📌 AI Data Center Expansion Faces a Test from Financing Costs
Bank of America expects global AI data center capacity to reach about 219 gigawatts by 2030, roughly 5 times its current level. Among the additional capacity listed in the forecast, Amazon accounts for 42 gigawatts, Google for 27 gigawatts, and Microsoft and SPCX for 24 gigawatts each. The expansion plans are substantial, but capacity forecasts reflect expectations for construction and do not mean that all these projects will generate sufficient returns.
Another analysis views the cost of capital as a potential bottleneck to expansion: the four major cloud providers’ capital expenditure plans for 2026 amount to up to about USD 720 billion, and banks are reportedly also preparing about USD 60 billion in financing to support Broadcom-related AI equipment. Meanwhile, the 10-year U.S. Treasury yield has exceeded 5%, which could raise data center project financing costs and the hurdle for recouping investments. Interest rates are only one part of project costs; the information currently available does not give the financing rates, utilization rates, or returns of specific projects, so it is not yet possible to judge whether the investments are worthwhile.
Therefore, whether the pace of expansion can be sustained depends not only on how much companies are willing to invest, but also on whether the additional capacity can generate enough revenue to cover construction and financing costs. The related forecasts and financing figures come from social media reports and still need to be verified against original reports and company disclosures.
▌ Sources