In the ever-evolving landscape of artificial intelligence, Texas has emerged as a key player, attracting numerous large-scale data centers to its remote areas. These facilities are crucial for the intense computing required in AI training. However, a shift is happening as the next wave of AI development moves closer to urban centers.
Dallas-based DataBank is seizing this opportunity by constructing data centers closer to densely populated areas, enabling quicker delivery of AI models to users. This strategic move aligns with the company’s recent announcement of securing $2 billion in construction financing from a consortium of banks led by Mitsubishi UFJ Financial Group. The funds will be used to develop three data center buildings in Red Oak, a suburb located 20 miles from Dallas.
Additionally, MUFG is spearheading efforts to raise an extra $600 million for a fourth data center building at the same site. This financing will be sourced from the private placement market, targeting Wall Street investors.
DataBank’s Expansion Plans
The planned four-building complex, offering a total computing capacity of 240 megawatts, has already attracted a major hyperscaler as its tenant. Although DataBank CEO Raul Martynek did not disclose the tenant’s identity, he mentioned it could be one of the tech giants like Amazon, Google, Microsoft, Oracle, or Meta. Martynek noted that the project’s success underscores the growing need for inference facilities that enhance user interaction with AI by providing proximity and high-speed networking capabilities.
Inference computing, responsible for processing prompts in AI models, is gaining traction in the data center market. According to JLL, inference computing workloads are projected to rise from 9% in 2025 to 37% by 2030, while training workloads will slightly decrease from 14% to 13%. JLL’s Carl Beardsley emphasized, “The next phase is once your models have learned, and they’ve trained up, to convert that over to inference and be closer to the population center.”
Lending Challenges Amid AI Demand
Despite the escalating demand for AI computing facilities, lenders are exercising caution due to substantial debts linked to major AI and data center projects. Earlier this year, Business Insider reported that JPMorgan Chase and MUFG faced difficulties in selling portions of $38 billion in debt associated with two Oracle-occupied data center campuses.
Martynek acknowledged that the slowdown in the syndication market influenced DataBank’s financing timeline for its Red Oak project. As a result, the loan for the fourth building was separated from the syndication process and will be financed through a private placement.
The Red Oak campus is part of a larger plan to develop an eight-building, 480-megawatt data center facility. DataBank aims to complete the first four buildings by late 2027, with the initial building expected to be operational by the third quarter of 2026.
