Data Centers – analysis by Mark J. Welch (Aug 2026)
Enron. That’s the most important keyword when evaluating AI data centers. https://en.wikipedia.org/wiki/Enron_scandal
How many of the mega-datacenters are owned, operated, controlled and contracted entirely through Enron-style style LLC & REIT shell companies, which could default on all environmental, power, water & community commitments?
Do not assume nor misrepresent names of “ultimate customer(s).” It is critical to understand that all the big-tech companies are customers of each other, as well as investors, guarantors, and competitors of each other. (Facebook/Meta Apple Amazon Nvidia Google/Alphabet Microsoft Oracle OpenAI Anthropic etc.). The US government now also holds equity in some of these companies.
Who is promising what, to whom? Who is: the legal property owner; tenant; sub-tenants; customer(s); lenders; guarantors; insurers. What happens when someone attempts to shift ownership/control/liability to different entities?
No government officials or employees should sign any non-disclosure agreement (NDA). Do not allow the applicant/operator to conceal information as “proprietary.”
What insurance & bond requirements, legal liability, hold-harmless demands? Who can enforce?
What happens if facility fails to renew bond, insurance? What if the facility defaults on payment of utility bills, taxes, etc.?
Government must clearly ask about all infrastructure requirements for the intended use of the property, during planning + construction (phases) + startup + operation. Overhead wiring, trenching for pipes and conduit, water supply and discharge, drainage, grading, permanent and temporary roadways for construction.
What services are required but conditional?
Anticipate shutdown/closure options (recycling, ewaste, etc.)
What specific monitoring and supervision is required?
Some other documents to review
“Local Guidelines for Data Center Development” (Urban Land Institute Data Center Product Council; Hannah Miet)
- p.7-8: Reasonable explanation of the difference between data center types (corporate/enterprise, colocation/retail, wholesale, telecom, and hyperscale).
- p.15: “Power purchase agreements (PPAs): long-term contracts with renewable energy providers (the sellers) to purchase electricity from specific wind or solar farms.”
- p.17 and 21: Sound mitigation
- p.19: “Regulations developed for office or warehouse uses often impose excessive parking and plumbing fixture counts on data centers, which necessitate variances and complicate the planning process.” Yes, but: what if the property is later converted to a different use?
- p.19: Fire Department considerations: Battery fires? Diesel fuel storage; generators.
- p.26: Glossary of terms.
“AI Infrastructure Community Evaluation Framework (AICEF)” (appears to be a crude AI-generated template; the checklist on page 12 might be useful).
https://www.lindenwood.edu/files/future-caucus/ai-community-evaluation-aicef3.pdf
Mark J. Welch is a former journalist, attorney, teacher, and data center salesperson. From 2011-2021, Mark worked for Hurricane Electric, which converted two former factories in Fremont, California into colocation data centers. HE is also one of the world’s largest global internet bandwidth providers, selling connectivity globally at 320 major exchange points and at telecom data center in more than 200 cities worldwide. (https://pop.he.net/)
