The physical supply chain behind artificial intelligence (AI) is proving to be a lucrative source of revenue, with the demand for data centers skyrocketing. As AI continues to gain prominence, the need for chips, data centers, and other components has surged, leading to a boom across the entire supply chain.
Chips, designed by companies like Nvidia and AMD and manufactured primarily by Taiwan Semiconductor Manufacturing Company, require various parts and suppliers to be organized and connected. These components are then housed in specialized buildings designed to accommodate the unique requirements of AI technology.
Data centers, which play a crucial role in supporting AI, have seen a significant surge in demand. Companies like Amazon, Google, and Microsoft dominate the cloud capacity market, but new players are entering the scene due to the rise of AI cloud computing. DataBank, a leading player in the data center industry, leases out space at over 65 data centers in 27 markets, providing valuable insights into the trends shaping Big Tech.
DataBank CEO Raul Martynek, who has witnessed previous boom cycles in the tech industry, shared his perspective on the demand for AI computing power, hardware challenges, and the question of return on investment in AI.
Over the past 18 months, the demand for AI data centers has experienced a significant upswing. The introduction of ChatGPT in November 2022 triggered a surge in demand, with hyperscalers and major technology companies accelerating their AI initiatives. The sale of GPUs by companies like Nvidia directly correlates with the need for data center capacity, creating sustained demand that has outpaced supply for the past 12 months. As a result, data center pricing has doubled, and vacancy rates have dropped to low single-digits.
Planning for data centers involves thinking years in advance. Martynek explained that while he focuses on meeting the current year’s budget, he also has to consider the next 24 to 36 months. DataBank is currently developing over 800 megawatts of data center capacity, with plans for completion between 2025 and 2028. Pre-leasing, a common practice in the industry, has extended from 12 to 24 months due to the scarcity of data center capacity caused by the AI boom.
Fluctuations in the AI hardware market, such as potential delays or cancellations of chip models, have minimal impact on DataBank’s planning. The company can adjust the pace of bringing capacity online based on customer demand. While some customers may be annoyed by hardware disruptions, others appreciate the opportunity to maximize the useful life of their chips.
The question of return on investment in AI remains unanswered, as it is still early in the technology’s adoption. Martynek compared AI’s transformative potential to that of the internet or mobile phones, suggesting that its true impact may not be fully realized for several years. DataBank focuses on the credit quality and financial stability of its customers, ensuring they can effectively utilize the sophisticated and expensive AI technology.