. Customers deploying GPU clusters for AI applications, such as large language models, are asking for up to 70KW per rack, and certain data center designs cannot support them”, he stated.
He stressed that while current data center infrastructure supports the world’s needs, AI workloads demand newer, specialized designs. Retrofitting older data centers for AI is challenging. Customers now seek high power in a small footprint, necessitating innovative cooling techniques like water cooling, over air cooling.
Martynek highlighted how data center infrastructure is running the world, supporting the many applications today, and legacy infrastructure will remain relevant, “Legacy workloads can focus on digital transformation, cloud and technology adoption which will drive data center growth”, he said.
Describing the data center boom in the US, surging from 100 megawatts in the mid-90s to a current 17 gigawatts, Martynek provided an explanation for the growing demand “The data center boom emerged after the internet and can be attributed to three key drivers: Enterprise Outsourcing: Organizations outsourcing their data centers to third-party providers; Public Cloud Growth: The dramatic growth of public cloud services, fueling demand for data centers as public cloud providers outsource their data centers to companies like DataBank and Technology Adoption: The number of applications available today requiring data center support has driven data center adoption. Now, with the advent of AI, we are witnessing unprecedented demand for data center capacity, particularly driven by AI’s adoption and impact.”
Martynek also spoke about its recent $533m raise and its plans to tap into the data center boom in the US. “This year has been particularly successful for us, with AI emerging as a significant driver in the US market. We are currently undertaking expansions in 15 different markets, ranging from a small 2MW expansion to a substantial 40MW build, contributing to over 180MW of new data center capacity.”
DataBank operates the largest geographic footprint of data centers in the US, boasting over 70 facilities across 26 markets, covering more than 2 million square feet. Its centers serve over 3000 customers, with diverse purposes, including data center interconnects functioning as carrier hotels, 18 peering data centers, enterprise-class facilities, and master-planned, purpose-built centers catering to critical IT gear on a single-tenant basis. Please watch the full interview here.
