Generative AI Software’s Hidden Cost Is Big Paperwork

The Securities and Exchange Commission finally voted into place its long-expected greenhouse gas disclosure rule.

Technology will be key for tracking, analyzing, and reporting emissions. Maybe generative AI programs like ChatGPT — if not for solving the problem, then maybe for contributing to it because of the amount of electrical power the companies running the systems consume.

“Larger commercial real estate companies that are public to start, and the commercial REITs, as well as potentially the REITs that own interest in commercial real estate, whether in CMBS or another format,” are likely to be immediately responsible for additional reporting, disclosures, and financial statement requirements, Anna Pinedo, a partner with law firm Mayer Brown, recently told GlobeSt.com.

Furthermore, smaller CRE firms outside the legal requirements may find themselves having to satisfy larger tenants who do need to disclose greenhouse gas information and will want it from the landlord. So, even the smaller CRE firm might need some extensive data to pass along.

That brings us to generative artificial intelligence. The technology has suggested a lot of promise for CRE businesses. It still is in early phases. “From a generative AI perspective in real estate, we’re barely scratching the surface,” Ram Srinivasan, managing director of the global consulting group within JLL, told GlobeSt.com earlier this year. “There’s a lot of buzz around it and a lot of investment dollars chasing it, and a lot of executive attention on it.”

The technology is typically provided by vendors, often through a hosted services. And it uses a lot of electricity, which means greenhouse gas emissions. Alex de Vries, Ph.D. candidate at VU Amsterdam and founder of the digital-sustainability blog Digiconomist, has estimated that AI technology overall could come to annually consume as much electricity as the entire country of Ireland, as IEEE Spectrum reported. There is even some data suggesting that generation of one AI image can use as much power as it takes to fully charge a smartphone.

But many of the AI companies have become increasingly opaque over their activities. How will CRE firms get the data they need? How can the users verify the answers they receive?

These are some of the problems coming from the AI companies. Significant customers of the CRE firms will need data from a landowner’s power consumption, including what is used their reliance and use of AI.