Handling Data Center Demand Risk

By Branko Terzic

In the restructured case the local electric utility is obligated to extend transmission or distribution level facilities to the new AI data center. The new data center must then procure its own power supply from the competitive market of merchant generators, or it must build its own generation obtaining all the required permits and authorizations.  Historic public utility line extension policies would require the date center to fund the line extension for costs beyond those in the embedded cost of service underlying the applicable tariff.  So, panicked claims aside, there’s no consumer impact there.

In the case of construction of a new AI data center in the franchise territory of a traditional vertically integrated utility, the rate regulators (utility commissions) makers have multiple tools at hand to ensure that residential consumers are not affected unreasonably by a future data center closure. The line extension policies would be the same in both cases.

The question of obtaining new generation capacity can be handled in multiple ways. The data center may have the opportunity to build its own dedicated facility much as many industrial customers and factories have done in the past. In that case the utility would provide only backup power at fair and appropriate rates.

In the case of the utility making an investment in new generation intended in large part to serve AI centers in its territory, the utility would need to file an application to the regulator (the state Public Service Commission (PSC) or Public Utility Commission (PUC)) for a certificate of Public Convenience and Necessity (CPCN) for the construction of a new power plant. In those proceedings, that state PSC would determine which customers would benefit from the new power plant and design rates under the “cost causer is cost payer” principle.

A risk that the new data center would have a limited life could be accommodated with power plant depreciation rates matched to the same life estimate if other customers would not need the capacity at the time the data center closes.

The late Joseph Brennan, a cost of capital expert witness, proposed that different customer classes could have different cost of capital depending upon the risk that the specific customer or customer class might not remain a customer of the system as long as another customer.  For example, Brennan’s point was that if a utility’s generation plant lost 30% of its load because an auto manufacturer moved overseas, then a fairer solution for the regulator might be to allow the utility a higher return on equity in rates for that fleeting customer because of the built-in risk. That has obvious ex ante-ex post considerations, but the challenge is not new, and regulators have seen these problems before.

In the restructured market case – whereas we said, the utilities cannot own generation and had to divest their existing generation to merchant suppliers – any new generation capacity, built by those merchant suppliers in anticipation of AI data center demand, becomes “extra capacity” (and system reserve capacity).  If the AI centers do not buy up this power, the excess power generated from it can be marketed by the merchant supplier (as a price taker) to other customers in the ISO market if that AI demand does not manifest. Or the merchant supplier can shut the generation down and incur a loss until market conditions improve. A moment’s thought should suggest that this extra competition should be good for the market. The investors in merchant generation are likely aware of the risk and have made their investment with that consideration.  The end users – the residential, commercial, and industrial customers – are shielded from the costs of excess supply.

In the case of the traditional vertically integrated utility, let’s suppose the utility is approved to build new generation capacity for AI data centers. The reasonable costs of this new generation are included in the utility’s rate base.  Thereafter, if the AI demand proves illusory the state utility regulator has a number of options. One option for the state regulator is just to recognize that the system now has more “reserve capacity” to meet growth and contingencies.  The Public Utility Fortnightly magazines of the 1980’s discussed the current national situation of reserve capacities in the 30% range. At the time there were numerous studies by EPRI and others showing that the costs of carrying additional reserve capacity was more economic than the cost of extended outages. This may be truer today than it was then. Today, outages in systems with thin reserve capacity margins from power plant retirements and the low system inertia from intermittent generation create major headaches for regional system operators and customers.

Another option for the state regulator is to find that some generating capacity of the regulated utility is no longer “used and useful” or should be treated as “stranded investment.” Although the reason why the excess generation came about is an important consideration in any proceedings, both situations have well known regulatory treatments that may either eliminate or minimize customer impacts.

In sum, 1) electricity markets can work and 2) utility regulators have the tools needed to protect consumers from utility investments in rate base which do not provide service.